Colby Hill

Week in Review: August 17–21 — What’s Next for Investors

Colby Hill — Sun, 23 Aug 2026 14:04:05 GMT

In This Issue

Full Recap of Last Week
Upcoming Earnings Week
Macro Event Calendar
Macro Trends: Rates, Inflation & the Business Cycle
Market Volatility
Analyst Upgrades & Downgrades
Top 5 Big News Events
Capital Flows: Insiders, Whales & ETF Money

1. Full Recap of Last Week

The bond market ran the show last week, and stocks paid the price. All three major indexes finished lower even after a strong Friday bounce.

S&P 500: closed at 7,674.37, down 111.39 points or 1.4% for the week
Nasdaq Composite: closed at 26,180.45, down 548.71 points or 2.1% — the weakest of the three
Dow Jones Industrial Average: closed at 53,277.01, down 455.40 points or 0.8%, its second straight weekly decline

The trouble started midweek when the 30-year Treasury yield pushed to its highest level in nearly two decades. Rising long-term yields hit the most expensive parts of the market hardest, which is why the tech-heavy Nasdaq lagged. Friday brought relief — the Dow jumped 518 points after a survey showed U.S. business activity growing at its fastest pace in more than four years.

Earnings results: It was retail week, and the read on the American consumer was “resilient but picky.” Home Depot reported Tuesday, Target and Lowe’s Wednesday, and Walmart Thursday. Walmart was the disappointment, posting its weakest sales growth in more than six years and missing on U.S. comparable sales — shares dropped sharply.

Target and Home Depot both delivered sales gains and described shoppers who will still spend when the price is right. Off-price retailer Ross Stores was the bright spot, jumping 8.1% on a better-than-expected quarter. Moderna surged 17% on positive Phase 3 results for its personalized mRNA cancer vaccine developed with Merck.

Zooming out, Q2 2026 has been an unusually strong earnings season. As of FactSet’s late-July update, 86% of reporting S&P 500 companies had beaten EPS estimates — well above the 5-year average of 78% and the 10-year average of 76%.

Macro results: The data was mostly solid. The Philadelphia Fed manufacturing index surged to 47.4 against a consensus of 25, and the NY Empire State index came in at 20.6 versus 11 expected. Initial jobless claims fell to 206,000, better than the 210,000 forecast. Housing was the weak spot — July housing starts fell 12.4% month-over-month to 1.239 million, badly missing the 1.35 million estimate, though building permits rose 5%.

The big macro event was Wednesday’s release of the July FOMC minutes, which revealed a far more hawkish Fed than markets had assumed. More on that in Section 4.

Elsewhere: WTI crude rose 4.76% on the week to finish near $86.75, with Brent above $93. Gold surged 3.12% to roughly $4,637 an ounce. And Bitcoin had its best week in two years, soaring more than 17% to around $77,000.

2. Upcoming Earnings Week (Aug 24–28)

Q2 earnings season is in its final stretch, but this week saves the biggest name for last. Here’s the day-by-day:

Monday 8/24: A quiet start. PDD Holdings and Napco Security Systems report before the open. Nothing noteworthy after the close.
Tuesday 8/25: Bank of Montreal, Bank of Nova Scotia and Dick’s Sporting Goods before the open. After the close: Intuit, Zoom Communications, HEICO, Box, nCino and Semtech.
Wednesday 8/26: The main event. Before the open: Abercrombie & Fitch, Bath & Body Works, Kohl’s, Williams-Sonoma, J.M. Smucker, Li Auto and Dycom. After the close: NVIDIA, Salesforce, CrowdStrike, HP, Synopsys, Okta, Agilent, Nutanix, Veeva Systems and Urban Outfitters.
Thursday 8/27: Before the open: Best Buy, Dollar General, Dollar Tree, Burlington Stores, Toronto-Dominion Bank, Hormel Foods and Bilibili. After the close: Marvell Technology, Workday, Autodesk, Ulta Beauty, Affirm, Gap, Rubrik, SentinelOne and Elastic.
Friday 8/28: No noteworthy reports scheduled.

What to watch: NVIDIA on Wednesday afternoon is the week’s single biggest catalyst. Wall Street is looking for year-over-year earnings growth of roughly 99% to $2.09 per share on revenue growth of about 97% to $92.07 billion for its fiscal Q2 2027. Susquehanna’s Christopher Rolland has a Positive rating and $275 target, pointing to hyperscaler capex that is expected to nearly double in 2026.

Marvell on Thursday is the other AI bellwether — and an interesting contrast. Marvell has returned about 180% year-to-date through August 20 versus roughly 17% for NVIDIA. Stifel expects a beat, looking for revenue near $2.71 billion and holding a $350 target.

Also worth noting: this is a heavy week for the mid-tier consumer. Dollar General, Dollar Tree, Best Buy and Gap will round out the retail picture that Walmart and Target started last week.

3. Macro Event Calendar (Aug 24–28)

A back-loaded week, with everything that matters landing Wednesday and beyond.

Monday 8/24: No major U.S. releases.
Tuesday 8/25: Conference Board Consumer Confidence for August. Given how much last week’s retail earnings focused on the value-seeking consumer, this one carries more weight than usual.
Wednesday 8/26: The big one. Core PCE Price Index for July — the Fed’s preferred inflation gauge — plus personal income, personal spending, durable goods orders and the second estimate of Q2 GDP. All at 8:30 a.m. ET.
Thursday 8/27: Weekly initial jobless claims. Jackson Hole Economic Symposium begins.
Friday 8/28: Jackson Hole continues. The preliminary annual benchmark revision to non-farm payrolls is also scheduled.

On Jackson Hole: The Kansas City Fed’s annual symposium runs August 27–29 in Wyoming under the theme “Financial Innovation: Implications for Payments and Policy.” The agenda leans toward digital payments, stablecoins and tokenization rather than near-term rate policy, and Chair Kevin Warsh has consistently avoided explicit forward guidance.

Historically, though, this venue has produced surprises — Powell used it in 2020 to announce flexible average inflation targeting and in 2022 to push back hard on rate-cut expectations. With the September FOMC meeting less than a month out, markets will be listening even if the bar for a real signal is high.

On core PCE: June’s reading was encouraging — core PCE rose just 0.1% month-over-month after three straight 0.3% gains, pulling the annual rate down to 3.3% from 3.4%. Consensus expects a 0.2% monthly gain in July, leaving the annual rate at 3.3%. A hotter print raises the odds of a September hike; a softer one buys the Fed patience.

4. Macro Trends: Rates, Inflation & the Business Cycle

Rates. This is the single most important thing to understand right now: the market is pricing in rate hikes, not cuts. The Fed has held the federal funds rate at 3.50%–3.75% for five consecutive meetings. Futures ended last week pricing roughly 9 basis points of tightening for the September FOMC meeting and about 23 basis points cumulatively for the rest of 2026.

Wednesday’s July minutes explained why. The FOMC voted 9–3 to hold, its most fractured vote in years, with Cleveland’s Beth Hammack, Minneapolis’s Neel Kashkari and Dallas’s Lorie Logan all dissenting in favor of a quarter-point increase.

The hawkishness ran deeper than the dissenters — the minutes noted that “many participants” believed further tightening would likely be necessary if inflation didn’t come down. Kansas City’s Jeffrey Schmid and St. Louis’s Alberto Musalem, neither of whom voted in July, both later indicated they would have backed a hike.

Chair Warsh remains the patient voice, emphasizing data dependence and repeating that the Committee is “not constrained by market prices.” But he’s clearly leading a divided committee.

Separately, the long end of the curve is doing its own thing. The 30-year Treasury yield touched a near-20-year high last week, driven by a combination of inflation worries, fiscal concerns and a heavy calendar of AI-related corporate debt issuance. Treasury Secretary Scott Bessent responded by doubling the size of the department’s long-term debt buybacks from $2 billion to at least $4 billion per operation.

Inflation. Headline PCE is running at 3.7% annually and core PCE at 3.3% — both meaningfully above the Fed’s 2% target. The bigger near-term problem is energy. WTI crude is above $86 and Brent above $93, up two consecutive weeks on U.S.–Iran tensions. Rising fuel costs squeeze consumers and retailers simultaneously, which is exactly what showed up in last week’s retail earnings commentary.

Business cycle. The picture is genuinely mixed, which is what makes this moment tricky. Manufacturing surveys are booming (Philly Fed at 47.4 is a very strong number), the labor market is holding up (claims at 206,000), and business activity is growing at a four-year high.

But housing starts collapsed 12.4% in July, consumers are trading down, and corporate profit growth — while excellent — is increasingly concentrated in AI-adjacent names. In short: an economy strong enough to keep inflation sticky, with pockets of real weakness underneath.

5. Market Volatility

Volatility ticked higher last week but stayed well short of panic levels.

VIX: closed the week at 16.02, up from 14.26 the prior week. That’s a meaningful jump in percentage terms but still a historically calm absolute reading. For context, sustained readings above 20 typically signal genuine market stress.
CNN Fear & Greed Index: slipped into neutral territory around 52–55, down from roughly 57 earlier in the week. Sentiment cooled as bond yields climbed, but investors are neither euphoric nor fearful.
Options positioning: The Fear & Greed put/call component contributed to the move toward neutral, consistent with modest hedging demand rather than aggressive downside protection buying.

The takeaway: last week’s selloff was a repricing driven by rates, not a fear event. Equity volatility stayed contained even as bond volatility spiked — a divergence worth watching. If long-end yields keep climbing and the VIX starts catching up, that’s when the character of this pullback changes.

6. Analyst Upgrades & Downgrades

Notable Upgrades

Apple (AAPL) — Rothschild & Co Redburn raised it to Buy from Neutral and lifted its price target dramatically to $400 from $260. One of the more aggressive target hikes of the week.
Okta (OKTA) — Wells Fargo moved to Overweight from Equal Weight, target to $180 from $150. Timely, with Okta reporting Wednesday.
Astera Labs (ALAB) — Northland upgraded to Outperform from Market Perform with a $350 target, an AI-infrastructure call.
Duolingo (DUOL) — DA Davidson raised it to Buy from Neutral with a $160 target.
Equifax (EFX) — Rothschild & Co Redburn upgraded to Buy from Neutral, target to $235 from $214.
Ternium (TX) — Morgan Stanley moved to Overweight from Equal Weight, target to $65 from $55.

BMO Capital also initiated coverage on the AI complex, starting NVIDIA at Outperform with a $340 target, Broadcom at Outperform with $455, and Marvell at Outperform with $250.

Notable Downgrades

Teradyne (TER) — Baird cut it to Neutral from Outperform with a $420 target, citing valuation and a lack of major growth catalysts until 2028. Shares fell nearly 4% Friday.
York Space Systems (YSS) — JPMorgan downgraded to Neutral from Overweight and cut its target in half, to $17 from $38. The harshest call of the week.
Ameresco (AMRC) — Baird moved to Neutral from Outperform, target to $32 from $36.
Weibo (WB) — Citigroup cut to Neutral from Buy with a $7.80 target.
Frontline (FRO) — Danske Bank downgraded to Sell from Hold with a $39.15 target.

Also notable: Piper Sandler cut AppLovin’s price target to $325 from $385 while keeping a Neutral rating, questioning the sustainability of its growth after Q2 results. And DZ Bank initiated SpaceX at Sell with a $100 target, warning of “crash risk in the valuation orbit” — the stock debuted at $135 in June and peaked at $225.64 days later.

7. Top 5 Big News Events

1. The 30-year Treasury yield hit a near-20-year high. This was the story of the week and the direct cause of the equity selloff. Long-term borrowing costs at these levels challenge every valuation model built on cheap money, which is why the Nasdaq took the worst of it. Inflation fears, fiscal concerns and a wave of AI-related debt issuance all contributed.

2. Treasury doubled its bond buyback program. Secretary Bessent announced the department would at least double long-term debt buybacks to $4 billion per operation, and hinted at more. He also discussed currency interventions involving the yen and peso plus new fiscal consolidation plans. The immediate market reaction was muted for stocks but explosive for Bitcoin.

3. Fed minutes revealed a deeply divided, hawkish committee. A 9–3 vote with three regional presidents dissenting for a hike, plus “many participants” seeing further tightening as likely if inflation doesn’t cool. This meaningfully shifted the rate outlook toward tightening.

4. Walmart’s miss reframed the consumer story. The world’s largest retailer posted its slowest sales growth in over six years and missed on U.S. comps. Combined with Target’s and Home Depot’s “resilient but value-focused” commentary, the picture is of a consumer who hasn’t quit but is clearly under pressure from inflation and high fuel costs.

5. U.S.–Iran escalation pushed oil past $90. President Trump threatened an “economic D-Day” on Iran, with a blockade in place and tougher sanctions on Iranian oil exports threatened. Iran called its potential response “devastating” and finalized a preferential trade deal with Oman. Brent finished above $93, its second straight weekly gain — and energy prices feed directly back into the inflation problem in item 3.

Honorable mention: Bitcoin’s 17% weekly surge to roughly $77,000 — its best week in two years — lifted Robinhood 11.1%, Coinbase 8.5% and Strategy 7.6%. Meanwhile, Meta shares fell sharply as four states pursued up to $1.4 trillion in penalties over alleged consumer protection and children’s privacy violations.

8. Capital Flows: Insiders, Whales & ETF Money

ETF flows. The dominant theme is a rotation toward defense and yield. Bond ETFs gathered $13.2 billion in a single recent week, with mixed-maturity funds pulling in $6.9 billion and roughly $193 billion in cumulative year-to-date inflows. U.S.-listed ETFs overall attracted about $41 billion in the week ending August 14.

On the equity side, flows have been choppy — the SPDR S&P 500 ETF Trust (SPY) saw large outflows in early August while commodity funds like GLD and SLV drew strong inflows. Most S&P sector ETFs saw net inflows recently with technology and energy the notable exceptions. Translation: money is moving out of the crowded growth trade and into bonds and hard assets.

Whale watching. Bill Ackman’s Pershing Square drew down its Alphabet position this quarter, taking profits. Worth noting given how consensus the mega-cap tech trade has become.

Notable insider selling. August has seen heavy selling at AI-adjacent names. Magnetar Financial, a 10% owner and director at CoreWeave, sold 2,133,481 Class A shares on August 12 for roughly $231.6 million. CoreWeave CEO Michael Intrator sold about $27.4 million of stock on August 11. Arista Networks director and 10% owner Andreas Bechtolsheim sold 300,000 shares on August 5 for approximately $61 million. DST Global Advisors sold roughly $33.4 million of Chime Financial across August 11–12.

Notable insider buying. Thinner, as usual. Walters Group, a 10% owner and director at Zentalis Pharmaceuticals, bought 4,335,000 shares on August 14 for $15.17 million. Opendoor Technologies CEO Kasra Nejatian purchased 27,625 shares on August 14 for just over $100,000.

The pattern is worth flagging: insiders at the highest-flying AI infrastructure names are taking money off the table in size, while buying is concentrated in smaller, beaten-down names. Insider selling has many innocent explanations — diversification, tax planning, pre-scheduled 10b5-1 plans — but the asymmetry between the two columns is notable.

The Bottom Line

Last week wasn’t about stocks — it was about bonds. A 30-year Treasury yield at a two-decade high, a Fed whose own minutes show it debating hikes rather than cuts, and oil above $90 all point to the same uncomfortable conclusion: the disinflation story has stalled, and the market is only starting to price that in.

Yet earnings are excellent, manufacturing is booming, and the VIX at 16 says nobody is panicking. That gap between a hawkish rate backdrop and calm equity markets is the tension to watch. This week resolves some of it — core PCE on Wednesday morning tells us whether inflation is cooling, NVIDIA on Wednesday afternoon tells us whether the AI capex cycle is still compounding, and Jackson Hole gives Chair Warsh a stage he may or may not use. Three catalysts, one week.

This newsletter is for informational purposes only and is not financial advice. Always do your own research before making investment decisions.

Subscribe now

HOOD: The Brokerage That Stopped Being a Brokerage

Colby Hill — Sun, 23 Aug 2026 11:47:11 GMT

August 23, 2026 | Deep Dive by Colby

OPENING HOOK

Why this story matters right now

Institutions added 209 million shares of Robinhood between May and August 2026. That is a 37% increase in institutional share count in a single quarter, in a stock that already had 2,247 institutional holders. Nothing about that is normal. Index funds do not move that fast. That is active money making a decision.

On Friday, August 21, the market caught up. HOOD closed at $108.13, up 13.7% on volume of 50.5 million shares — roughly 2.6 times its 30-day average. The stock opened at $101.10, dipped to $98.77 in the first hour, and then never looked back, grinding to a high of $109.71 before settling near the top of the range. That is not a gap-and-fade. That is accumulation.

Here is a number worth sitting with. In the second quarter of 2026, Robinhood generated $156 million from event contracts — prediction markets. A year earlier that line was worth roughly $15 million. It grew more than tenfold, and for the first time in company history, prediction markets out-earned both equities and crypto. The company that Congress dragged through hearings over payment for order flow now makes more money letting people bet on outcomes than it does letting them buy Apple.

Robinhood Markets is a $97 billion company that most investors still price as a retail brokerage. It is not one anymore. It is a distribution platform that has decided the product is irrelevant — stocks, options, crypto, prediction contracts, private startups, tokenized securities — as long as 28.4 million funded accounts are pointed at it. Friday was the day the market started underwriting that thesis. The question is whether it should.

INVESTMENT THESIS

The core reasons to own this stock

One — the revenue mix has genuinely diversified, and the market has not repriced for it. Two years ago the bear case was simple: Robinhood is a levered bet on crypto volumes. That case is dead. Crypto revenue in Q2 2026 came in at $100 million, down 38% year over year, and now represents just 12.9% of transaction revenue. Total revenue still grew 32% to $1.308 billion. A company that grows a third while its former flagship product shrinks by nearly forty percent has proven something structural. The multiple has not caught up to the durability.

Two — prediction markets are a genuinely new revenue pool, and Robinhood owns the retail front door. Kalshi runs the exchange; Robinhood owns the customer. Roughly 87% of Kalshi’s $39.7 billion in trailing volume was sports-related, which means the addressable market is not political junkies — it is the same demographic that already lives inside the Robinhood app. That $156 million quarter came from a standing start.

Three — the private markets push converts an inaccessible asset class into a ticker. Robinhood Ventures Fund II (RVII) priced an 8 million share IPO at $25, raising $225.5 million with up to $255.5 million on the greenshoe, and it holds stakes in 80 private companies with a Y Combinator early-stage tilt. Because it is structured as a business development company trading on the NYSE, there are no accreditation requirements and no minimums. Robinhood took the single most exclusive product in finance and made it buyable with $25 and a phone.

Four — the balance sheet buys time that competitors do not have. Robinhood ended the quarter with roughly $13.67 billion in cash, a 57% adjusted EBITDA margin, and 23.6% return on equity. This is a growth story funded entirely by its own profitability. It can launch three funds, enter the UK, and build a tokenization stack without touching capital markets.

BUSINESS OVERVIEW

What Robinhood actually does

Robinhood Markets was founded in 2013 by Vladimir Tenev and Baiju Bhatt on a single idea: commission-free trading delivered through a mobile app good enough that people would use it for fun. Today it operates out of Menlo Park with roughly 2,900 employees and serves 28.4 million funded customers holding $369 billion in total platform assets, up 32% year over year.

The revenue model has three legs. Transaction-based revenue — the largest at $776 million in Q2, up 44% — comes from order flow and spreads across options ($342 million, up 29%), event contracts ($156 million), equities, and crypto ($100 million, down 38%). Net interest revenue contributed $389 million, up 9%, earned on customer cash, margin lending, and securities lending. The third leg is subscription: Robinhood Gold now counts 4.8 million subscribers, up 39%, paying a recurring fee for higher interest on cash, bigger instant deposits, and research access.

That third leg matters more than its size suggests. Gold is the mechanism by which Robinhood converts a transactional relationship into a banking relationship. Net deposits of $21.7 billion in the quarter — a 28% annualized growth rate against beginning platform assets — tell you that customers are no longer treating the app as a casino they visit. They are moving their actual money there. Every dollar that arrives becomes float that earns net interest whether or not anyone trades.

INDUSTRY OPPORTUNITY

Three regulatory doors opening at the same time

The unusual thing about Robinhood in 2026 is that three separate regulatory regimes are loosening simultaneously, and the company has positioned itself in front of all three. Prediction markets came first. The CFTC reversed its posture on event contracts in early 2025, describing prior interpretations as a legal sinkhole, and by August 2026 the agency convened its first Innovation Advisory Committee under Chairman Michael Selig specifically to write durable rules. Tenev sat at that table. When a regulator moves from prohibition to rulemaking, the incumbent that showed up early usually writes the compliance template everyone else copies.

The second door is tokenization. The SEC is working on tailored rules for crypto contracts alongside an innovation exemption for tokenized securities. Robinhood already runs tokenized equity products in Europe. If the U.S. framework lands favorably, the company does not build a new business — it ports an existing one into its largest market. That is a meaningfully different risk profile than a competitor starting from a whitepaper.

The third is geographic. Crypto trading is now rolling out to eligible UK customers through Bitstamp UK, the FCA-registered entity Robinhood acquired. International expansion for a U.S. brokerage has historically been a graveyard, because the regulatory cost of each market exceeds the revenue. Buying a licensed local operator and plugging it into an existing app inverts that math.

COMPETITIVE ADVANTAGE

Why Robinhood might win

The moat is not the trading engine. It is that Robinhood is the only firm at scale whose customers will click on something new. Charles Schwab has more assets and better pricing on nearly everything, and its median customer would never touch an event contract on a Federal Reserve decision. Interactive Brokers has superior execution and a global footprint, and it markets to professionals who do not want a fourth tab in the app. Coinbase owns crypto-native users but has no equities business worth naming, which means every dollar of its revenue rides the same cycle. Webull and eToro have the demographic but roughly a tenth of the distribution.

Robinhood sits alone in the intersection: a young, high-engagement, mobile-first user base large enough — 28.4 million funded accounts — to make a brand-new product line commercially meaningful in a single quarter. The prediction markets result is the proof. A $156 million business did not exist eighteen months ago. No incumbent could have built that revenue that fast, because no incumbent has customers who would try it.

The second, quieter advantage is regulatory. Robinhood spent 2021 through 2023 absorbing the worst enforcement environment any retail broker has faced in decades. What it purchased with that pain is a compliance apparatus and a set of relationships that let it launch a BDC, a UK crypto venue, and a CFTC-regulated event contracts business inside twelve months. The scar tissue turned out to be the asset.

FINANCIAL DEEP DIVE

The numbers: what they mean and what to watch

Metric                     Value
-------------------------  --------------------
Last Close (Aug 21, 2026)  $108.13 (+13.7%)
Day Range                  $98.77 - $109.71
52-Week Range              $63.52 - $153.86
Market Cap                 $97.2 billion
P/E (trailing)             47.9x
P/B                        10.25x
Dividend Yield             None
Q2 2026 Revenue            $1.308B (+32% YoY)
Q2 2026 Net Income         $561M (42.9% margin)
Adj. EBITDA Margin         57%
Return on Equity           23.6%
Cash on Hand               ~$13.67 billion
Shares Outstanding         899.1 million
Avg. Volume (30-day)       19.4 million
Aug 21 Volume              50.5 million (2.6x)

The margin structure is the story inside this table. A 42.9% net margin and 57% adjusted EBITDA margin on a business with 2,900 employees is software economics attached to a financial services license. Compare the quarterly progression: Q2 2025 revenue of $989 million at a 39.0% net margin became Q2 2026 revenue of $1.308 billion at 42.9%. Robinhood grew revenue 32% and expanded margin by nearly four points at the same time. That is operating leverage working exactly as the bull case requires.

Now the uncomfortable part. At 47.9 times trailing earnings and 10.25 times book, this stock is priced for that to continue indefinitely. Trailing twelve-month net income runs roughly $2.07 billion, or about $2.30 per share. For the current multiple to hold, earnings need to compound at 25-30% for several years — and the single largest swing factor, net interest revenue at $389 million, is a rate-sensitive line growing at only 9%. If the Fed cuts aggressively, that line compresses and the growth arithmetic gets much harder.

Watch the 52-week range for perspective on how violent this name is. The stock touched $153.86 in October 2025 and $63.52 in March 2026 — a 59% peak-to-trough drawdown inside six months. Friday’s close of $108.13 still sits 30% below the high. Investors who bought the October top are down meaningfully despite everything described in this article going right.

MANAGEMENT TEAM

Who is running this?

Vladimir Tenev has been chief executive since January 2013 — a 13-year tenure that makes him one of the longest-serving founder-CEOs in financial technology. He now holds the Chairman, CEO, and President titles simultaneously, which is a governance concentration worth naming plainly: there is no independent chair checking the person setting strategy.

The alignment, however, is close to ideal. Tenev holds roughly 6.16% of the company, worth about $4.55 billion — more than a thousand times his $3.0 million annual compensation package, of which only 1.1% is salary. Nearly everything he earns is equity. When a CEO’s personal balance sheet is 99.9% correlated with the share price, the incentive to chase short-term revenue at the expense of franchise value is structurally weak.

Two items deserve scrutiny. First, Tenev sold 375,000 shares on July 6, 2026 for roughly $44 million. Against a $4.55 billion position that is well under 1% and consistent with routine diversification, but it happened three weeks before earnings and is worth tracking if it repeats. Second, Tenev is separately Executive Chairman and co-founder of Harmonic, an AI reasoning company. Divided attention at the top of a $97 billion company navigating three simultaneous regulatory transitions is a real risk, not a footnote. On the finance side, CFO Shiv Verma has been explicit that the private-markets effort is early — the two funds are staffed by three investment professionals with plans to hire more, and in his words, they are just getting started.

BULL CASE

5 reasons this stock could dramatically outperform

One — prediction markets compound from a small base into the largest transaction line. Event contracts went from roughly $15 million to $156 million in four quarters and already outgrossed equities and crypto. If the CFTC’s rulemaking under Chairman Selig produces a workable federal framework in 2027, the sports-betting demographic — which drove 87% of Kalshi’s $39.7 billion in volume — becomes fully addressable through a brokerage account rather than a sportsbook. A $600 million to $800 million annual run rate by 2028 is not aggressive under that scenario.

Two — U.S. tokenized securities go live and Robinhood is first. The company already operates tokenized equities in Europe. An SEC innovation exemption would let it port a working product rather than build one, and 24/7 tokenized trading of U.S. equities would generate spread revenue on hours when every competitor is closed. Watch for the SEC framework in the first half of 2027.

Three — the private markets flywheel becomes a fee business, not a marketing stunt. RVII raised $225.5 million at $25 per share holding 80 private companies. Management has said explicitly it intends to accelerate launch cadence. Each fund generates management fees, drives platform assets, and gives active traders another reason to stay inside the app. Three to four funds a year at increasing size turns a headline into an annuity.

Four — Gold subscriptions convert the business into recurring revenue. 4.8 million subscribers, up 39%, is a subscription base larger than most standalone fintech companies. Subscription revenue carries a far higher multiple than transaction revenue. If Gold crosses 8 million subscribers, the market will start valuing a meaningful slice of Robinhood on SaaS comparables rather than broker comparables — and that re-rating alone is worth more than any single product launch.

Five — the institutional bid has already turned. Institutions went from 568.5 million shares in May 2026 to 777.8 million in August, an increase of 209 million shares, or 37%, in three months. Fintel counts 2,247 institutional holders including BlackRock, Vanguard, State Street, Fidelity, Geode, and Susquehanna. Positioning of that magnitude ahead of a catalyst-rich twelve months creates a structural bid under the stock that did not exist in March at $63.

BEAR CASE

5 reasons this could disappoint

One — the valuation leaves no room for a single bad quarter. At 47.9x trailing earnings and 10.25x book, HOOD trades at a multiple that requires 25-30% earnings growth for years. The historical price-to-book mean for this business sits near 10, and in a genuine risk-off tape brokers derate toward mid-teens P/E in weeks, not months. The March 2026 low of $63.52 — reached only five months ago — is what that repricing looks like.

Two — net interest revenue is a rate bet nobody is pricing. $389 million of quarterly revenue, roughly 30% of the total, is earned on customer cash and margin balances and it grew only 9%. That is the slowest-growing major line in the business and it is entirely a function of Fed policy. A 200 basis point cutting cycle takes a material bite out of the highest-margin revenue Robinhood has, and no product launch offsets it quickly.

Three — prediction markets are one adverse ruling from a much smaller business. New York has already sued Kalshi. The state-versus-federal fight over whether sports event contracts constitute gambling is unresolved, and Robinhood accesses the market through KalshiEX LLC rather than owning the exchange. If a court or a state coalition forces sports contracts off the platform, the fastest-growing revenue line in the company shrinks dramatically and the growth narrative loses its centerpiece.

Four — crypto is still a meaningful drag and the yield war is real. Crypto revenue fell 38% to $100 million. Competitors including Galaxy’s GalaxyOne are pushing aggressive yield offerings that directly attack Robinhood’s net interest economics. Being forced to pay up on customer cash to defend deposits is the single most efficient way to compress a 57% EBITDA margin, and it does not show up until it does.

Five — the stock is a high-beta instrument regardless of fundamentals. A 59% drawdown from $153.86 to $63.52 in six months happened while the business was posting record results. Fintel’s own data notes the shares were at $107.94 in August 2025 and $95.10 in August 2026 — an 11.9% decline over a full year in which revenue grew more than 30%. Owning HOOD means accepting that operational excellence and share price can decouple completely for twelve months at a stretch.

VALUATION

Building the model: Bear, Base, Bull

Scenario  Key Assumption                           Target  Prob.
--------  ---------------------------------------  ------  -----
Bear      Rate cuts hit NII; derate to 25x         $58     25%
Base      Revenue +25%, EPS ~$2.85 at 42x          $120    50%
Bull      US tokenization live; EPS ~$3.40 at 55x  $187    25%

The probability-weighted expected value works out to roughly $121 per share — about 12% above Friday’s $108.13 close. That is a far more modest number than the 13.7% single-day move implies, and it is the honest read. The Street broadly agrees: 28 analysts carry a consensus Buy with an average target of $119.82, Goldman Sachs raised its target to $123 on Friday, and the FactSet mean sits at $124.73. The dispersion is what matters more than the midpoint — the low target on the Street is $57 and the high is $163.60, a spread of nearly 3x.

Read that spread carefully. It means the analyst community is not disagreeing about execution; it is disagreeing about which company this is. If Robinhood is a cyclical brokerage with a good year, $58 is defensible. If it is a distribution platform monetizing four new asset classes, $187 is defensible. The base case says the truth lands closer to the platform, but that the market has already paid for most of the first leg of that transition. You are being offered roughly $79 of upside against $50 of downside — better than even, but not a layup.

KEY METRICS TO WATCH

The five signals that tell you if the thesis is working

01 — Event contracts revenue. Q2 2026 printed $156 million. Green flag is sequential growth above 20% in Q3, confirming the line is a business and not a one-quarter political-cycle artifact. Yellow flag is any sequential decline, which would suggest the tenfold jump was a pull-forward rather than a new baseline.

02 — Net interest revenue trajectory. Currently $389 million growing 9%. Green flag is holding flat or better through a cutting cycle, which would prove deposit growth is offsetting rate compression. Yellow flag is a sequential decline of more than 5%, because 30% of revenue rolling over changes the entire earnings model.

03 — Gold subscriber count. 4.8 million, up 39%. Green flag is crossing 6 million within three quarters, which puts the SaaS re-rating argument genuinely in play. Yellow flag is growth decelerating below 25%, signalling the conversion funnel from funded account to subscriber has saturated.

04 — Institutional share count. Fintel shows 777.8 million shares held as of August 2026 versus 568.5 million in May. Green flag is the number holding above 750 million through the next two 13F cycles, confirming the accumulation was conviction rather than a tactical trade. Yellow flag is a retracement below 650 million, which would mean the smart money took the pop.

05 — Crypto revenue stabilization. $100 million, down 38%. Green flag is any return to year-over-year growth, which would remove the last structural drag on the transaction line. Yellow flag is a further decline paired with rising customer-cash yield costs, because that combination is the yield war showing up in the P&L.

MY TAKE

The bottom line

I have watched a lot of companies claim they are becoming platforms. Almost none of them produce evidence. Robinhood produced evidence: a revenue line that went from nothing to $156 million a quarter and passed both of the company’s legacy businesses on the way. You cannot fake that. It means the distribution asset — 28.4 million funded accounts that will actually try a new product — is real and monetizable, which is the entire thesis in one sentence.

What gives me pause is not the business. It is the price. At 47.9 times earnings with 30% of revenue riding on the Fed, HOOD is priced as though the platform transition is already complete. It is not. It is maybe halfway through, and the second half depends on regulatory outcomes at the SEC and CFTC that no management team controls. The stock fell 59% in six months earlier this year while every operating metric improved. That will happen again.

The institutional accumulation is the detail I keep coming back to. 209 million shares in three months is not a rotation — it is a decision, made by people with better information than the tape, positioned ahead of a catalyst calendar that runs through 2027. Friday’s 13.7% move on 2.6x volume suggests that positioning is starting to get paid.

So I own the thesis and I respect the multiple. This is a stock to hold in size you can watch drop 40% without changing your mind, because at some point in the next eighteen months, it probably will.

Robinhood has already proven it can build a new business from zero to $156 million a quarter — the question is not whether the platform is real, but whether 48x earnings is the price of admission or the price of the exit.

ANALYST RATING

⚡ SPECULATIVE BUY

A 2-3% starter position with an 18-24 month horizon, added to only on drawdowns below $85 — the expected value is positive but the dispersion between the $57 and $164 Street targets means position size, not conviction, is what protects you here.

Sources & Further Reading

This is not financial advice. Always do your own research.

Cipher Digital, Inc. ($CIFR) Q2 2026 Earnings Call Review

Colby Hill — Sat, 22 Aug 2026 15:37:38 GMT

Reported: Tuesday, August 4, 2026, before the open | Call: 8:00 a.m. ET
Speakers: Tyler Page (Chief Executive Officer), Greg Mumford (Chief Financial Officer), Courtney Knight (Head of Investor Relations)
Overall Grade: C+

1. THE SCORECARD

METRIC                       ACTUAL        CONSENSUS      RESULT
--------------------------------------------------------------------
Revenue                      $24.84M       $32.52M        MISS (-23.6%)
Revenue YoY                  -43.0%        --             vs $43.6M LY
Revenue QoQ                  -28.7%        --             vs $34.8M Q1
GAAP EPS                     -$0.65        -$0.11 / -$0.23  MISS
GAAP net loss                -$267.5M      --             vs -$114.3M Q1
Gross profit                 $9.79M        --             vs $28.2M LY
Gross margin                 39.4%         --             vs 64.8% LY
Warrant remeasurement        -$150.5M      --             non-cash
Bitcoin mined (Odessa)       346 BTC       --             207 MW live
--------------------------------------------------------------------
Two consensus EPS figures circulated for this quarter: -$0.11 from
the brokerage feed and -$0.23 from Investing.com. Both are shown
because they were not reconcilable. The result is a miss on either.

Cipher does not issue formal financial guidance. What it guides is a construction schedule, so that is what the guidance table tracks.

GUIDANCE LINE                          STATUS        DETAIL
---------------------------------------------------------------------------
Formal financial guidance              NOT ISSUED    None given
Black Pearl Ph.1 initial capacity      AHEAD         Aug, 2 months early
Barber Lake Ph.1 (~168 IT MW)          MAINTAINED    Sept, rent from Oct
Stingray first delivery                MAINTAINED    1H 2027
Reveille + Ulysses                     MAINTAINED    270 gross MW, 2027
Contracted avg. annualized NOI         MAINTAINED    ~$793M, Oct26-Sep36
Need for additional equity             MAINTAINED    "do not expect to"
Total portfolio                        RAISED        5.3 GW, 11 sites
CapEx per critical IT megawatt         AT RISK       $9-11M, "creep higher"
ERCOT Batch Zero timing                DELAYED       Governor's letter

DATE            CLOSE       MOVE       NOTE
-----------------------------------------------------------------------
Aug 3  (Mon)    $24.16      --         Last close before the print
Aug 4  (Tue)    $20.38      -15.6%     Reported before the open
Aug 5  (Wed)    $18.71      -8.2%
Aug 6  (Thu)    $18.21      -2.7%
Aug 7  (Fri)    $17.18      -5.7%
Aug 10 (Mon)    $16.33      -4.9%      -32.4% in five sessions
Aug 18 (Tue)    $16.10      --         30-year Treasury yield 5.32%
Aug 21 (Fri)    $15.77      --         -34.7% from the pre-print close
-----------------------------------------------------------------------
52-week range: $5.71 - $30.14. The June 22 high was $30.14.

The tell: there was no bounce. A 15.6% drop on an earnings miss is ordinary. What is not ordinary is five consecutive red sessions afterward taking the cumulative decline to 32.4%, with no recovery in the two weeks since. That is not a market digesting a bad print, it is a market marking down the terminal value of the thesis. Two forces did the work. First, the revenue line made it undeniable that the legacy Bitcoin business is being dismantled faster than the data centre business is coming online, leaving a gap where earnings should be. Second, the same rate shock that hit every leveraged AI infrastructure name on August 18 hit a company whose entire model depends on issuing project bonds. Cipher is a developer that finances construction with debt; when the long end sells off, its cost of building rises and its exit multiple falls at the same time.

Grade rationale (C+): the operational half of this quarter was genuinely excellent. Cipher delivered data centre capacity two months early at the tenant’s request, priced its third project bond at its tightest coupon yet into a widening credit market, added up to 1.1 gigawatts of future capacity, and secured 96% to 100% of the equipment on its two nearest projects in a market where everyone else is complaining about supply. Against that: revenue fell 43% year over year and missed by 24%, gross margin has halved in four quarters, the reported loss more than doubled sequentially, and not one dollar of the $793 million of contracted annual NOI had reached the income statement during the period. A quarter where the operating story and the reported story point in opposite directions is the definition of a C. The plus is for the early delivery, which is the single most valuable thing a developer can prove.

2. REVENUE COMPOSITION: THE DELIBERATE COLLAPSE

Revenue fell to $24.84 million from $34.84 million in Q1 and $43.57 million a year ago. Management wants this understood as a choice rather than a deterioration, and on the facts they are right: the decline came from decommissioning Bitcoin mining at Black Pearl to clear the site for contracted data centre development. Cipher is deliberately destroying its only revenue line to build the thing that replaces it.

What remains of the mining business sits entirely at Odessa.

ODESSA (LAST OPERATING MINING SITE)     Q2 2026
------------------------------------------------------------
Operating capacity                      207 MW
Total hash rate                         ~11.6 EH/s
Fleet efficiency                        ~17.2 J/TH
Bitcoin mined in the quarter            ~346 BTC
Bitcoin held on balance sheet           $38M
Power arrangement                       Fixed-price PPA,
                                        Vistra Luminant
Further capital investment planned      None
Status                                  HPC conversion talks
                                        with multiple tenants

The strategic logic is sound. Odessa is already energised, which makes it a far shorter path to powered HPC capacity than any greenfield site. Page said a conversion could deliver first megawatts “very end of 2027” if a lease is signed within a couple of months. But until that lease exists, Odessa is a declining cash contribution attached to a site the company would rather repurpose, and the market is valuing it accordingly.

The gap in the middle is the problem. Contracted rent began at Black Pearl in August, after quarter end. Barber Lake rent begins in October. So Q2 was the trough quarter by construction: mining revenue already withdrawn, lease revenue not yet arrived. Anyone modelling Cipher on Q2 run-rate revenue is modelling a quarter that will never repeat in either direction.

3. WHERE THE LOSS ACTUALLY CAME FROM

The headline is that the net loss more than doubled sequentially, from $114.3 million to $267.5 million. That framing is accurate and misleading in equal measure.

LINE ITEM                            Q2 2026       Q1 2026
------------------------------------------------------------------
Revenue                              $  24.8M      $  34.8M
Gross profit                         $   9.8M      $  17.1M
Gross margin                         39.4%         49.2%
Warrant remeasurement (non-cash)     -$150.5M      +$ 43.6M
Interest income                      $  36.0M      --
Interest expense                     -$  67.0M     -$  59.0M
GAAP net loss                        -$267.5M      -$114.3M
GAAP EPS                             -$   0.65     -$   0.28
------------------------------------------------------------------
Net loss excluding the warrant
remeasurement in each period         ~-$117.0M     ~-$157.9M

Strip the warrant remeasurement out of both quarters and the loss did not double. It narrowed, from roughly $158 million to roughly $117 million. The entire deterioration in the headline number, and more, is a non-cash mark on warrant liabilities that swung $194 million quarter over quarter because the share price moved.

This is worth being blunt about: Cipher’s reported EPS is currently a derivative of its own stock price. When the shares rise, the warrant liability marks up and creates a loss. When they fall, it creates a gain. Q4 2025’s $734 million loss and Q1 2026’s smaller loss were both distorted by the same mechanism in opposite directions. Until those instruments are settled, the GAAP bottom line carries very little information about the operating business, and investors reacting to the headline EPS miss were reacting to an accounting artefact wrapped around a real revenue miss.

The real revenue miss, however, is not an artefact. Revenue came in 23.6% below consensus, and gross margin fell from 64.8% a year ago to 39.4%. That is what a shrinking hash rate against fixed site costs looks like.

4. BALANCE SHEET, CASH FLOW AND FUNDING

This is Cipher’s strongest section and the reason the equity has not de-rated further.

ITEM                                       Q2 2026       CHANGE
---------------------------------------------------------------------
Total assets                               $7.5B         +75% vs YE25
Property and equipment                     $2.13B        +$1.5B
Construction in progress                   $1.68B        +$1.4B
Restricted project cash                    ~$3.7B        --
  reserved for construction                ~$3.2B        --
  DSRA / interest during construction      ~$526M        --
Unrestricted cash                          $832M         +$204M
Bitcoin holdings                           $38M          --
Total unrestricted liquidity               $870M         --
Corporate + project debt outstanding       ~$6.0B        --
  of which convertible notes               $1.47B        two series
Revolver ($200M + $50M accordion)          undrawn       --
CapEx in the quarter                       ~$400M        ~$50M pipeline
Accounts payable                           $289M         from $40M

The Stingray financing is the proof point. Cipher priced $810 million of project-level senior secured notes at a 6% coupon, its lowest to date, in a quarter when credit spreads were widening across the sector. The book was roughly eight times oversubscribed and the deal funded approximately 98% of project costs while reimbursing $56.7 million of previously funded expenditure.

“Barber Lake established that the projects could be financed. Black Pearl demonstrated its repeatability. Stingray, our third successful project financing and our lowest coupon to date, demonstrated scalability and improving capital efficiency.”

Mumford was equally clear about the discipline behind it:

“We need each project to stand on its own, and it needs to generate attractive risk-adjusted returns.”

Two structural features matter more than the coupon. The notes amortise during the base lease terms, so debt service is matched to contracted rent rather than sitting as a bullet at maturity. And the existing committed funding is fixed rate and medium term, which insulates the projects already under construction from exactly the rate move that hit the stock on August 18. Mumford made this point directly when asked about credit markets: Cipher’s funded projects are already priced; it is the next deals that face the widening.

On the equity question, which was asked twice:

“We remain well capitalized to execute on our near-term commitments, and based on current forecasts, we do not expect to require additional equity.”

The qualifier is load-bearing. Mumford went on to explain that this holds only absent a large new lease, because signing “multiple hundreds of megawatts, up to a gigawatt” would require an equity down stroke Cipher does not currently carry. So the company is fully funded for what it has already signed, and would need capital for the thing shareholders most want it to do. That tension is not a criticism, it is simply the shape of the business.

5. EIGHT-QUARTER TREND

QUARTER      REVENUE     GROSS PROFIT   GROSS MGN   NET INCOME
---------------------------------------------------------------------
Q3 2024      $  24.1M    $   9.0M       37.5%       -$  86.8M
Q4 2024      $  42.2M    $  24.0M       56.9%       +$  17.5M
Q1 2025      $  49.0M    $  34.1M       69.6%       -$  39.0M
Q2 2025      $  43.6M    $  28.2M       64.8%       -$  45.8M
Q3 2025      $  71.7M    $  45.0M       62.7%       -$   3.3M
Q4 2025      $  59.7M    $  35.5M       59.4%       -$ 734.2M
Q1 2026      $  34.8M    $  17.1M       49.2%       -$ 114.3M
Q2 2026      $  24.8M    $   9.8M       39.4%       -$ 267.5M

What the table shows that a single quarter cannot: Cipher peaked in Q3 2025 at $71.7 million of revenue and 62.7% gross margin, and has given back nearly two thirds of that revenue in three quarters. The margin decline is the more revealing line, because it falls in a smooth staircase from 69.6% to 39.4% — this is not a bad quarter, it is a business being wound down on a schedule. Q2 2026 revenue of $24.8 million is almost exactly the $24.1 million Cipher printed in Q3 2024, two years earlier. On the income statement, the company has round-tripped to where it started. Everything that has been created in the interim sits on the balance sheet, in the $1.68 billion of construction in progress and the leases behind it, and none of it appears in this table yet. That is the entire investment case in one observation.

6. CONTRACTED NOI AND THE LEASE PIPELINE

MILESTONE                          CAPACITY       TIMING
------------------------------------------------------------------
Operating and contracted           907 MW         Live / in build
Black Pearl Ph.1 initial           --             Delivered Aug, early
Barber Lake Ph.1                   ~168 IT MW     Sept, rent from Oct
Stingray                           --             1H 2027
Reveille + Ulysses                 270 gross MW   2027
Colchis, Mikeska, McLennan         ~2.0 GW        2028-2029
Apollo, Barber Lake exp.,
  Stingray exp., Milsing           up to 2.1 GW   2030+
------------------------------------------------------------------
Total portfolio                    ~5.3 GW        11 sites
Future development pipeline        ~4.4 GW
Avg. annualized contracted NOI     ~$793M         Oct 2026 - Sep 2036

Three executed campus leases with investment-grade counterparties are expected to generate approximately $793 million of average annualised net operating income over the decade from October 2026 to September 2036. Against a current enterprise value of roughly $11.7 billion, that is about 14.8 times contracted NOI for the leased portion alone, with the entire 4.4 gigawatt pipeline as free option value.

The near-term catalyst is the 477 megawatts sitting at Odessa, Reveille and Ulysses. Page was unusually direct about the odds:

“All of those are past whatever milestones they would need to get a lease. The milestones remaining are do we like the terms and the tenant and we work through the legal documents? I expect that they will all be signed.”

7. WHAT MANAGEMENT ACTUALLY SAID

On the flywheel. Page’s framing of the quarter, and the sentence the company clearly wanted to land:

“What has become increasingly clear over the past few months is how each step forward in the progress of our flywheel is now reinforcing the next. The leases we’ve signed are giving prospective tenants more confidence to come to the table. The financings we’ve completed are strengthening our standing with capital markets.”

On what Cipher actually is. The company renamed itself from Cipher Mining to Cipher Digital, and the positioning follows:

“We control the full value chain of developing and delivering turnkey data centers to hyperscalers, from land and power origination to engineering, construction, and operations, which is what allows us to move at the speed and precision hyperscalers require.”

On the early delivery, which is the quarter’s real headline. Page explained why two months matters more than it sounds:

“Any developer can sign a lease. Far fewer can deliver ahead of schedule when a tenant asks for a faster timeline without cutting corners or sacrificing quality. What we have now proven is that Cipher has the operational depth across all of our teams to compress a delivery timeline on demand and without compromise.”

He is not overselling this. In a sector where the dominant news flow is slippage, delivering early at a tenant’s request is a differentiator that shows up in future lease terms and future bond pricing. Cipher is now able to point to a demonstrated construction track record in front of both tenants and creditors.

On the demand environment.

“This is the best demand environment we’ve ever seen. Terms continue to improve for developers.”

On procurement, where Cipher differs from peers. Most developers outsource. Cipher does not, and Page argued that is why its disclosure is more specific than the sector norm:

“Most of our competitors outsource procurement. We still manage that supply chain in-house... one of the reasons you get more transparency from us is that we actually know where we stand with procurement at the sites. We’re not sort of subject to the vagaries of the market and hoping to hear something positive from our outsource provider.”

8. THE Q&A THAT MATTERED

Stephen Byrd, Morgan Stanley - opened on the letter Texas Governor Greg Abbott had issued the previous afternoon, which introduced fresh uncertainty into ERCOT’s data centre interconnection batch process. Page said the company was “a little bit surprised” but not shocked, framing it as the predictable consequence of Texas becoming the most sought-after data centre market in the country:

“Texas is the most sought-after place for data centers right now because it’s the best environment. It’s got the best setup. They have been a bit of a victim of their own success as they try to whittle down the queue and just sheer volume of development that is sought there.”

Why it matters: this is the single largest external risk in the story. Cipher’s pipeline is concentrated in Texas by design, and roughly 2 gigawatts of it depends on Batch Zero inclusion. A delay is survivable; an adverse structural outcome is not. Page’s counter-argument, that scarcity favours well-prepared developers who have already funded deposits and submitted studies, is credible but self-serving.

Stephen Byrd, Morgan Stanley - follow-up on behind-the-meter generation, and the most striking answer of the call:

“It’s potentially extraordinarily meaningful... We have sites with, let’s say, readily available access to extraordinarily large quantities of natural gas... I think the potential is larger than the rest of our portfolio. How that actually comes into being will be determined by a lot of things like supply chain and financing and engineering. It’s too early for me to give exact forecasts on size, but it’s potentially enormous.”

Why it matters: the CEO just said an opportunity not included anywhere in the 5.3 gigawatt portfolio could be larger than everything currently in it. That is either the most under-appreciated line in the quarter or a well-hedged aspiration. It deserves tracking either way, because behind-the-meter gas is the only obvious route around the interconnection queue.

Paul Golding, Macquarie - on leasing Reveille and Ulysses. Page said demand “has never been stronger” with multiple interested parties at both sites, but was explicit that Cipher is holding out:

“We remain focused on securing the right deal for Cipher, not just the first deal available.”

Why it matters: this is the correct posture and also the reason the stock is under pressure. Every quarter Cipher waits for better terms is a quarter without a signed lease to point to. Management is optimising for lease economics; the market is scoring on lease count.

Richard Chough, J.P. Morgan - on how the two-month acceleration was achieved. Page credited an in-house, largely ex-hyperscaler construction and procurement team, and disclosed a structural detail: at that site Cipher operates under a cap on expenses, so burst labour to accelerate delivery “typically doesn’t fall on us.”

Why it matters: the early delivery cost Cipher less than it appears. The tenant absorbed the acceleration cost, which makes the achievement more repeatable, not less.

Jonathan Petersen, Jefferies - on whether Black Pearl Phase 2 also moves up. Page said no, the remaining deadlines are unchanged. He also pressed on the equity question, prompting Mumford’s clarification that Q2 CapEx was roughly $400 million of which only about $50 million went to the development pipeline, and that a large new lease would require an equity contribution.

Why it matters: the honest read is that Cipher is funded for its committed book and would need to raise to grow beyond it. Investors should expect dilution to accompany the next major lease announcement, and should probably welcome the trade.

Chris Brendler, Rosenblatt Securities, and Michael Chen (for John Todaro), Needham - both pushed on capital cost per megawatt against the historical $9 million to $11 million band. Page conceded the direction:

“We definitely see inflation on labor and equipment across the board. As we build budgets for the next build, I would expect our budgeting to go up, and that probably translates into a higher cost per megawatt... I would say we’re running at the contingency level given that inflation.”

Why it matters: “running at the contingency level” means current projects are consuming the buffer built into their budgets. There is no cushion left on Black Pearl, Barber Lake and Stingray. Future leases will need higher rents to hold returns constant, which is achievable in this demand environment but is a real headwind to the NOI-per-megawatt assumption embedded in most models.

9. SITE-BY-SITE PROGRAM STATUS

SITE          STATUS                          EQUIPMENT SECURED
----------------------------------------------------------------------
Black Pearl   Ph.1 initial delivered early    ~96% (Ph.1 + Ph.2)
Barber Lake   Ph.1 delivery Sept, rent Oct    100%
Stingray      Earthwork, underground elec.    ~75%
Odessa        207 MW mining, conversion talks n/a
Reveille      Interconnect approved, leasing  n/a
Ulysses       Interconnect approved (PJM)     n/a
Colchis       Deposits funded, Batch Zero     n/a
Mikeska       Deposits funded, Batch Zero     n/a
McLennan      Deposits funded, Batch Zero     n/a
Apollo        Batch Zero studied load         n/a
Milsing       Expected in Batch One           n/a

Two hires signal where the company thinks its risk sits. Bill Blevins joined as Head of Grid Strategies from ERCOT, where he ran large load interconnections — precisely the process now bottlenecking the pipeline. Mohammed Aboulela arrived from Google, where he spent eleven years on technical due diligence, design and delivery across more than five gigawatts of data centre capacity. Cipher is hiring against the interconnection queue and against hyperscaler build standards, which is the correct read of its own constraints.

10. VALUATION

METRIC                                   CIFR
--------------------------------------------------------------
Share price (Aug 21 close)               $15.77
Market capitalization                    $6.54B
Shares outstanding                       415.0M
52-week range                            $5.71 - $30.14
Price / book                             11.62x
Trailing P/E                             n/m (loss-making)
Approximate enterprise value             ~$11.7B
EV / contracted avg. annualized NOI      ~14.8x
EV per total portfolio megawatt          ~$2.2M / MW
EV per operating + contracted megawatt   ~$12.9M / MW
--------------------------------------------------------------
EV approximated as market cap plus ~$6.0B of corporate and
project debt less $832M of unrestricted cash. Restricted project
cash is excluded because it is committed to construction.

The two megawatt-based figures bracket the argument. At roughly $12.9 million per operating-or-contracted megawatt, Cipher trades above its own stated $9 million to $11 million build cost, meaning the market is paying a premium to replacement cost for the leased book and assigning meaningful value to the leases themselves. At roughly $2.2 million per portfolio megawatt across all 5.3 gigawatts, the market is assigning close to nothing to the 4.4 gigawatt pipeline. Which number is right depends entirely on how many of those pipeline megawatts convert into signed leases, and Batch Zero is the gate.

11. RISKS AND WHAT TO WATCH

Near-term. The ERCOT batch process. Governor Abbott’s letter arrived the day before the call and pushed out a decision Cipher expected within days. Roughly 2 gigawatts across Colchis, Mikeska and McLennan depend on Batch Zero inclusion, and Apollo and the Barber Lake and Stingray expansions are submitted into the same process. Management expressed “strong conviction” all three core sites qualify. That conviction is not yet confirmed.

Near-term. Lease signings at Odessa, Reveille and Ulysses. Page said he expects all 477 megawatts to be leased. Each quarter without an announcement makes that harder to underwrite, and the stock is now priced as though none of it converts.

Near-term. The Q3 print is the first quarter with contracted rent flowing from both Black Pearl and Barber Lake. This is the number that establishes whether the $793 million NOI figure is a real run rate or a modelling exercise.

Medium-term. Cost inflation. Management is running at contingency on current builds and expects budgets to rise. If rents do not rise with them, returns on future leases compress.

Medium-term. Rate sensitivity. Cipher funds construction with project bonds. Committed projects are fixed-rate and insulated, but every future project prices at the market’s clearing spread. The August 18 session, when the 30-year touched 5.32% and CIFR fell 9% intraday on no company news, is the mechanism.

Structural. Warrant remeasurement will keep making reported earnings meaningless in both directions until those instruments are resolved. Investors and headline writers will keep reacting to numbers that carry no operating information.

Structural. Concentration. Texas is the pipeline. That was a deliberate and, until recently, extremely profitable bet. It is now a single-jurisdiction regulatory exposure at exactly the moment the jurisdiction is rewriting its rules.

12. BOTTOM LINE

The operating case. Cipher delivered capacity two months early when a tenant asked, which is the hardest thing to do in this industry and the easiest thing for a prospective tenant to verify. It priced its cheapest bond yet, eight times oversubscribed, into a widening credit market. It has 96% to 100% of equipment secured on its two nearest projects while the rest of the sector queues for transformers. It added up to 1.1 gigawatts of future capacity and hired the ERCOT executive who used to run the interconnection process it now has to navigate. Total assets grew 75% in six months. On execution, this was a strong quarter.

The bear case. The income statement has round-tripped to 2024 levels. Revenue is down 43% year over year and missed consensus by 24%. Gross margin has fallen thirty points in four quarters. Reported EPS is a function of the share price rather than the business. The $793 million of contracted NOI is real and contracted and had produced exactly zero revenue as of June 30. And roughly 2 gigawatts of the pipeline now sits behind a regulatory process that a state governor intervened in the day before the call.

The synthesis. Cipher is a construction company that reports like a failing miner. Every dollar of value created in the last eighteen months sits in construction in progress, restricted project cash and lease contracts, none of which appear in revenue. The market has been willing to pay for that in advance — the stock is up substantially over the past year even after this drawdown — but Q2 was the quarter where the gap between the balance sheet and the income statement became too wide to ignore, and the rate move two weeks later removed the market’s patience for financing it.

For investors: the next two quarters resolve most of this. Barber Lake and Black Pearl rent hits the P&L in Q3 and Q4, which converts the NOI claim from a slide into a reported number. Batch Zero determines whether the pipeline is worth $2.2 million per megawatt or considerably more. And a signed lease at any of Odessa, Reveille or Ulysses would validate Page’s claim that holding out for better terms was the right call rather than an excuse. Until those arrive, this remains a story where the operational evidence and the financial evidence point in opposite directions, and reasonable people can weight them differently. The Q3 report is scheduled for November 2.

Grade: C+

This is not financial advice - for informational and educational purposes only.

CoreWeave, Inc. ($CRWV) Q2 2026 Earnings Call Review

Colby Hill — Sat, 22 Aug 2026 15:37:19 GMT

Reported: Tuesday, August 11, 2026, after the close | Call: 5:00 p.m. ET
Speakers: Michael Intrator (Chief Executive Officer), Nitin Agrawal (Chief Financial Officer)
Overall Grade: A-

1. THE SCORECARD

METRIC                      ACTUAL       CONSENSUS     RESULT
--------------------------------------------------------------------
Revenue                     $2.58B       $2.56B        BEAT
Revenue growth YoY          +112%        --            vs $1.21B LY
Revenue growth QoQ          +24%         --            --
Adjusted EPS                -$1.03       -$1.20        BEAT
Adj. operating income       $128M        $66M          BEAT (+94%)
Adjusted EBITDA             $1.50B       --            59% margin
GAAP net loss               -$626M       --            vs -$291M LY
GAAP gross margin           65.9%        --            vs 74.2% LY
Interest expense            $640M        --            vs $267M LY
Free cash flow              -$5.7B       --            CapEx $9.4B

Every guidance line moved up.

GUIDANCE LINE                    NEW               PRIOR         CALL
--------------------------------------------------------------------------
FY26 revenue                  $12.4B-$13.2B      not disclosed  RAISED
FY26 adj. operating income    $960M-$1.15B       not disclosed  RAISED
FY26 CapEx                    $35B-$39B          not disclosed  RAISED
FY26 exit ARR                 $18.5B-$19.5B      $18B-$19B      RAISED
FY26 year-end active power    >1.85 GW           >1.70 GW       RAISED
Q3 revenue                    $3.45B-$3.60B      --             NEW
Q3 adj. operating income      $200M-$260M        --             NEW
Q3 interest expense           $860M-$940M        --             NEW
Q3 CapEx                      $11.5B-$13.5B      --             NEW
Q4 adj. operating margin      low teens          --             NEW

DATE            CLOSE       MOVE       NOTE
-----------------------------------------------------------------------
Aug 11 (Tue)    $ 90.32     --         Reported after the close
Aug 12 (Wed)    $107.73     +19.3%     The earnings pop
Aug 13 (Thu)    $106.29     -1.3%      Traded as high as $117.49
Aug 14 (Fri)    $105.26     -1.0%
Aug 17 (Mon)    $106.00     +0.7%
Aug 18 (Tue)    $ 93.17     -12.1%     30-year Treasury yield 5.32%
Aug 19 (Wed)    $ 90.87     -2.5%
Aug 20 (Thu)    $ 89.76     -1.2%
Aug 21 (Fri)    $ 87.85     -2.1%      Below the pre-print close
-----------------------------------------------------------------------
Net move, Aug 11 close to Aug 21 close:            -2.7%

The tell: the market did not reject this quarter. It gapped the stock up 19.3% the morning after, held that level for three sessions, and then handed all of it back plus a little more when the 30-year Treasury yield hit 5.32%, its highest level since early 2024. CoreWeave finished the following Friday below where it traded before it reported a beat-and-raise. Nothing in the operating results changed between August 12 and August 21. What changed was the discount rate applied to a company that spent $9.4 billion on CapEx in a single quarter and guided to $35 billion to $39 billion for the year. This is the cleanest example this earnings season of a stock trading on its cost of capital rather than its income statement.

Grade rationale (A-): CoreWeave beat on revenue and adjusted EPS, doubled adjusted EBITDA year over year, delivered the sequential margin inflection management had been promising for three quarters, nearly tripled backlog, took price up 25% across SKUs mid-quarter and still saw demand exceed supply, and raised every single guidance line including CapEx. Offsetting that: GAAP gross margin compressed more than eight points year over year, adjusted operating income of $128 million was actually down from $200 million in the year-ago quarter, net loss more than doubled, interest expense grew 140% to $640 million and is guided to as much as $940 million next quarter, and free cash flow ran to negative $5.7 billion. This is an A quarter attached to a financing structure that is one rate regime away from being the whole story. It does not get an A because the leverage is no longer a footnote.

2. REVENUE COMPOSITION AND CUSTOMER MIX

Revenue of $2.58 billion grew 112% year over year and 24% sequentially. The more interesting number is what sits underneath it. CoreWeave is no longer purely a GPU rental business, and Q2 was the quarter that became visible in the disclosure.

REVENUE / ARR LINE                  Q2 2026        NOTE
--------------------------------------------------------------------
Total revenue                       $2,575M        +112% YoY
Non-GPU product ARR                 $400M+         Storage, CPU,
                                                   networking, software
Managed inference ARR               $100M+         Was $1M at launch
Managed inference, 2026 exit target $250M+         Guided on this call
Model training runs tracked         1B+            Cumulative milestone
New AI platform capabilities, Q2    7              Multiple firsts

The managed inference number is the one to sit with. It went from roughly $1 million to more than $100 million of booked ARR in the few months since launch, and management says growth is “constrained only by our near-term capacity” rather than by demand. That is a different kind of revenue than a five-year take-or-pay GPU contract: higher margin, shorter duration, and much closer to a software gross margin profile.

Customer concentration is also visibly loosening. Intrator named Caterpillar deploying NVIDIA Vera Rubin for physical AI training on autonomous construction equipment, Isomorphic Labs in life sciences, Flow Traders and IMC in systematic trading, Leidos in the public sector through CoreWeave Federal, and IBM, Nissan and ZF via the Monolith field engineering team. Separately, press coverage of the quarter noted a $21 billion agreement to supply AI cloud capacity to Meta through 2032 and a multi-year agreement with Anthropic. A year ago the bear case on CoreWeave was that it was a leveraged bet on a handful of labs. That case is materially weaker than it was.

3. THE MARGIN INFLECTION

Management spent three quarters telling investors margins would inflect in the second quarter. They did. But the shape of the inflection deserves precision, because the headline and the year-over-year comparison point in opposite directions.

LINE ITEM                     Q2 2026     Q2 2025     Q1 2026
---------------------------------------------------------------------
Revenue                       $2,575M     $1,213M     ~$2,076M
GAAP gross profit             $1,696M     $  900M     n/a
GAAP gross margin             65.9%       74.2%       n/a
Adjusted EBITDA               $1,500M     $  753M     n/a
Adjusted EBITDA margin        59%         62%         n/a
Adjusted operating income     $  128M     $  200M     $   21M
Adjusted operating margin     5.0%        16.5%       ~1.0%
Interest expense              $  640M     $  267M     n/a
GAAP net loss                 -$  626M    -$  291M    n/a
Adjusted net loss             -$  567M    -$  130M    n/a
---------------------------------------------------------------------
Q1 2026 revenue implied from the 24% sequential growth management
cited on the call. Operating expenses were $2.6B in Q2, including
$165M of stock-based compensation.

Read the sequential column and this is a triumph: adjusted operating income went from $21 million to $128 million, roughly six-fold, and beat the Street’s $66 million by 94%. Read the year-over-year column and adjusted operating income fell from $200 million while revenue more than doubled. Both are true. The reconciliation is that CoreWeave spent the last four quarters absorbing the ramp cost of building an installed base that is now large enough for new deployments to stop dominating the P&L.

Agrawal was explicit about why this is the turn:

“Perhaps most importantly, Q2 marked the quarter in which we saw margins inflect expanding sequentially as we had discussed over the past several quarters.”

Intrator framed the mechanics in a way that is worth quoting in full, because it is the actual bull case:

“Each new deployment is landing against a much larger installed base than it was even 1 quarter ago. As that base grows, each new build becomes a smaller part of the whole, while contracted revenue from existing deployments remains in place. This is how we are transforming scale into operating leverage.”

The pricing side reinforced it. Management pushed through an approximately 25% price increase across SKUs in July, and contracts signed during the quarter carried contribution margins five to ten percentage points above those added in recent quarters.

“Pricing and margins for our Blackwell and Vera Rubin SKUs are setting new highs, while pricing for prior generation SKUs is at or above where it was years ago.”

The one line that did not cooperate is GAAP gross margin, down from 74.2% to 65.9%. That is depreciation on a rapidly enlarging fleet working through cost of revenue, and it will keep compressing as long as CapEx runs at this pace. Adjusted EBITDA margin held at 59% against 62% a year ago, which is the more useful cash proxy.

4. BALANCE SHEET, CASH FLOW AND FUNDING

This is where the quarter gets genuinely uncomfortable, and it is the reason the stock round-tripped.

ITEM                                          Q2 2026
------------------------------------------------------------------
Cash, equivalents, restricted cash, mkt sec   $6.9B+
Capital raised during Q2                      ~$18B
Cumulative debt + equity capital secured      $32B+
CapEx, Q2                                     $9.4B
CapEx, Q3 guided                              $11.5B-$13.5B
CapEx, FY26 guided                            $35B-$39B
Construction in progress                      $11.9B (from $9.6B)
Free cash flow, Q2                            -$5.7B
Interest expense, Q2                          $640M
Interest expense, Q3 guided                   $860M-$940M
Weighted avg. cost of debt, 12-month change   ~-300 bps
Implied annualized interest savings           ~$1.1B

CoreWeave raised roughly $18 billion in a single quarter across debt, convertibles and equity, including an inaugural Eurobond and its first publicly-issued delayed draw term loan backed by HPC infrastructure. The most recent transaction, a second publicly syndicated term loan, was the first to include shorter-duration customer contracts in the collateral pool, which matters because it unlocks financing for the higher-ASP, higher-margin short-dated business.

“The transaction, which priced during one of the most dislocated weeks for credit this year was met with meaningful interest. Despite the environment at the time, we chose to complete the transaction at its full size. Our spreads have retraced since.”

Agrawal also laid out the unit-economics argument that underwrites the whole model, and it is the single most important paragraph on the call:

“The cost, primarily in the form of CapEx is front-loaded requiring a combination of debt, customer prepayments, and other corporate level capital to finance its build out. Once the cluster is delivered, contracted revenue ramps becoming predictable and highly cash flow generative.”
“So when an initial contract ends, the cluster no longer has any leverage, and we are free to recontract that cloud infrastructure or offer it to the market... Every resale or renewal is incremental on top of the returns already earned within the initial term.”

The residual-value question has been the core bear argument against every GPU cloud since 2023, and CoreWeave answered it with a specific data point: it recently signed an A100 contract extending into 2029 on a SKU introduced in 2020. If prior-generation silicon is still commanding attractive prices nine years after launch, the depreciation schedules the market has been applying are too aggressive. That is a genuinely important disclosure.

None of which changes the near-term math. Negative $5.7 billion of free cash flow against $6.9 billion of liquidity means CoreWeave must return to the capital markets continuously, at whatever price those markets set. When the long end sold off on August 18, the stock fell 12.1% in a session on no company-specific news.

5. EIGHT-QUARTER TREND

QUARTER      REVENUE     GROSS PROFIT   GROSS MGN   NET INCOME
---------------------------------------------------------------------
Q1 2025      $   982M    $   719M       73.3%       -$  315M
Q2 2025      $ 1,213M    $   900M       74.2%       -$  291M
Q3 2025      $ 1,365M    $   996M       73.0%       -$  110M
Q4 2025      $ 1,572M    $ 1,063M       67.6%       -$  452M
Q1 2026      ~$ 2,076M   n/a            n/a         n/a
Q2 2026      $ 2,575M    $ 1,696M       65.9%       -$  626M
---------------------------------------------------------------------
CoreWeave has been public since March 2025, so a full eight-quarter
public history does not yet exist. Q1 2026 revenue is implied from
management's 24% sequential growth comment.

What the table shows that a single quarter cannot: revenue has compounded at roughly 27% per quarter for six straight quarters, which is extraordinary and is not the interesting part. The interesting part is that gross margin has fallen in a near-straight line from 74.2% to 65.9% while net loss has widened from $291 million to $626 million. CoreWeave is converting balance sheet into revenue at an accelerating rate, and the income statement is absorbing the conversion cost in real time. The Q3 2025 quarter, with a net loss of only $110 million, now looks like the anomaly rather than the trend.

6. BACKLOG, CAPACITY AND PRICING

METRIC                            Q2 2026      CHANGE       NOTE
-----------------------------------------------------------------------
Revenue backlog                   $104.2B      +246% YoY    As of Jun 30
Net new commitments, early Q3     $25B+        --           Not in above
Backlog with delivery commenced   >50%         --           >2/3 by YE26
Active power                      1.5 GW       ~3x YoY      +500MW in Q2
Power added in June alone         300+ MW      --           > any prior qtr
Contracted power, at call date    4.2 GW       --           3.7 GW at Q2 end
Powered land / options / LOIs     1.5+ GW      --           Excluded above
Contracted power outside the US   1+ GW        --           Incl. 360MW APAC
2030 active power target          8+ GW        --           Reaffirmed

Backlog of $104.2 billion against trailing twelve-month revenue of roughly $7.6 billion is a coverage ratio that does not exist elsewhere in infrastructure. And the $25 billion of net new commitments signed in the first weeks of Q3 is not in that figure. More than half of the backlog is attached to contracts where delivery has already commenced, with management guiding that to more than two-thirds by year end, which is the disclosure that separates a real backlog from a press release.

“Our near-term capacity remains effectively sold out. That is translating into signed commitments on increasingly favorable terms from a broadening set of customers and is positioning CoreWeave to gain market share for years to come.”

On the scale of the Q2 build:

“To put that in perspective, we added more power in Q2 than any single neo-cloud operates in total today, according to third-party estimates.”

7. WHAT MANAGEMENT ACTUALLY SAID

On the quarter overall. Intrator opened without hedging:

“Q2 was an exceptional quarter for CoreWeave. We outperformed our plan across the board with the operating leverage we have been building beginning to show up clearly in our results.”

On why customers pay up. The competitive argument is now a cost argument rather than an availability argument, which is a meaningful shift:

“According to Signal65, CoreWeave delivers total cost of ownership estimated to be up to 47% lower than the average hyperscaler.”

On the structural demand thesis. Intrator’s framing of why AI compute demand does not plateau after training is the most useful strategic argument on the call:

“For the last several years, many organizations treated a model like a deliverable, train it, deploy it and move on. Enterprises no longer operate that way. Training, inference, evaluation and improvement now form a single continuous loop... Compute is no longer a onetime requirement concentrated at the beginning of a model’s life. It becomes an ongoing requirement that grows with every application in production and every cycle of improvement.”

On the supply chain. Asked how CoreWeave keeps securing inputs in a constrained market:

“End of the day, my job is to ensure that this company has the capacity to deliver the product that our clients require. And in order to do that, we need to aggressively manage a complicated supply chain. And that supply chain includes everything from land power and shell through GPUs and networking through memory, all of which is being challenged by the growth and expansion of artificial intelligence.”

8. THE Q&A THAT MATTERED

Samik Chatterjee, JPMorgan - on recontracting older fleets and input costs. Agrawal said the capacity coming up for renewal is “a very limited part of our fleet” and that ASPs on older generations remain at or above levels seen a year ago. He then delivered the line that explains the entire margin inflection:

“The increase in the value of output of the CoreWeave cloud has outpaced the value of the input increases that we are currently experiencing in the supply chain. And as a result of it, what you’re seeing is margins expand.”

Why it matters: CoreWeave is not absorbing component inflation, it is passing it through with a spread. That is pricing power, and it is the difference between a landlord and a commodity reseller.

Brad Zelnick, Deutsche Bank - on how much of the 5-to-10-point margin improvement is duration versus differentiation. Intrator declined to decompose it but pointed to platform quality, reliability, security and TCO, and added that many customers are now monetizing their own products, which changes their willingness to pay.

Why it matters: the honest answer is that it is both, and the duration piece is only financeable because of the DDTL 5.5 structure. The two are linked.

Irvin Liu (for Amit Daryanani), Evercore ISI - on data center moratoriums and the path to 8 gigawatts. Intrator’s answer was the most quotable of the call:

“We believe that the certain communities have moved forward with moratoriums. And we feel like moratoriums are -- they’re not going to impact the demand for this infrastructure, they are going to impact where this infrastructure gets built.”

Agrawal added the arithmetic: 4.2 gigawatts contracted plus roughly 1.5 gigawatts of powered land and executed LOIs puts CoreWeave near 6 gigawatts identified, halfway through 2026, against an 8 gigawatt target for 2030.

Why it matters: permitting risk is real but it is a siting problem, not a demand problem, and CoreWeave has enough optionality banked to route around individual jurisdictions.

Raimo Lenschow, Barclays - on edge AI and competitive encroachment following Meta news the prior day. Intrator leaned on information advantage:

“One of the things about CoreWeave that should never be underestimated is we sit at the epicenter of an incredible amount of information flow from across the entire industry, right? The hyperscalers use us, the labs use us. You’ve seen enterprise begin to scale within our platform.”

Agrawal followed with the cleaner rebuttal:

“Even with this increased competition, we are seeing demand, pricing and margin all expanding which is a signal for the growth in the CoreWeave kind of product as well as our growth overall in an already massive TAM that exists.”

Why it matters: the falsifiable test of a competition thesis is price. Price went up 25% in July. Until that reverses, the encroachment argument is theoretical.

Michael Turrin, Wells Fargo Securities - on the gap between 500 megawatts of added power and flat sequential revenue adds. Agrawal confirmed 300 of the 500 megawatts landed in June, meaning the revenue contribution is a Q3 and Q4 event. Intrator on the new silicon:

“Vera Rubin is a generation that is seeing the margin expansion right from the start... when we were talking about that 5% to 10% margin step function that we’re seeing, a lot of that is coming in, in the Vera Rubin SKU.”

Why it matters: this is the single best reason to believe the Q3 revenue guide of $3.45 billion to $3.6 billion. The capacity is already energized; it just landed too late in the quarter to bill.

Brett Knoblauch, Cantor Fitzgerald - on contract duration strategy after the DDTL 5.5:

“The execution of the DDTL 5 puts CoreWeave in a position where we get to populate the curve in terms of what we think is the most profitable configurations for term leasing. And so we want to sell our compute on long-term contracts. We also want to sell it on shorter-term contracts to extract additional margin.”

Why it matters: financing innovation is now driving revenue mix, not the other way around. That is a competitive moat that has nothing to do with GPUs.

9. POWER, SILICON AND SUPPLY CHAIN STATUS

ITEM                          STATUS
-------------------------------------------------------------------
Active power                  1.5 GW, +~500 MW in Q2
Contracted power              4.2 GW as of the call date
First self-build              Online later in 2026
International                 1+ GW contracted outside the US
APAC entry                    360 MW Indonesia, live in ~18 months
Vera Rubin NVL72              First cloud provider to validate
MLPerf                        New training and inference records
Gartner                       Named a Visionary, 2026 Cloud AI MQ
Memory / storage supply       Long-term Solidigm agreement signed
Omni (sovereign/enterprise)   First deal signed, scales in 2027

The vertical integration story is starting to matter. Powered land gives CoreWeave control over the shell, the cooling design and the rack layout, which is where the software-defined liquid cooling work that enabled the first Vera Rubin NVL72 validation actually lives. Management explicitly ties powered land to “enhanced long-term margins,” and the first self-build comes online before year end.

10. VALUATION

METRIC                                  CRWV
-------------------------------------------------------------
Share price (Aug 21 close)              $87.85
Market capitalization                   $48.4B
Shares outstanding                      551.5M
52-week range                           $60.55 - $153.20
Price / book                            9.63x
Trailing P/E                            n/m (loss-making)
Approximate enterprise value            ~$71B
EV / FY26 revenue guidance (midpoint)   ~5.6x
EV / FY26 exit ARR (midpoint $19B)      ~3.8x
EV / revenue backlog ($104.2B)          ~0.68x
-------------------------------------------------------------
EV approximated as market cap plus roughly $30B of long-term debt
less $6.9B of cash and equivalents. Debt figure from press coverage
of the quarter, not from the connector.

The stock sits 43% below its October 2025 high of $153.20 and 45% above its July 2026 low of $60.55. At roughly 3.8 times exit ARR, CoreWeave trades at a discount to most infrastructure software and at a premium to most REITs, which is approximately correct for a business that is neither. The backlog multiple of 0.68 times is the number bulls will point to; the honest response is that backlog converts to revenue only if CoreWeave can finance the CapEx required to deliver it, which is precisely what the August 18 session was about.

11. RISKS AND WHAT TO WATCH

Near-term. Interest expense is guided to $860 million to $940 million in Q3, up from $640 million. At the midpoint that is roughly 25% of guided Q3 revenue consumed by interest before a dollar of operating cost. Watch whether adjusted operating income of $200 million to $260 million actually covers the sequential step-up in financing cost, because on the guided numbers it does not.

Near-term. The Q3 revenue guide depends on the 300 megawatts energized in June converting to billed revenue on schedule. Construction in progress of $11.9 billion is the leading indicator; if it does not convert to property and equipment next quarter, the ramp slipped.

Medium-term. Rate sensitivity is now the dominant driver of the share price. A business funding $35 billion to $39 billion of annual CapEx with debt is a duration instrument whether or not management describes it that way. The August 18 move, minus 12.1% on a Treasury auction rather than a customer loss, is the template.

Medium-term. Component inflation is currently being passed through with a spread. Memory in particular is tightening across the industry. If the spread compresses, the 5-to-10-point contribution margin uplift on new contracts goes with it.

Structural. Residual value. Management made a strong case with the 2029 A100 contract, but the entire model assumes fleets can be recontracted profitably after the initial term retires the asset-level debt. That assumption has not yet been tested across a full cycle at scale.

Structural. Permitting and community opposition. Intrator is probably right that moratoriums move the map rather than shrink the demand, but moving the map costs time and capital, and CoreWeave’s 2030 target assumes a build cadence with very little slack.

12. BOTTOM LINE

The operating case. CoreWeave did exactly what it said it would do. Margins inflected sequentially in the quarter management said they would, adjusted operating income beat consensus by 94%, backlog nearly tripled, the company took price up 25% and demand still exceeded supply, managed inference went from a rounding error to $100 million of ARR in a few months, and every guidance line for the year moved higher. The customer list broadened from AI labs into industrials, life sciences, trading firms and the federal government. On execution, this was close to a flawless quarter.

The bear case. Free cash flow was negative $5.7 billion. Interest expense is about to approach a billion dollars a quarter. GAAP gross margin has fallen eight points in a year and adjusted operating income is lower than it was a year ago on more than double the revenue. CoreWeave has to raise capital more or less permanently, and the terms of those raises are set by a bond market that just repriced the long end to its highest yield in more than two years. None of that is fixable through operational excellence.

The synthesis. The gap between the quarter and the stock is the point. CoreWeave delivered an A- quarter and the shares finished ten days later below where they started. That is not the market disputing the numbers, it is the market repricing the cost of the capital required to produce them. Both things can persist for a long time: a company can compound revenue at 112% and still see its equity fall, if the discount rate rises faster than the earnings.

For investors: this is now a rates trade wearing an AI costume. The operating business is compounding faster than almost anything of comparable size, the backlog gives multi-year visibility that very few infrastructure companies can match, and the residual-value disclosure was genuinely reassuring. But position sizing should reflect that the single largest determinant of the share price over the next several quarters is the 30-year Treasury, not the Q3 revenue print. The next checkpoints are the Fully Connected developer conference in September and the Q3 report on November 9.

Grade: A-

This is not financial advice - for informational and educational purposes only.

MSFT: The Trillion Dollars Microsoft Added While Nobody Was Looking

Colby Hill — Sat, 22 Aug 2026 11:40:02 GMT

August 22, 2026 | Deep Dive by Colby

OPENING HOOK

Why this story matters right now

On June 25, 2026, Microsoft traded at $349.20. That was a fresh 52-week low, capping the company's worst month since the dot-com crash — down more than 21% in June alone and roughly 35% off the all-time closing high set the prior October.

Fifty-seven days later the stock closed at $483.24. That is a gain of 38.4%, and in dollar terms it is close to $1 trillion of market capitalization recreated out of thin air in under two months. The entire market cap of Broadcom, added back to one company, in less time than a school summer break.

Here is what makes the reversal remarkable: nothing about the AI spending question actually got answered. Microsoft did not stop building. It spent $41 billion on capital expenditures in a single quarter, up 70% year over year. What changed is that for the first time in this cycle, the revenue moved faster than the spending — and management lowered the capex guide while raising the growth guide. Wall Street had been waiting two years for exactly that sentence.

Microsoft Corporation is now the ninth-most-discussed ticker on X, a mega-cap that trades like a momentum name, and the single clearest referendum available on whether the AI infrastructure buildout produces returns or craters. The stock is trending because the argument finally has data on both sides.

INVESTMENT THESIS

The core reasons to own this stock

1. Azure crossed $100 billion and accelerated. Most businesses decelerate as they scale. Azure did the opposite — full-year revenue topped $100 billion for the first time, growing 41%, and the June quarter came in at 43%. Management guided the following quarter to roughly 45% in constant currency. A hundred-billion-dollar business growing faster than it did at eighty billion is a rare and violently repriceable thing.

2. The capex guide came down while growth went up. Calendar-2026 capital expenditure guidance fell from roughly $190 billion in April to about $175 billion in July. That $15 billion reduction is the entire bull case in one number. It says Microsoft believes it can meet demand without an unbounded arms race, which converts the AI story from a spending problem into a margin problem — and Microsoft has never lost a margin argument.

3. Copilot finally has the seat math. Microsoft 365 Copilot passed 30 million paid seats, adding 10 million in a single quarter after taking years to reach 20 million. Customers with 50,000-plus seats grew more than sevenfold year over year. This is the software attach rate that turns Azure's infrastructure spend into a high-margin annuity.

4. The valuation never re-rated with the fundamentals. At $483.24 the stock trades at roughly 26.8x trailing earnings — for a company that just grew revenue 17.8% and net income 31.3% for the full fiscal year, with a 39.7% net margin in the most recent quarter. The market is paying a market multiple for one of the two or three best businesses on earth.

BUSINESS OVERVIEW

What Microsoft actually does

Microsoft runs three reporting segments, and it is worth being precise about them because the AI narrative has flattened public understanding of the company into "Azure and Copilot." Productivity and Business Processes houses Office Commercial and Consumer, LinkedIn, and Dynamics — the annuity that pays for everything else. Intelligent Cloud contains Azure, server products, and enterprise services, and is the segment the entire stock now trades on. More Personal Computing covers Windows, Surface, Xbox and its cloud gaming layer, Edge, and the consumer surface area that generates distribution but not much investor attention.

The company was founded in 1975, employs roughly 223,000 people, and is headquartered in Redmond, Washington. What it actually sells, stripped to its economics, is a contractual claim on enterprise IT budgets. Microsoft's genius has never been inventing the category — it was not first to the cloud, first to search, or first to large language models. Its genius is being the vendor already inside the building when the category arrives, holding the identity layer, the compliance stack, and the procurement relationship.

That is precisely the mechanism now converting AI demand into Azure revenue. Commercial remaining performance obligation reached $678 billion, up 84% year over year with a weighted-average duration of 2.3 years. Even stripping OpenAI entirely out of the figure, RPO still grew 25%. Microsoft is not selling AI to the curious. It is selling contracted, multi-year compute to customers who already sign the paperwork.

INDUSTRY OPPORTUNITY

The hyperscalers now own two-thirds of a $143 billion quarterly market

The global cloud infrastructure market ran at roughly $143 billion per quarter in mid-2026, and the big three captured about two-thirds of it. AWS holds roughly 28% share, Azure 21%, and Google Cloud 15%. Those numbers matter less than the direction of travel: this is a market where a single quarter of relative growth moves share by a full point, and where the constraint is not demand but the physical availability of power, land, and silicon.

That constraint is the whole game. The industry-wide AI buildout has been estimated at $725 billion to $805 billion, and the reason Microsoft's stock lost a third of its value in the first half of 2026 is that investors could see the spending clearly and the returns dimly. Free cash flow made the argument for them — Microsoft's FY26 free cash flow came in at $66.99 billion, down 6.5% from the prior year, even as net income grew 31%. That gap is depreciation and construction, and for six months the market treated it as evidence of destruction rather than investment.

What flipped in July was the arrival of a second data point. When capex guidance falls $15 billion and Azure growth guidance rises to 45%, the ratio between the two inverts. Every incremental dollar of infrastructure is suddenly carrying more revenue. That is what a return on invested capital inflection looks like from the outside, and it is why one earnings print moved a $3 trillion company 15.5% in a single session on 110 million shares — roughly 4.6 times normal volume.

COMPETITIVE ADVANTAGE

Why Microsoft might win

The moat is distribution, and it is not close. Microsoft does not have to win the model race, the chip race, or the developer-mindshare race. It has to remain the default place where an enterprise's identity, documents, endpoints, and compliance obligations already live — and then sell AI as a line item on an existing invoice.

The August 20 Bloomberg report is the cleanest possible illustration. Meta Platforms — a company that spends tens of billions building its own AI infrastructure and trains its own frontier models — has quietly become one of Microsoft's largest AI customers, paying hundreds of millions of dollars a year to access external models through Azure AI Foundry and processing trillions of tokens per week. Meta's developers are using OpenAI models on Azure to evaluate the outputs of Meta's own systems. When your most sophisticated competitor is also your customer, the distribution argument stops being theoretical.

The competition is real and should not be minimized. Google Cloud grew 82% to $24.8 billion in the June quarter, roughly double Azure's rate, and Google's TPU stack gives it a genuine cost advantage on inference that Microsoft cannot match with merchant silicon. AWS grew 37% — slower than Azure but off a larger base, and Amazon has been aggressively courting OpenAI itself. Azure sits in the middle: not the fastest grower, not the biggest, but the only one that arrives pre-installed with Excel, Teams, Entra ID, and a procurement contract that renews itself.

Microsoft's relationship with OpenAI cuts both ways and deserves its own honesty. Microsoft holds a 26.79% fully-diluted economic interest, and OpenAI has contractually committed to $250 billion of Azure compute. But the April 27, 2026 restructuring ended exclusivity, capped the revenue-share payments, and eliminated Microsoft's right of first refusal on OpenAI's cloud capacity. What Microsoft retained is a four-month first-mover window on new model releases and the exclusive path for customers who want OpenAI models wrapped in Microsoft's security and productivity stack. That is a weaker position than 2023 — and still a better one than any competitor holds.

FINANCIAL DEEP DIVE

The numbers: what they mean and what to watch

Metric                     Value
-------------------------  -------------------
Last Close (Aug 21, 2026)  $483.24 (+0.43%)
52-Week Range              $349.20 - $553.72
Market Cap                 $3.59 trillion
P/E (TTM)                  26.8x
Price / Book               8.08x
TTM EPS (FY26)             $17.28
Dividend / Yield           $3.64 / 0.74%
FY26 Revenue               $331.8B (+17.8%)
FY26 Net Income            $133.7B (+31.3%)
Q4 FY26 Net Margin         39.7%
Q4 FY26 EPS                $4.74 vs $4.23 est.
FY26 Free Cash Flow        $66.99B (-6.5%)
Avg Volume (30-day)        35.6M shares

Start with the beat, because the pattern is what matters. Q4 FY26 earnings of $4.74 came in 12.1% above the $4.23 consensus — the eighth consecutive quarterly beat, and the largest of the eight. Revenue of $90.01 billion grew 17.7% year over year. Net income of $35.77 billion on a 39.7% net margin means Microsoft converted nearly forty cents of every revenue dollar into profit while simultaneously running the largest construction program in corporate history.

Now sit with the uncomfortable number. Free cash flow fell 6.5% to $66.99 billion in a year when net income rose 31.3%. The delta is capital expenditure — $41 billion in the June quarter alone, up 70%, with roughly two-thirds of it going to short-lived assets, primarily CPUs and GPUs. Short-lived is the operative word. GPUs depreciate on a schedule measured in a handful of years, not the decades you get from a building. Every dollar spent there must earn its return fast or it becomes an expense with extra steps.

The bookings picture requires the same care. RPO of $678 billion grew 84%, which is a genuinely stunning headline — until you notice that excluding OpenAI, the growth was 25%, and that commercial bookings grew 18% ex-OpenAI versus only 10% including it. Read that inversion carefully: the OpenAI contract is so large it drags down the growth rate of the rest of the book on a comparison basis while inflating the absolute total. Concentration risk in one customer, on one balance sheet, that itself burns cash. Microsoft returned over $43 billion to shareholders in FY26 including $22.3 billion of buybacks, so the cash generation is not in doubt. The question is whose money is funding the demand.

One more number, and it may be the most important one in this piece. Per Fintel, 8,827 institutions now report holdings in Microsoft totaling 7.14 billion shares. In May 2026 — weeks before the $349.20 low — that figure sat at 5.41 billion. Institutions have added roughly 1.73 billion shares, a 31.9% increase, into and out of the worst drawdown Microsoft has suffered since 2000, with 1.13 billion of those shares added in the most recent month alone. The largest holders are the names you would expect: BlackRock, Vanguard, State Street, Geode, Fidelity, JPMorgan, and Morgan Stanley. Retail was arguing about the capex line. The institutions were buying the panic.

MANAGEMENT TEAM

Who is running this?

Satya Nadella has been CEO since February 2014 — roughly twelve and a half years, and long enough that his tenure now includes a full cloud transition, a full AI transition, and two separate 30%-plus drawdowns in the share price. The stock has appreciated approximately 1,000% under him. He is also Chairman, which concentrates authority at exactly the moment the company is making the largest capital allocation bets in its history.

His fiscal 2025 compensation totaled $96.5 million, up 22% year over year, of which roughly $84.2 million was performance stock tied to total shareholder return, Azure growth, and Microsoft Cloud revenue. That is a defensible structure — the metrics are the ones investors actually care about — but it is worth noting that Azure growth as a comp metric creates an incentive to buy revenue with capex, which is precisely the behavior the market punished in June.

Direct ownership is thin. Nadella controls roughly 0.01% of shares outstanding: approximately 790,000 shares held directly plus another 109,720 through family LLCs and generation-skipping trusts. That is over $400 million of stock and also, in percentage terms, effectively nothing. Alignment here runs through the annual grant, not through a founder's stake — normal for a mega-cap, but a different kind of skin in the game than you get from an owner-operator.

BULL CASE

5 reasons this stock could dramatically outperform

1. The capex-to-revenue ratio inverts and stays inverted. If calendar-2026 capex lands near the revised $175 billion while Azure compounds at 45%, free cash flow re-accelerates hard in FY27 off a depressed $66.99 billion base. The market has not priced an FCF recovery — it has priced an FCF decline continuing. A single quarter of expanding free cash flow alongside 40%-plus Azure growth would force a re-rating on the same earnings.

2. Copilot seat growth compounds off a small base. Thirty million paid seats sounds large until you set it against roughly 400 million Microsoft 365 commercial seats. That is under 8% penetration on a product that added 10 million seats in one quarter and grew 50,000-plus-seat customers sevenfold. At $30 per seat per month, every additional 10 million seats is roughly $3.6 billion of near-pure-margin annual revenue. The path from 30 million to 100 million is arithmetic, not invention.

3. Analyst targets sit meaningfully above the price. Consensus across 56 analysts is a Strong Buy with an average target of $569.56 — about 18% above the current quote. JPMorgan raised its target from $550 to $625, and Wells Fargo went from $650 to $700 on August 12. The high end of the range reaches $870. When the sell side is chasing a stock upward after a 38% move, the flows tend to follow the revisions.

4. Meta is the template, not the exception. If a company with Meta's internal AI capability is paying Microsoft hundreds of millions annually for external model access through Azure Foundry, the addressable market for "AI models as a metered utility" is far larger than the enterprise-adoption curve suggests. Every well-capitalized technology company becomes a candidate Azure AI customer regardless of what it builds internally.

5. The old high is only 13% away. The stock closed October 28, 2025 at its all-time high and sits 12.7% below the $553.72 intraday peak today, despite fundamentals that are materially better than they were then. Azure has crossed $100 billion since. Copilot has 50% more seats. Net income is up 31%. Simply returning to the prior high on improved fundamentals is a 15% move that requires no multiple expansion at all.

BEAR CASE

5 reasons this could disappoint

1. The RPO number is a concentration problem wearing a growth costume. Take OpenAI out and the 84% becomes 25%, and bookings growth goes from 10% to 18% — meaning the single largest contract in the book is diluting the underlying trend. OpenAI has committed to $250 billion of Azure compute while burning cash at extraordinary rates. If that commitment is renegotiated, extended, or impaired, the most impressive number in Microsoft's release deteriorates immediately.

2. Google Cloud is growing at twice Azure's rate. 82% versus 43%. Google's vertically-integrated TPU stack gives it structural cost advantages on inference that Microsoft, buying merchant silicon, cannot replicate. Google Cloud took share from 14% to 15% in a single quarter. If that continues for four more quarters, the "Azure is winning AI" thesis needs rewriting, and Microsoft's premium multiple goes with it.

3. Two-thirds of the capex is short-lived assets. GPUs and CPUs depreciate on aggressive schedules. Microsoft is committing roughly $175 billion of calendar-2026 capital against demand curves nobody can forecast beyond a few quarters. If AI workload growth decelerates even modestly in FY28, the company will be carrying enormous depreciation against a revenue line that stopped compounding — and margins compress from 39.7% quickly.

4. Free cash flow already declined once. FY26 free cash flow fell 6.5% to $66.99 billion. That happened in the best demand environment the company has ever seen. The bear question is simple: if FCF falls when everything goes right, what does it do when something goes wrong? Microsoft pays a $3.64 annual dividend and bought back $22.3 billion of stock in FY26. Those commitments compete directly with the buildout.

5. The stock has already made the move. From $349.20 to $483.24 in under two months is +38.4%, and the rally has been fading — the stock closed at $506.06 on August 10 and has drifted 4.5% lower since, on volume running below its 30-day average. Buying a mega-cap after a 38% rip, at 26.8x, into an October 28 earnings print, is a very different risk profile than buying it in June was.

VALUATION

Building the model: Bear, Base, Bull

Scenario  Key Assumption                            Price Target  Probability
--------  ----------------------------------------  ------------  -----------
Bear      Azure slows to 30%, capex re-accelerates  $370          25%
Base      Azure ~40%, FY27 EPS ~$20.25 at 28x       $567          50%
Bull      Azure holds 45%+, FCF inflects, 33x       $710          25%

The probability-weighted expected value works out to $553.50 — roughly 14.5% above the August 21 close of $483.24, and within a dollar of the 52-week high. That is not a screaming bargain and it is not a trap. It is a fairly-priced compounder with genuinely asymmetric tails, and the asymmetry is the point: the bear case takes you to $370, a level the stock already visited eight weeks ago and recovered from, while the bull case requires only that management does in FY27 what it just demonstrated it could do in Q4 FY26.

The base case deserves scrutiny because 28x is where the argument actually lives. Microsoft has traded between roughly 20x and 38x forward earnings over the past five years. At 26.8x trailing today, the market is assigning almost no premium for a business that just compounded net income at 31%. Twenty-eight times forward on $20.25 is not an aggressive assumption — it is the assumption that Microsoft remains an ordinary mega-cap. If the free cash flow inflection arrives, that number is too low. If OpenAI's commitments wobble, it is too high.

KEY METRICS TO WATCH

The five signals that tell you if the thesis is working

01 — Azure growth rate, reported October 28. Management guided to approximately 45% in constant currency. Green flag: 44% or better, confirming acceleration is structural. Yellow flag: anything in the high thirties, which would mean the June quarter was the peak and the re-rating was premature.

02 — Free cash flow direction in Q1 FY27. FY26 free cash flow was $66.99 billion, down 6.5%. Green flag: a quarter of year-over-year FCF growth, which would validate the entire capex-discipline argument. Yellow flag: another sequential decline, which would tell you the $175 billion guide is a floor rather than a ceiling.

03 — Capex versus the $175 billion calendar-2026 guide. The reduction from $190 billion was the catalyst. Green flag: reaffirmed or trimmed again at the October print. Yellow flag: any language about "meeting demand" that implies an upward revision — that would unwind the July repricing in a session.

04 — Copilot paid seats. Thirty million now, up 10 million sequentially. Green flag: 40 million or more by the October report, sustaining the net-add pace. Yellow flag: net adds below 7 million, which would suggest the Q4 surge was a promotional pull-forward rather than a demand curve.

05 — RPO growth excluding OpenAI. The headline was 84%; ex-OpenAI it was 25%. Green flag: ex-OpenAI RPO growth holding above 25%, proving the book is broadening. Yellow flag: any deceleration below 20%, which would mean the backlog is one customer wearing a trench coat.

MY TAKE

The bottom line

I have watched a lot of narrative reversals and this one is unusually clean. In June, Microsoft was a cautionary tale about capital destruction — worst month since 2000, down 35% from the high, free cash flow falling while net income rose, and a management team that could not articulate a return on $190 billion of planned spending. In July, the same company reported a 12% earnings beat, guided Azure up to 45%, and guided capex down by $15 billion. Same business, same quarter, opposite story. The stock moved 15.5% in a session on 4.6 times normal volume, and the market has spent the three weeks since deciding whether it believes it.

What I actually think happened is that Microsoft solved the perception problem before it solved the economics problem. The capex cut is real and it is meaningful, but $175 billion is still an extraordinary number, two-thirds of it aimed at assets that depreciate on a five-year clock. Meanwhile Google Cloud is growing at 82% — double Azure's rate — and the single most impressive figure in Microsoft's release collapses from 84% to 25% the moment you remove one customer. Those are not small caveats. They are the reason this is trading at 26.8x instead of 35x.

But the distribution argument keeps winning. Meta — a company that trains frontier models on its own infrastructure — is paying Microsoft hundreds of millions a year for external model access. Copilot is under 8% penetrated on Microsoft's own installed base and just added 10 million seats in three months. Microsoft does not need to build the best model or the cheapest chip. It needs to remain the invoice that enterprises already pay, and there is no evidence anywhere in this quarter that it is losing that position.

At $483.24 you are paying a market multiple for the best distribution asset in software, 13% below its old high, with a $553 probability-weighted target and a downside that the stock already tested and recovered from eight weeks ago. That is a reasonable place to own a core position. It is not a reasonable place to be a hero.

You are not being asked to bet that Microsoft wins AI — you are being asked to bet that it stays the company enterprises write the check to, and that bet has never once been wrong.

ANALYST RATING

🟢 GROWTH BUY

Core position sizing of 4–6% with a 24-month horizon; I would add on any retest of the $440–$460 zone rather than chase into the October 28 earnings print, and I would trim into $560 if Azure guidance ever slips below 40%.

Sources & Further Reading

This is not financial advice. Always do your own research.

BMNR: The $13 Billion Ethereum Vault Trading at One Times Its Own Assets

Colby Hill — Fri, 21 Aug 2026 11:20:28 GMT

August 21, 2026 | Deep Dive by Colby

OPENING HOOK

Why this story matters right now

Here is a number worth sitting with. Seven employees.

That is the entire headcount at BitMine Immersion Technologies, a company sitting on 5.82 million Ethereum tokens worth north of $13 billion. Seven people. One vault. Roughly 4.8% of every ETH that will ever circulate, held on a single corporate balance sheet in Las Vegas.

The stock closed Thursday at $21.57, up 6.6% on the session, on 64.9 million shares — nearly double its 34.2 million average. It traded to $22.67 in Thursday’s after-hours and Friday’s pre-market as Bitcoin ripped past $71,000 and Ethereum cleared $2,290. BitMine was among the most-discussed tickers on X and StockTwits all day.

But the headline is not the rally. It is the price tag. BitMine’s market cap is $13.0 billion against a crypto-and-cash position the company pegs at $11.4 to $13.2 billion depending on the tape that hour. The stock carries a price-to-book ratio of 1.01. You are being asked to pay approximately one dollar for one dollar of Ethereum, and the staking business, the buyback, and Tom Lee are thrown in for free.

That is either the cleanest asymmetry in the digital asset space or a warning that the entire treasury-company model has stopped working. This deep dive argues it is closer to the former than the tape suggests — but the bear case here is not academic, and it deserves the same specificity as the bull case.

INVESTMENT THESIS

The core reasons to own this stock

1. The NAV floor is real and it is close. BitMine trades at roughly 1.01x book, against a sector median that has historically run well above 3x. Every dollar of ETH appreciation flows to book value nearly one-for-one, and at parity the stock functions as a leveraged-optionality wrapper on Ethereum rather than a speculative premium you have to grow into. Downside is anchored by an asset with a live, liquid, 24-hour price.

2. The staking business turned the pile into a cash flow. MAVAN — the Made in America Validator Network — has over 5.06 million ETH staked and is throwing off $212 million to $299 million in annualized staking revenue depending on the seven-day yield window. A treasury company that earns yield on its treasury is structurally different from one that just holds. This is the single most underappreciated line in the story.

3. The buyback is enormous relative to the float. Management has authorized $4 billion and has already retired roughly 19 to 21 million shares, including 11.6 million since July 1 — which the company claims is the largest common-stock repurchase completed by any Ethereum or Bitcoin treasury company. With 603.2 million shares outstanding and a 599.9 million float, buying back at or below NAV is accretive to ETH-per-share by definition.

4. Institutions have been buying the entire decline. Fintel data shows 550 institutional holders now own 304.1 million shares — up from 200.3 million in June. That is roughly 104 million shares added in two months while the stock fell more than fifty percent from its 52-week high. BlackRock, Vanguard, Morgan Stanley, Geode, ARK, Susquehanna, and Jane Street are all on the register.

BUSINESS OVERVIEW

What BitMine actually does

BitMine Immersion Technologies is, functionally, a publicly traded Ethereum vault with a yield attachment. The company describes itself as focused on acquiring, holding, and actively managing ETH as its primary treasury reserve asset, with a secondary business in digital ecosystem consultancy and advisory. It was founded in 1995, is headquartered in Las Vegas, employs seven people, and is classified by the exchanges as a financial conglomerate rather than a technology company. That classification is more honest than most.

The legacy immersion-cooled mining operation that gave the company its name is being wound down. What replaced it is a strategy Tom Lee, the Fundstrat co-founder who chairs the board, has been executing with unusual discipline: buy ETH relentlessly, stake nearly all of it, and when the stock trades below the value of the coins, buy the stock instead. Chief Executive Chi Tsang, who came out of HSBC and the digital asset world, runs the operating side.

The mechanics are simple enough to fit in a sentence. BitMine issues equity when the stock trades at a premium to net asset value, converts the proceeds into Ethereum, stakes that Ethereum through MAVAN to generate a running yield, and repurchases shares when the stock trades at a discount. It is a capital allocation machine with one asset and two levers, and right now the discount lever is the one being pulled.

INDUSTRY OPPORTUNITY

The digital asset treasury shakeout is separating survivors from casualties

The digital asset treasury company — the DAT — went from novelty to crowded trade to graveyard in about eighteen months. At the peak, dozens of small caps announced crypto treasury pivots and were rewarded with premiums of three, four, even ten times the value of the coins they held. That premium was the entire business model: issue stock high, buy coins, repeat.

Then the premiums collapsed. Research from NYDIG on how DATs fail identifies the mechanism precisely — when market cap falls below net asset value, the equity issuance flywheel stops turning, and companies that funded accumulation with leverage or convertible debt are forced to sell coins into weakness. Strategy itself began selling Bitcoin in 2026 to cover preferred dividend obligations, with sales priced below its own average cost basis. The premium era, as one industry piece put it, is over.

That is the macro backdrop, and it cuts both ways for BitMine. The bad news is that BitMine is trading at parity because the whole category re-rated. The good news is that BitMine is one of the few names in the category that can survive parity — because it has staking income, a funded buyback, and a balance sheet that is not levered against the asset. The shakeout is not a threat to a company with cash flow and no debt maturity wall. It is a consolidation opportunity.

COMPETITIVE ADVANTAGE

Why BitMine might win

The moat is scale plus yield, and the two compound each other.

Start with scale. At 5.82 million ETH, BitMine is the largest corporate Ethereum holder on the planet by a wide margin and is closing on management’s stated target of 5% of total supply. That position is not replicable — anyone attempting to build it today would move the price against themselves the entire way up. It also confers something subtler: BitMine’s stake is large enough that it constitutes a meaningful supply overhang removal from the tradeable float of Ethereum itself, which is a structural tailwind to the very asset it holds.

Now add yield. This is what separates BitMine from Strategy, the archetype of the category. Strategy holds Bitcoin, an asset that generates nothing, and has therefore been forced to sell coins to service preferred dividends. BitMine holds Ethereum, an asset that pays roughly 2.75% to 2.83% to stake, and has built its own validator network to capture that yield domestically rather than renting infrastructure. $212 million to $299 million of annualized staking revenue against zero debt service means BitMine can hold indefinitely without touching principal. Strategy structurally cannot. That is the whole ballgame.

Against smaller Ethereum treasury peers, BitMine’s advantages are index inclusion — Russell 1000 membership opens institutional flows the microcaps cannot access — and the credibility of the ARK and Pantera backing on the register. Against the ETFs, BitMine offers something a spot ETH ETF cannot: a staking yield, an active buyback, and the ability to trade below the value of the underlying.

FINANCIAL DEEP DIVE

The numbers: what they mean and what to watch

Metric                      Value
--------------------------  -----------------------
Last Close (Aug 20, 2026)   $21.57 (+6.6%)
Pre-Market (Aug 21)         $22.67
52-Week Range               $12.80 - $65.60
Market Cap                  $13.00B
Price / Book                1.01x
P/E (trailing)              -2.53
Shares Outstanding          603.2M
Float                       599.9M
Avg. Volume (30d)           34.5M
Aug 20 Volume               64.9M
ETH Held                    5.82M (~4.8% of supply)
ETH Staked                  5.06M+
Annualized Staking Revenue  $212M - $299M
Buyback Authorized          $4.0B
Institutional Shares Held   304.1M (550 holders)
Employees                   7

The trailing P/E of -2.53 is the number most likely to be misread, so handle it carefully. It is negative because mark-to-market losses on the ETH position ran through the income statement during the drawdown from $65.60 to $12.80. It tells you almost nothing about the operating business and everything about where Ethereum traded last quarter. For a treasury company, price-to-book is the only valuation ratio that means anything, and at 1.01x the answer is that you are paying par.

Volume is the tell right now. Thursday’s 64.9 million shares against a 34.5 million thirty-day average is the second consecutive session of roughly double-normal turnover — Wednesday printed 75.7 million on a +8.4% move from $18.28 to $20.24. Two days, 140 million shares, and an 18% advance off the base. That is not retail chatter. That is size being put to work.

What to watch next is the gap between book value and price. The stock bottomed at $12.80 on June 30 and has climbed roughly 68% since, but it remains 67% below its October 2025 high of $65.60. The entire re-rating from here is a function of two variables: where Ethereum trades, and whether the market is willing to pay any premium at all above NAV for a company that stakes and buys back. At parity, you are getting the second variable for nothing.

MANAGEMENT TEAM

Who is running this?

Tom Lee is the reason this stock has a following, and it is worth being clear-eyed about what that means in both directions. As co-founder of Fundstrat Global Advisors and one of the most visible bulls in financial media for a decade, Lee brings distribution, credibility with institutional allocators, and a genuine research pedigree. He has been publicly explicit about the thesis — setting a $2,500 Ethereum target and arguing that tokenization and agentic AI are breaking the multi-year ETH-versus-BTC downtrend. He chairs the board. He is not a passive name on a letterhead.

He is also a risk factor. A treasury company whose valuation is partly a function of its chairman’s media presence carries key-person exposure that does not appear on any balance sheet.

Chi Tsang is the chief executive and runs operations. His background spans traditional banking at HSBC and the digital asset sector, which is the correct résumé for a business that is fundamentally a capital markets operation wearing a technology company’s name. The ARK Investment Management and Pantera positions on the shareholder register provide a layer of sophisticated oversight that most microcap-turned-megacap treasury vehicles lack entirely.

The alignment question is the honest gap in this section. Insider equity ownership and compensation structure at BitMine are less transparent than one would want for a company of this size, and any investor sizing a position should treat that as an open item rather than a settled one.

BULL CASE

5 reasons this stock could dramatically outperform

1. Ethereum reprices and the whole thing re-rates one-for-one. With 5.82 million ETH on the balance sheet and roughly 603 million shares, every $100 move in Ethereum is worth approximately $0.96 per BMNR share in book value. Lee’s $2,500 target against a current $2,290 print is modest by his standards. A move to $3,000 adds roughly $4.1 billion to book — over $6.80 per share — before any multiple expansion.

2. The NAV discount closes. BitMine sits at 1.01x book with some sources marking it nearer 0.80x. The category median has run above 3x in better tapes. Getting back to even a modest 1.3x premium is a 30% move with no change whatsoever in the underlying asset price. That is pure multiple arithmetic, and it is the fastest path to a re-rating.

3. The buyback consumes the float. $4 billion authorized against a $13 billion market cap means management is empowered to retire roughly 30% of the company at current prices. They have already taken out 11.6 million shares since July 1. Every share retired below NAV mechanically increases ETH-per-share for everyone who stays.

4. MAVAN scales into a standalone earnings story. Staking revenue of $212 million to $299 million annualized on a $13 billion market cap is a 1.6% to 2.3% yield on the equity today. If BitMine begins reporting MAVAN as a distinct operating segment — and if it eventually stakes third-party ETH alongside its own — the market gets a reason to value the company on earnings rather than on assets. That is the single highest-leverage narrative shift available.

5. Institutional accumulation continues at this pace. Institutions added roughly 104 million shares between June and August, taking their total to 304.1 million — over half the float. If that rate persists through the September and Q4 13F cycles, the available supply for retail and for the buyback compresses simultaneously, and the stock gets tight.

BEAR CASE

5 reasons this could disappoint

1. Ethereum goes down and takes the entire thesis with it. This is not a diversified business. It is one asset. Ethereum fell far enough between October 2025 and June 2026 to drive BMNR from $65.60 to $12.80 — an 80% drawdown. There is no operating segment, no recurring subscription revenue, and no product cycle to cushion the next one. A retest of the June low is a 41% loss from Thursday’s close and requires nothing more than ETH giving back this month’s rally.

2. The DAT model breaks below NAV and stays broken. NYDIG’s research on how these companies fail is blunt: a market cap under NAV is an existential threat, because the equity issuance flywheel that funds accumulation stops turning entirely. BitMine is sitting exactly at that line. If the stock spends 2027 at 0.8x book, the company cannot grow ETH-per-share through issuance and is left with only the buyback — which eventually exhausts.

3. Dilution history is real and the market remembers. BitMine got to 5.82 million ETH by issuing an extraordinary amount of stock. Share count went from a microcap base to 603.2 million shares. Investors who bought at $65.60 in October 2025 have been diluted and marked down simultaneously. Nothing prevents a return to issuance the moment a premium reappears, and the disclosed strategy explicitly contemplates it.

4. The staking yield is not contractual. MAVAN’s $212 million to $299 million range moves with Ethereum’s protocol-level validator yield, which has compressed steadily as more ETH is staked network-wide. A drop from 2.75% to 1.5% would cut staking revenue nearly in half without any management misstep. Add slashing risk, validator concentration risk, and an unsettled U.S. regulatory posture toward staking-as-a-service, and the yield line is softer than it reads.

5. Key-person and governance concentration. Seven employees managing $13 billion. A chairman whose public profile is a load-bearing element of the equity story. Limited transparency on insider alignment and compensation. Any of those alone would be a discount factor. Together they justify the market refusing to pay a premium — and they are the reason the NAV gap may not close on the timeline bulls expect.

VALUATION

Building the model: Bear, Base, Bull

Scenario  Key Assumption              Price Target  Probability
--------  --------------------------  ------------  -----------
Bear      ETH to $1,600; 0.85x book   $13.50        30%
Base      ETH near $2,300; 1.1x book  $24.00        45%
Bull      ETH to $3,000; 1.4x book    $39.00        25%

The probability-weighted expected value works out to roughly $25.60 — about 19% above Thursday’s $21.57 close and near Friday’s pre-market print. That is a positive expected value, but it is not the eye-watering asymmetry the sell side implies with its $31.87 to $37.33 consensus targets across three published notes.

What the table actually shows is more interesting than the average. The bear case is bounded by an asset with a real, liquid price — you are not modeling a business going to zero, you are modeling Ethereum going to $1,600 and the market applying a modest discount to a coin pile. The bull case requires two things to happen at once: the asset appreciates and the market agrees to pay a premium for the wrapper. That conjunction is why I weight it at only 25%. The single most important sentence in this entire model is that the downside is priced against hard assets and the upside requires a change of sentiment. Those are very different kinds of risk, and the market is currently pricing them as if they were the same.

KEY METRICS TO WATCH

The five signals that tell you if the thesis is working

01 — Price-to-book ratio. This is the master signal. Above 1.2x is a green flag that the market is paying for the wrapper and the equity issuance flywheel can restart. Sustained below 0.90x is a yellow flag that the DAT discount is structural, not cyclical, and the buyback is the only remaining lever.

02 — ETH-per-share. Track total ETH held divided by shares outstanding. Rising is the entire point of the buyback and the only honest measure of whether management is creating value. Any quarter where ETH-per-share falls — meaning issuance outran accumulation — is a yellow flag regardless of what the stock price did.

03 — Weekly buyback disclosure. Management has been reporting repurchases with unusual frequency. Continued execution at or above the recent pace of roughly 11.6 million shares per seven weeks is green. A pause, especially while trading below book, signals either a cash constraint or a change of strategy and should be treated as yellow.

04 — MAVAN annualized staking revenue. Watch the seven-day yield figure the company publishes. Holding at or above $250 million annualized is green. A slide toward $150 million on protocol yield compression removes the single feature that most distinguishes BitMine from every Bitcoin treasury company and would be a serious yellow flag.

05 — Institutional share count in quarterly 13F cycles. The 304.1 million share figure across 550 holders is the accumulation thesis in one number. Continued growth through the next filing window confirms the smart money is still buying the discount. A decline of more than 10% in aggregate institutional shares would be the clearest possible signal that the thesis has broken.

MY TAKE

The bottom line

I have looked at a lot of digital asset treasury companies over the last two years and most of them were the same trade dressed in different tickers — take a shell, announce a crypto pivot, harvest the premium, dilute the buyers who showed up late. BitMine is a different animal, and the difference is not Tom Lee’s media presence. It is the staking revenue.

Every Bitcoin treasury company on earth has the same terminal problem: the asset generates nothing, so the moment the premium disappears, the only way to fund obligations is to sell coins. Strategy is living that problem right now. BitMine holds an asset that pays roughly 2.75% to stake, built its own validator network to capture it, and generates $212 million to $299 million a year without touching principal. That converts an unsustainable structure into a sustainable one. It is the difference between a company that can wait out a bear market and a company that gets liquidated by one.

What I cannot get past is the concentration. Seven employees. One asset. One chairman carrying a meaningful share of the equity narrative. A -2.53 trailing P/E that exists because the balance sheet is a mark-to-market instrument. This is not a business you underwrite the way you underwrite a software company or an industrial — it is a leveraged sentiment expression on Ethereum with a yield kicker and a buyback, and it should be sized exactly that way.

The 1.01x book is what makes it interesting rather than reckless. You are not paying up for a story. You are paying par for the coins and getting the staking network, the $4 billion buyback authorization, the Russell 1000 access, and Tom Lee’s Rolodex thrown in. That is a genuinely favorable place to take a swing — as long as you accept that the swing is on Ethereum, not on BitMine.

At one times book, you are not betting on the company — you are betting on Ethereum with a staking yield and a buyback attached, and the only real question is whether you wanted that bet in the first place.

ANALYST RATING

⚡ SPECULATIVE BUY

A 1–2% starter position with a 12–18 month horizon, sized so that a retest of the $12.80 June low is survivable rather than catastrophic. Add only on a sustained move above 1.2x book with ETH-per-share confirmed rising.

Sources & Further Reading

Tom Lee’s Bitmine now owns 4.8% of Ethereum supply after latest ETH purchase — CoinDesk

Bitmine Immersion Stock Moves Higher as Ethereum and Bitcoin Rally — Benzinga

Tom Lee’s Bitmine Launches MAVAN, Instantly Becomes World’s Largest Ethereum Staking Platform

How DATs Die — NYDIG Research

BMNR Institutional Ownership — Fintel.io

This is not financial advice. Always do your own research.

Nebius Group ($NBIS) Q2 2026 Earnings Call Review

Colby Hill — Fri, 21 Aug 2026 00:29:17 GMT

Reported: Wednesday, August 12, 2026, before the open | Call: 8:00 a.m. ET

Speakers: Arkady Volozh (Founder & CEO), Dado Alonso (CFO), Gili Naftalovich (IR), Marc Boroditsky (CRO), Andrey Korolenko (Chief Product & Infrastructure Officer), Roman Chernin (Chief Business Officer), Tom Blackwell (Communications)

Overall Grade: A

1. THE SCORECARD

Headline vs. Consensus

METRIC                    ACTUAL       CONSENSUS/PRIOR   RESULT
Group revenue             $582.3M      $105.1M LY        +454% YoY - BEAT
Revenue QoQ               +46%         $399.0M in Q1'26  Sequential growth
Nebius AI Cloud revenue   $574.9M      98% of group      +514% YoY
ARR (exit June)           $3.00B       $1.92B exit Mar   +56% QoQ / +598% YoY
Adjusted EPS              $(0.12)      Est. $(0.82)      BIG BEAT (+$0.70)
GAAP EPS                  $(0.68)      $2.11 LY          LY had $597.4M gain
GAAP net loss             $(190.4)M    +$502.5M LY       Not comparable
Group adj. EBITDA         $236.2M      $(21.0)M LY       $257M SWING
Group adj. EBITDA margin  41%          32% in Q1'26      +900bps QoQ
AI Cloud adj. EBITDA      $285.7M      -                 50% SEGMENT MARGIN
Operating cash flow       $2.3B        -                 Positive, scaling
CapEx                     $5.7B        -                 GPUs + DC fit-out
Cash & equivalents        $8.0B        -                 Pre-July facility

FY2026 Guidance - All Reaffirmed, One Raised

METRIC                      FY26 GUIDE        ACTION
Year-end ARR                $7.0B - $9.0B     REAFFIRMED (2.3x-3.0x from $3.0B)
Group revenue               $3.0B - $3.4B     REAFFIRMED
Group adj. EBITDA margin    ~40%              REAFFIRMED
CapEx                       $20B - $25B       REAFFIRMED
Year-end connected power    800MW - 1GW       REAFFIRMED
Year-end contracted power   5 GW              RAISED

Stock Reaction

SESSION               CLOSE      MOVE
Aug 11 (pre-print)    $193.23    -
Aug 12 (report day)   $259.20    +34.1%
Aug 13                $255.04    -1.6%
Aug 14                $277.68    +43.7% vs. pre-print
Aug 19                $223.90    +15.9% vs. pre-print
Aug 20 (intraday)     ~$219      Round-tripped half the pop

The tell: the market repriced NBIS 34% higher in a single session, pushed it to $277 within two days, then gave back more than half of it over the following week. That is not the market rejecting the quarter - the quarter was excellent. It is the market discovering that a company guiding to $20-25 billion of CapEx trades on the AI infrastructure financing narrative, not just on its own results.

Grade rationale (A): Revenue up 454%, ARR up 598% and 56% sequentially, adjusted EBITDA swinging $257 million to a 41% group margin, AI Cloud at a 50% segment margin, $2.3 billion of operating cash flow, more than $9 billion of customer prepayments landing this year, four contracts averaging over $1 billion each, an auction clearing 15% above any price the company had ever charged, and every guidance line reaffirmed with contracted power raised to 5 GW. No line in this print got worse. The only things standing between this and an A+ are a $20-25 billion CapEx obligation that still has to be financed and a year-end ARR target requiring more than a doubling in six months.

2. REVENUE COMPOSITION AND THE THREE DEAL TYPES

SEGMENT              Q2'26 REV    % OF TOTAL   NOTE
Nebius AI Cloud      $574.9M      98.0%        +514% YoY
TripleTen            $10.0M       1.7%         -19% YoY, exiting U.S. legacy
Avride / other       ~$(2.6)M     ~0.3%        Early stage, dilutes margin
TOTAL                $582.3M      100%         -

The more useful framework Volozh laid out is not by segment but by deal type, because it explains the pricing power:

DEAL TYPE                   DURATION      YIELD / MW      ROLE
Midterm AI cloud            1-3 years     $20M - $25M     Core. 4 closed in Q2,
                                                          >$1B each. Prepay covers
                                                          50-60% of the CapEx.
Short-duration premium      Up to 6 mo    $40M - $50M+    Time-boxed training runs,
                                                          post-training sprints.
Long-term investment-grade  Multi-year    -               Financing instrument.
                                                          Backs $40B backlog debt.

The four landmark deals were with Reflection, Cohere, a scaled U.S. neolab, and a large U.S.-based quantitative trading firm. Marc Boroditsky was explicit that none were inbound: “All of these wins were earned wins, not inbound walk-ins. We had multiple engagement cycles before closing.” In each case the customer already had a supplier - in some cases a hyperscaler - and added Nebius for next-generation capacity.

3. MARGIN AND UNIT ECONOMICS

METRIC                       Q2'26      Q1'26      Q2'25
Group adj. EBITDA            $236.2M    $129.5M    $(21.0)M
Group adj. EBITDA margin     41%        32%        n/m
AI Cloud adj. EBITDA margin  50%        -          -

The nine-point sequential jump in group margin came from three places, and only one of them is raw capacity: higher utilization on existing infrastructure, the early contribution of the asset-light model, and Token Factory plus the recent acquisitions. Alonso was clear that the gap between the 50% AI Cloud margin and the 41% group margin is Avride and TripleTen - two early-stage businesses diluting the consolidated number.

The forward margin comment is the one to file away. Alonso: “We have visibility into pricing and expect capacity coming online from our own data center to begin improving margins in the second half of next year.” Owned data centers carry better economics than colocation, and that shift is a 2027 event.

4. BALANCE SHEET, FUNDING AND CAPEX

ITEM                        Q2'26        DETAIL
Operating cash flow         $2.3B        Positive and scaling
Capital expenditures        $5.7B        GPUs, GPU hardware, DC expansion
Cash & equivalents          $8.0B        End of period
Customer prepayments FY26   >$9B         ~70% of Q2 deals had prepayment
ATM equity issued           $2.8B gross  12.7M shares @ ~$224 avg
ATM remaining               12.3M sh     As of June 30
Asset-backed facility (Jul) $775M        SOFR + 250bps, IG customer cash flows
Committed backlog           $40B         Borrowing base for repeat deals
Corporate-level debt        ~None        Flagged as untapped capacity

The funding stack is the most underappreciated part of this quarter. A company guiding to $20-25 billion of CapEx normally has a financing problem. Nebius is covering it in layers: operating cash flow first, then more than $9 billion of customer prepayments that reduce external need dollar for dollar, then asset-backed debt against a $40 billion committed backlog, and only then equity. The $2.8 billion ATM issuance at an average $224 per share is real dilution - roughly 4.7% of shares outstanding - but it was executed near the top of the stock’s range rather than into weakness.

5. SIX-QUARTER TREND

QUARTER    REVENUE    YoY      QoQ     EXIT ARR   ADJ. EBITDA
Q2 2026    $582.3M    +454%    +46%    $3.00B     $236.2M (41%)
Q1 2026    $399.0M    +684%    +75%    $1.92B     $129.5M (32%)
Q4 2025    $228M      +547%    +56%    $1.20B     -
Q3 2025    $146M      +355%    +39%    $551M      -
Q2 2025    $105.1M    -        +107%   -          $(21.0)M
Q1 2025    $50.9M     -        -       -          -

What the table shows that a single quarter cannot:

Revenue has grown more than eleven-fold in five quarters, from $50.9 million to $582.3 million, with no sequential decline at any point. ARR has gone from $551 million to $3.0 billion in three quarters. And adjusted EBITDA crossed from a $21 million loss to a $236 million profit inside twelve months - the profitability inflection is not a projection, it already happened.

The one honest caveat: sequential revenue growth decelerated from +75% to +46%. That is a function of when capacity comes online, and management said capacity deployed late in Q2 begins contributing in Q3. TTM revenue stands at $1,355.3 million.

6. BACKLOG, CAPACITY AND PRICING

METRIC                       Q2'26 / LATEST    DETAIL
Committed backlog            $40B              Investment-grade contracts
Contracted power (year-end)  5 GW              RAISED this quarter
Connected power (year-end)   800MW - 1GW       Reaffirmed
Capacity build rate 2027     >1 GW / year      "One of just a few companies"
Auction clearing price       +15%              Above any price ever charged
Auction vs. Blackwell pipe   +20%              Above current pipeline pricing
Deal prepayment coverage     50-60% of CapEx   On the four landmark deals

A backlog of $40 billion against TTM revenue of $1.36 billion is not a normal ratio, and it is worth being precise about what it means: this is contracted commitment from investment-grade counterparties that Nebius can borrow against, not revenue that converts on a predictable schedule. Its primary function today is collateral.

The auction is the more interesting disclosure. Nebius took a small slice of capacity to market and let buyers set the price. It cleared 15% above anything the company had previously charged and 20% above its own Blackwell pipeline pricing. In a market where, as Boroditsky put it, “we have several buyers for every GPU,” that is real-time evidence of pricing power rather than a management assertion about it.

7. WHAT MANAGEMENT ACTUALLY SAID

Volozh on the core strategic choice - build ahead, sell late

“We build capacity ahead of contracts and use our multi-tenant cloud and our software stack to support our customers... Our strategy is working. We choose when to sell, to whom we sell and on what terms, and how we finance everything.”
“But most importantly, we could sell today our entire 2027 capacity on these terms if we wanted to. But we are not doing this. We see that we can achieve higher value by retaining some capacity to serve shorter-term and immediate client needs.”

This is the single most important passage on the call. Nebius is telling investors it is deliberately leaving 2027 revenue uncontracted because it believes pricing goes higher. That is a bet, and it is the opposite of what a company worried about demand would do.

Volozh on the capacity ambition

“Our future capacity pipeline effectively makes Nebius one of just a few companies in the world able to build more than 1 gigawatt of new capacity a year. And we plan to do so in 2027.”

Volozh on why new entrants do not worry him

“The whole AI cloud market is growing massively from hundreds of billions of dollars per year to probably a trillion... the market as a whole is growing by tens of gigawatts per year. It’s like multiple of order larger. And the hyperscalers are those who build a lot of this. But even then, they definitely cannot build all of it... somebody needs to build the rest of the market, and that will be us.”
“New players coming to the market don’t change the market for us. They actually just, I think, validate the market.”

Volozh on the asset-light model

“After we announced our asset-light model, we have had dozens of inquiries from potential partners who have significant capacity and enough capital but don’t know how to build and how to sell it, with GPU increasingly becoming an investable asset... By having the right tech platform and access to the market, we can help them.”

Partners finance, build and operate; Nebius supplies the platform and the demand. High-margin revenue with minimal balance sheet capital. Management called it early-stage but flagged it as a 2027 unlock.

Alonso on the financing philosophy

“Our approach is not about making a call on the markets. Really, our approach is to match the right financing instrument with the right assets, while remaining disciplined about 3 things: our cost of capital, minimizing shareholder dilution and maintaining a strong and healthy balance sheet.”

Chernin on why open models drive the inference business

“Open-source models are valuable, not simply because they are available, but because they can be tuned, trained and eventually optimized for a particular use case of a particular customer, a particular business. What has changed is that the quality of open models is improving rapidly.”

8. THE Q&A THAT MATTERED

Nebius runs Q&A through an IR-moderated portal rather than an open operator queue, so questions are attributed by analyst and firm but read by IR. These are the exchanges that moved the analytical picture.

Ryan Lantz, Morgan Stanley - on 2027 pricing and when guidance arrives

Alonso: “You should expect the deals we closed in this quarter with more than $20 million per megawatt deals and less than 2-year payback period to start coming online from late Q4 and onwards. This can serve as the baseline for pricing early next year as we could have sold out our planned capacity already today. However, we are choosing not to do so, reflecting our confidence with respect to future pricing dynamics.”

Why it matters: a sub-two-year payback on $20M+/MW contracts is the most useful modeling input on the call. It tells you the capital cycle turns fast enough that the $20-25 billion CapEx number is less alarming than it looks in isolation. Formal 2027 guidance comes later this year.

Arsenije Matovic, Wolfe Research - on debt appetite in a volatile market

Alonso: “The $775 million facility we completed in July, priced at SOFR plus 250 basis points was backed by deployed GPU’s infrastructure and contracted cash flows from an investment-grade customer. It demonstrates that even in a more volatile market, there is a strong demand to finance these contracted cash flows on attractive terms.”

Why it matters: SOFR+250 on GPU-backed paper is a genuinely good print, and it establishes that Nebius can term out its build against contracts rather than diluting shareholders every quarter.

Tyler Radke, Citi - on how 2027 capacity gets allocated

Boroditsky: “Our current emphasis is taking care of existing customers, followed by new logos and then terms, terms in order of priority, including price, then upfront prepayment, then duration... We have also, at the same time, tactically shortened how far in advance we sell capacity. So in effect, selling closer to deployment, which, at the end of the day, has improved the pricing that we can derive while preserving our agility.”

Why it matters: “Selling closer to deployment” is the operational expression of the pricing bet. It raises realized price and it raises risk - if the market softens between build and sale, that unsold capacity sits.

Alex Duval, Goldman Sachs - on confidence in the capacity ramp

Korolenko: “Almost all of that power will come online over the next 3, maybe to 3.5 years out of the contracted power... the vast majority of our contracts is the cloud contracts, where we have the flexibility where within the region, we can deliver it, reducing the dependency on any one location. So over-provisioning the sites, over-provisioning the capacity is our highest priority.”

Why it matters: the 5 GW contracted-power figure is a three-plus-year pipeline, not a 2026 asset. Do not model it as near-term revenue.

Brett Knoblauch, Cantor - on the gap between power and revenue

Korolenko: “Connected power represents the data centers, and there are a few steps coming from the data centers to the revenues. You have to commission the data center, build the network, build the clusters, deploy the platform, then onboarding the customers, and then the revenue generation starts. So that takes a few months... there is a gap between the connected power and the revenues coming.”

Why it matters: the most important caveat management volunteered. The 800MW-1GW of connected power by year-end largely monetizes across the first half of 2027, not in Q4 2026. It is also the mechanical reason the $7-9 billion year-end ARR target can coexist with a $3.0-3.4 billion revenue guide.

Ryan Lantz, Morgan Stanley - on the New Jersey site

Blackwell: “Obviously, there’s a broader set of issues and debate around data centers in the U.S. These are things that we track very closely... DataOne is seeking final approval for the amendment to their originally approved site layout plan. And the amendment... was a result of the decision to switch from the project’s power source to Bloom.”

Why it matters: the Vineland public hearing adjourned without a vote. Management framed it as routine and said the Bloom fuel-cell switch improves the project, but local permitting is now a live risk category for every U.S. data center developer.

9. CAPACITY: PROGRAM STATUS

ITEM                     STATUS
Vineland, NJ (DataOne)   Building shell finished early summer; engineering
                         fit-out progressing. Power source switched to Bloom
                         fuel cells; site-layout amendment awaiting approval
                         after a hearing adjourned without a vote. Management
                         says no significant timeline impact. Part of the 2026
                         connected-power plan, not 2027.
GPU generation           Vera Rubin already in Nebius labs and hitting expected
                         results. Step up from Grace Blackwell described as
                         technically easier than the prior transition.
                         Deployment begins late 2026 / early 2027.
Power mix                Blend of grid and behind-the-meter; hundreds of MW of
                         behind-the-meter generation already secured.
Acquisitions integrated  Eigen AI ($331.4M), Tavily ($189.4M), ClarifAI ($97.4M)
                         - all folded into the Token Factory inference effort.
Tavily developer base    2.5M+ developers, up from 1M in February.

10. VALUATION

METRIC                        VALUE
Share price (Aug 20 intraday) ~$219
Market capitalization         ~$58.6B
Shares outstanding            271.9M
Cash & equivalents (Q2 end)   $8.0B
TTM revenue                   $1,355.3M
P/S (TTM)                     ~43x
EV (approx., net of cash)     ~$50.6B
EV/Sales (TTM)                ~37x
EV / exit ARR ($3.0B)         ~17x
EV / FY26 revenue midpoint    ~16x
EV / FY26 year-end ARR guide  ~6x - 7x
P/B                           5.9x
52-week range                 $62.01 - $299.86
From 52-week high             -27%
From 52-week low              +253%

The valuation on Nebius is unusually sensitive to which metric you anchor on. On trailing sales it trades near 37x EV/Sales, expensive by any conventional standard. On exit ARR it is roughly 17x. On the company’s own year-end ARR guidance it is 6-7x - unremarkable for a fast-growing infrastructure business.

That spread is the entire debate. If Nebius lands $7-9 billion of ARR by December, today’s price is not demanding. If ARR ends the year nearer $5 billion because capacity slipped a quarter, the multiple roughly doubles and the stock has a problem. Everything rests on the deployment schedule.

11. RISKS AND WHAT TO WATCH

Near-term (next one to two quarters)

Q3 ARR against the $7-9 billion year-end target. To stay credible the exit-Q3 number needs to be comfortably north of $4.5 billion. This is the single most important disclosure at the November print.

The connected-power-to-revenue lag. Korolenko was candid that it takes months. Watch for any language extending that gap.

Vineland permitting. A stalled layout amendment would be the first concrete evidence that U.S. data center politics can move Nebius’s schedule.

CapEx pacing against the $20-25 billion guide. Q2 spent $5.7 billion; the back half has to accelerate substantially.

Medium-term

Financing mix. Prepayments and asset-backed debt are the shareholder-friendly sources. Heavy ATM use at lower prices would be a meaningfully worse outcome than the $224 average achieved in Q2.

The 2027 pricing bet. Management is deliberately holding capacity back on the view that prices rise. If AI compute pricing softens in 2027, unsold capacity becomes a margin problem rather than an upside option.

Customer concentration. Four deals over $1 billion each is a strong quarter and also a concentration risk.

Structural

GPU depreciation and generational transitions. Vera Rubin deploying from late 2026 raises the question of residual economics on the Blackwell fleet.

Competition from hyperscalers who are simultaneously partners, customers and competitors.

The gap between GAAP and adjusted results. GAAP EPS was $(0.68) against adjusted $(0.12). Track both.

12. BOTTOM LINE

Nebius delivered the cleanest quarter of any name in this cohort, and the market’s reaction said as much - a 34% single-session gain, then $277 within two days.

The operating case: revenue up 454% to $582.3 million, AI Cloud up 514%, ARR up 598% to $3.0 billion with 56% sequential growth, adjusted EBITDA swinging $257 million year over year to a 41% group margin and a 50% segment margin, $2.3 billion of operating cash flow, more than $9 billion of customer prepayments funding the build, a $775 million asset-backed facility at SOFR+250, four contracts averaging over $1 billion, an auction clearing 15% above any prior price, contracted power raised to 5 GW, and every guidance line reaffirmed.

The bear case: $20-25 billion of CapEx still has to be financed, GAAP EPS was $(0.68), the ATM added roughly 4.7% dilution, the year-end ARR target requires more than doubling in six months, connected power takes months to convert into revenue, four billion-dollar customers is real concentration, and local permitting is now a genuine variable.

The synthesis is that Nebius has crossed from a capacity story into a cash-generating business while still carrying the capital intensity of a capacity story. What separates this quarter from the last several is that the profitability inflection is behind it rather than ahead of it - and management is confident enough about 2027 pricing that it is refusing to sell capacity it could sell today.

For investors: at roughly 6-7x guided year-end ARR the valuation is not the problem. The deployment schedule is. Every material risk in this story routes back to whether capacity lands on time, and the November ARR print is where that gets settled. Nothing about the AI demand picture is in question here. The execution question is entirely about megawatts.

Grade: A

This is not financial advice - for informational and educational purposes only.6. BACKLOG, CAPACITY AND PRICING

SoFi Technologies ($SOFI) Q2 2026 Earnings Call Review

Colby Hill — Thu, 20 Aug 2026 16:57:49 GMT

Reported: Wednesday, July 29, 2026, before the open | Call: 8:00 a.m. ET

Speakers: Anthony Noto (CEO), Chris Lapointe (CFO)

Overall Grade: A-

1. THE SCORECARD

Headline vs. Consensus

METRIC                    ACTUAL       CONSENSUS/PRIOR   RESULT
Adjusted net revenue      $1.20B       ~$1.15B           +40% YoY - BEAT
GAAP revenue              $1,218.7M    $854.9M LY        +42.6% YoY
Adjusted EPS              $0.12        Est. $0.11        BEAT (incl. -$0.05 tax)
Adjusted net income       $160M        -                 13% margin, 11th
                                                         straight profitable qtr
Adjusted EBITDA           $358M        Est. ~$333M       +44% YoY, 30% margin
GAAP net income           $156.6M      $97.3M LY         +61% YoY
Total members             15.8M        11.7M LY          +35% / +1.1M in qtr
Total products            24.4M        17.2M LY          +42% - RECORD GAP
Cross-buy rate            51%          35% LY            Existing-member share
Total originations        $14.8B       -                 RECORD
Total deposits            $45.5B       $40.2B QoQ        +$5.3B sequential
Net interest margin       5.98%        5.94% QoQ         +4bps
Tangible book value/share $7.34        $4.72 LY          +56%
Total capital ratio       18.8%        10.5% required    Large excess

Guidance - Revenue Raised, Profit Held

METRIC                   FY26 GUIDE           ACTION
Adjusted net revenue     $4.75B - $4.85B      RAISED (30% -> 32-35% growth)
Adjusted EBITDA          ~$1.6B               MAINTAINED (33-34% margin)
Adjusted EPS             ~$0.60               MAINTAINED (assumes 22% tax)
Effective tax rate       22%                  +700bps vs. original guide
Rate assumption          1-2 HIKES in 2026    Was 2 cuts at guidance set

Stock Reaction

SESSION                CLOSE      MOVE
Jul 28 (pre-print)     $16.74     -
Jul 29 (report day)    $15.25     -8.9%  (intraday low $14.88 = 52-WK LOW)
Jul 30                 $16.47     +8.0%
Jul 31                 $16.31     -1.0%
Aug 4                  $18.70     +11.7% vs. pre-print
Aug 20 (intraday)      ~$18.14    +8.4% vs. pre-print

The tell: SoFi printed records on nearly every operating line, raised revenue guidance, and made a 52-week low the same morning. Six sessions later it was 12% above where it started. That round trip is the clearest evidence that the sell-off was a positioning and rate-narrative event rather than a verdict on the quarter.

Grade rationale (A-): Adjusted net revenue up 40%, adjusted EBITDA up 44% at a 30% margin, the eleventh consecutive profitable quarter, record $14.8 billion of originations, the first $10 billion personal loan quarter, student loans up 170%, cross-buy at 51% versus 35% a year ago, tangible book value per share up 56%, charge-offs improving 70 basis points, and a revenue guidance raise. The minus is for holding EBITDA and EPS guidance flat while raising revenue - a deliberate choice to reinvest, but one that leaves the earnings ramp unchanged - and for a tax rate 700 basis points above plan that cost $0.05 of quarterly EPS.

2. SEGMENT DETAIL

SEGMENT                  Q2'26 NET REV   GROWTH        NOTE
Lending                  ~$649M          -             Record originations
Financial Services       $466M           +29% YoY      Deposits + interchange
Technology Platform      $85M            +13% QoQ      New clients, Money core
Fee-based share of total 39%             -             Capital-light mix rising

The composition matters more than the total. Fee-based revenue now contributes 39% of the mix - the capital-light portion that supports a platform multiple rather than a bank multiple. Interchange alone hit $44 million, up 67% year over year, on more than $28 billion of annualized spend across money and credit card products.

The Technology Platform segment is the one to watch. It grew 13% sequentially and began onboarding SoFi Money onto its own cloud-native banking core - effectively using itself as the reference customer for a product it wants to sell to large institutions. That is a credible proof point, and it is the segment with the highest long-run multiple if it scales.

3. ORIGINATIONS - THE RECORD QUARTER

PRODUCT                Q2'26        GROWTH          NOTE
Personal loans         $10.7B       -               First $10B quarter
Student loans          $2.7B        +170% YoY       Highest in company history
Home loans             $1.4B        +74% YoY        1/3 from home equity
TOTAL ORIGINATIONS     $14.8B       -               RECORD
  On balance sheet     $11.7B       -               Drives future NII
  Loan platform (LPB)  $3.1B        -               Capital-light, fee-based

Noto explained the split between balance sheet and loan platform in a way that is worth internalizing, because it is the single most-questioned capital allocation decision at SoFi:

“When people ask like why are you putting loans in the balance sheet, why are you not doing everything through loan platform business, it’s because we want to make sure we have revenue in the future that’s visible that can deliver no matter what, and it’s 100% in our control. We layer loan platform business on top of that, not just in PL, but now SMB and closed-end seconds, and it’s a very incremental additive return on top of something that’s very visible.”

He also offered a credit data point that rarely gets airtime: “over $5 billion of cash net interest income from what we generated in Q1 of 2024 through Q2 of ‘26... it’s 2x greater than what the noncash revenue was recorded as a premium over time, which shows that the loans are really performing.”

The Loan Platform Business expanded beyond personal loans into small business and home equity, including a three-year, $3 billion agreement with Basepoint Capital.

4. CREDIT AND THE BALANCE SHEET

METRIC                        Q2'26      PRIOR QTR    NOTE
Personal loan net charge-off  3.7%       4.4%         -70bps, ex-delinquent
                                                      loan sales
Total deposits                $45.5B     $40.2B       +$5.3B sequential
Net interest margin           5.98%      5.94%        +4bps
Personal loans marked at      104.7%     105.4%       -71bps on higher
                                                      discount rate
Total capital ratio           18.8%      -            vs. 10.5% minimum
Target long-run capital       Low-to-mid teens        Room to deploy
Tangible book value/share     $7.34      $4.72 LY     +56%

Two things here cut in opposite directions. Charge-offs improved 70 basis points sequentially to 3.7%, which is the single best credit datapoint SoFi has posted in this cycle and directly rebuts the thesis that a 69% origination ramp is being bought with looser underwriting.

Against that, Lapointe flagged the fair value mark honestly: “our personal loans were marked at 104.7%, down 71 basis points from the prior quarter. This was driven by an increase in the discount rate, which was up due to higher benchmark rates.” That is a rate effect, not a credit effect, but it is a real headwind to reported revenue if rates keep rising.

The $5.3 billion deposit increase is quietly one of the most valuable items in the print - it displaces higher-cost warehouse funding and is the mechanism behind a 5.98% net interest margin.

5. EIGHT-QUARTER TREND

QUARTER    REVENUE      YoY       QoQ      NET INCOME   NET MARGIN
Q2 2026    $1,218.7M    +42.6%    +10.8%   $156.6M      12.9%
Q1 2026    $1,100.4M    +42.6%    +7.3%    $166.7M      15.2%
Q4 2025    $1,025.1M    +39.6%    +6.6%    $173.5M      16.9%
Q3 2025    $961.6M      +37.9%    +12.5%   $139.4M      14.5%
Q2 2025    $854.9M      -         +10.8%   $97.3M       11.4%
Q1 2025    $771.8M      -         +5.1%    $71.1M       9.2%
Q4 2024    $734.1M      -         +5.3%    $332.5M      45.3%*
Q3 2024    $697.1M      -         -        $60.7M       8.7%

* Q4 2024 net income was inflated by a one-time tax valuation
  allowance release and is not comparable.

What the table shows that a single quarter cannot: revenue has grown for eight consecutive quarters without a single sequential decline, from $697.1 million to $1,218.7 million - a 75% increase. Growth has been remarkably steady in the high-30s to low-40s percent range rather than decelerating.

The honest caveat is on the right-hand side. Net margin peaked at 16.9% in Q4 2025 and has declined for two straight quarters to 12.9%. Some of that is the higher tax rate; some is deliberate reinvestment. Either way, the pattern of revenue accelerating while margin compresses is what the market reacted to. TTM revenue stands at $4,305.8 million.

6. THE PRODUCTIVITY LOOP AND SOFI PLUS

METRIC                    Q2'26      PRIOR         NOTE
Members                   15.8M      11.7M LY      +35%
Products                  24.4M      17.2M LY      +42%
Products added vs members  2.0x      -             First time double
Cross-buy rate            51%        35% LY        Of new products
Invest products           3.9M       +38% YoY      1% match for Plus members
SoFi Plus paid members    206,000    0 (relaunch)  85% existing members
SoFi Plus annualized rev  $24M       -             One quarter post-relaunch

This is the section that justifies the “everything app” framing. Products grew at double the rate of members for the first time - meaning SoFi is now monetizing its existing base faster than it acquires new users, which is the entire economic argument for the model.

Noto gave the most concrete evidence of the flywheel working on a mature cohort:

“I looked back at the Q1 2021 cohort. Of the money members in the Q1 2021 cohort that became money members that quarter, with the launch of SoFi Plus, we increased the product per member of that cohort by one product in the quarter, which is, if you think about it, that’s almost 5 years old, that customer that was acquired, and they added one product for the entire cohort in the quarter.”

Adding a full product per member to a five-year-old cohort in a single quarter is a genuinely unusual result, and it is the strongest single data point in the release.

On SoFi Plus specifically, Noto set an aggressive marker: “I’d be disappointed if we are not at 1 million SoFi Plus members generating annual revenue of $120 million a year from now.” That is a 5x subscriber target on a product one quarter old.

7. WHAT MANAGEMENT ACTUALLY SAID

Noto on why revenue went up but profit guidance did not

“We don’t want to over-optimize for the actual EPS that we hit or the actual EBITDA that we hit relative to these bigger growth opportunities... And while we couldn’t be more positive about the demand for our products, the performance that we’re driving with them, the credit trends we’re seeing, the spending trends we’re seeing as well as the investing trends. All of those things are green and up and to the right.”

Noto on the rate environment - the underappreciated macro shift

“We entered the year expecting rate cuts. We’re now expecting 2 rate increases. We’ve not only maintained our guidance for revenue and profitability, we’ve exceeded expectations on the revenue side... If there are no rate increases, we probably have upside to the bottom line. But we have the cushion in our guidance to be prepared to deliver in regardless of the environment.”

This is a bigger deal than it sounds. SoFi built its 2026 plan assuming two cuts, now assumes one to two hikes, and still raised revenue guidance. That is a four-hike swing in assumptions absorbed without a guidance cut.

Noto on the new business lines launched off-plan

“This year, we’ve launched a number of things that were never in our 2026 plan. We just launched Big Business Banking. We launched SMB. We launched SoFi USD. We launched SoFi Crypto. And we’re leaning into SoFi Plus because of how well it’s doing.”

Noto on how Big Business Banking came about

“We were looking to partner with marketplaces and market makers in Crypto and everyone we talked to said, can you please be our actual bank? Can you build a fiat and crypto banking capability that’s API-driven. And that’s where that business was born out of.”

Lapointe on capital independence

“We believe that we can operate comfortably within our target capital range without the need to raise capital under our current operating plan.”

8. THE Q&A THAT MATTERED

Dan Dolev - on why EBITDA guidance was not raised

Q: Can you walk us through the rationale of not raising the EBITDA? Results are super strong. A lot of investors are asking.

Lapointe: “We ended up raising guidance from a revenue perspective, about $100 million to $200 million... There are just too many large attractive growth areas for us to invest versus adding even more profitability. The profitability opportunity is not going to go away, but choosing not to invest today would come at the expense of capturing that growth opportunity in the future.”

Why it matters: this was the first question asked and it is the crux of the stock’s reaction. Management is explicitly choosing growth over reported earnings. That is defensible for a company compounding products at 42%, but it means investors waiting for operating leverage to appear in EPS will keep waiting.

Kyle Peterson, Needham - on loan platform capacity and capital

Lapointe: “We’re able to meet all of our contractual commitments and then some. And where you’re going to start to see more meaningful growth is in some of these other asset types... we’re self-funded through our guidance period and the medium-term operating plan that we’ve laid out.”

Why it matters: LPB demand exceeds what SoFi chooses to supply. That is a good problem, and the expansion into SMB and home equity is where incremental capital-light volume comes from.

Pete Christiansen, Citi - on the commercial products and the cross-buy flywheel

Noto: “SMB really was borne out of the fact that a large percentage of our members actually are small business operators... So the SMB business is very much synergistic to the rest of our business... We’ll obviously follow this up with checking and savings in SMB and other products that are ancillary to that. And so it will add to the flywheel.”

Why it matters: the new commercial lines are not a diversification away from the consumer base - they are being sold into it. That keeps customer acquisition cost near zero, which is the whole point of the productivity loop.

Will Nance, Goldman Sachs - on the SoFi Plus customer profile

Noto: “All of the 206,000 SoFi Plus members that we reported in the quarter, 85% of them are existing members... we’re actually seeing increases in their deposits on the back of that SoFi Plus subscription sign-up, and we’re seeing increases in AUM as well, and we’re seeing increases in spending.”

Why it matters: SoFi Plus is not acquiring new customers, it is deepening existing ones - higher deposits, higher AUM, higher spend. That is the highest-return form of growth available to the company, and it explains the aggressive 1-million-member target.

Matt Coad, Truist - on spending down excess capital

Lapointe: “We exited the quarter at an 18.8% total risk-based capital ratio, which is our binding constraint. That’s well above the regulatory limit of 10.5%... we believe that having a risk-based capital ratio in the low to mid-teens is appropriate for this business longer term.”

Why it matters: the gap between 18.8% and “low to mid-teens” is several billion dollars of deployable capital. That is the fuel behind the balance-sheet origination growth, and it means SoFi can keep this pace for multiple quarters without raising equity.

9. NEW INITIATIVES - PROGRAM STATUS

INITIATIVE            STATUS
SoFi Plus             Relaunched as paid. 206,000 members, $24M annualized
                      revenue in the first quarter. Target: 1M members and
                      $120M annualized within a year.
Big Business Banking  Launched. API-driven real-time payments for commercial
                      and crypto-market clients, settling via SoFi USD.
SoFi USD              Proprietary payment rail. Settles Big Business Banking
                      and SoFi Crypto today; Mastercard debit/credit
                      settlement beginning in the coming weeks.
SMB lending           Launched, added to the Loan Platform Business.
                      Checking and savings to follow.
Loan Platform Bus.    Expanded into small business and home equity. Three-year,
                      $3B agreement with Basepoint Capital.
Technology Platform   SoFi Money migrating onto the cloud-native banking core -
                      the reference implementation for enterprise clients.
Brand                 Multiyear Notre Dame Athletics partnership; first brand
                      on Fighting Irish jerseys across all 26 varsity sports.

10. VALUATION

METRIC                        VALUE
Share price (Aug 20 intraday) ~$18.14
Market capitalization         ~$23.1B
Shares outstanding            1,291.6M
Tangible book value/share     $7.34
Price / tangible book         ~2.5x
TTM revenue                   $4,305.8M
P/S (TTM)                     ~5.4x
P/S (FY26 guide midpoint)     ~4.8x
P/E (TTM, GAAP)               ~38.8x
Forward P/E (FY26 $0.60 adj.) ~30x
P/B                           2.1x
52-week range                 $14.88 - $32.73
From 52-week high             -45%
From 52-week low              +22%

SoFi is priced in the awkward middle - too expensive to be valued as a bank at 2.5x tangible book, not cheap enough to be an obvious platform bargain at 4.8x forward revenue and 30x forward earnings.

The bridge between those two framings is the fee-based mix. At 39% of revenue and rising, with the Technology Platform re-accelerating and SoFi Plus scaling, the capital-light share is growing. If that continues, the bank multiple becomes the wrong anchor. If lending stays the dominant engine, 2.5x tangible book is a full price.

11. RISKS AND WHAT TO WATCH

Near-term (next one to two quarters)

Whether EBITDA and EPS guidance get raised at the Q3 print. Management held them while raising revenue. A second quarter of the same would confirm that reinvestment is open-ended.

The tax rate. A 22% effective rate is 700 basis points above the original plan and cost $0.05 in the quarter. Any further drift directly hits EPS.

Rate path. Management now assumes one to two hikes. Higher benchmark rates lower the fair value mark on the personal loan book, as they did this quarter.

Charge-offs. 3.7% is a good number. The vintage written at $14.8 billion of originations will season over the next four to six quarters.

Medium-term

Capital deployment pace. Moving from 18.8% to low-to-mid teens funds a lot of balance-sheet growth, but it also removes the cushion that has let SoFi avoid raising equity.

Technology Platform concentration. The segment is small at $85 million and its growth depends on landing large institutional clients, which is a long, lumpy sales cycle.

SoFi Plus targets. A 5x member target in twelve months is a public marker management will be held to.

Structural

SoFi is still primarily a lender. Roughly 61% of revenue is not fee-based, and lending is cyclical, capital-intensive, and credit-sensitive regardless of how good the app is.

Competitive intensity in digital banking continues to rise, and customer acquisition cost is the variable that would break the productivity loop math.

Net margin has compressed for two consecutive quarters even as revenue accelerated.

12. BOTTOM LINE

SoFi delivered the strongest operating quarter in its history and made a 52-week low on the morning it reported. Both facts are real, and the six sessions that followed - a 12% recovery - suggest the market has already resolved which one mattered.

The operating case: adjusted net revenue up 40%, adjusted EBITDA up 44% at a 30% margin, the eleventh consecutive profitable quarter, a record $14.8 billion of originations including the first $10 billion personal loan quarter, student loans up 170%, deposits up $5.3 billion sequentially, net interest margin at 5.98%, charge-offs improving 70 basis points, tangible book value per share up 56%, products growing at double the rate of members for the first time, cross-buy at 51%, and a revenue guidance raise delivered while flipping the rate assumption from two cuts to two hikes.

The bear case: EBITDA and EPS guidance were held flat while revenue went up, meaning the earnings ramp did not move; the effective tax rate came in 700 basis points above plan and cost $0.05; net margin has now compressed for two straight quarters; the personal loan fair value mark fell 71 basis points on higher rates; and roughly 61% of revenue still comes from a capital-intensive, credit-cyclical lending business.

The synthesis is that SoFi has reached the point its whole strategy was built toward - products compounding faster than members, cross-buy above 50%, a five-year-old cohort adding a full product in one quarter - and management has decided to spend the resulting operating leverage on new businesses rather than bank it. Whether that is the right call depends entirely on whether Big Business Banking, SMB, SoFi USD and SoFi Plus become real revenue lines or expensive experiments.

For investors: the quarter answered the demand question decisively. It did not answer the margin question, and management explicitly declined to. Watch whether EBITDA guidance moves at the Q3 print - that is the signal that the investment cycle has a ceiling.

Grade: A-

This is not financial advice - for informational and educational purposes only.

Hims & Hers Health ($HIMS) Q2 2026 Earnings Call Review

Colby Hill — Thu, 20 Aug 2026 16:56:57 GMT

Speakers: Andrew Dudum (Co-Founder & CEO), Yemi Okupe (CFO), Mo ElShenawy (CTO), Bill Newby (Director of IR)

Overall Grade: B

1. THE SCORECARD

Headline vs. Consensus

METRIC                    ACTUAL       CONSENSUS/PRIOR   RESULT
Revenue                   $753.2M      $544.8M LY        +38% YoY - BEAT
Revenue QoQ               +23.9%       $608.1M in Q1'26  Re-acceleration
GAAP EPS                  $(0.37)      Est. $(0.05)      BIG MISS
GAAP net loss             $(86.3)M     +$42.5M LY        Incl. ~$81M one-offs
Adjusted EBITDA           $60.3M       -                 8% margin, +1pt QoQ
Gross margin (GAAP)       63.8%        76.4% LY          -1,260bps
Subscribers               2,891,000    2,439,000 LY      +19% / +300k net adds
U.S. revenue              $621.8M      -                 +16% YoY, accelerating
International revenue     $131.4M      $7.5M LY          +17x (Eucalyptus ~$40M)
Marketing % of revenue    34%          39% LY            -500bps, more efficient
Operating cash flow       $(35.9)M     -                 Weight-loss inventory
Free cash flow            $(68.2)M     $(69.4)M LY       Roughly flat
Cash + ST investments     $841M        -                 Post $225M Euc. payment

Guidance - Raised on Revenue

METRIC                  GUIDE               ACTION
Q3'26 revenue           $880M - $900M       +47% to +50% YoY
Q3'26 adj. EBITDA       $75M - $95M         10% margin at midpoint
FY26 revenue            $3.1B - $3.3B       RAISED (+32% to +41%)
FY26 adj. EBITDA        $275M - $325M       9% margin at midpoint
FY26 international rev  >= $600M            At/near breakeven by design
2030 ambition           >= $6.5B revenue    $1.3B adj. EBITDA

Stock Reaction

SESSION                 CLOSE      MOVE
Aug 10 (report day)     $31.77     Reported after the close
Aug 11                  $30.51     -4.0%
Aug 12                  $29.97     -1.8%
Aug 13                  $28.77     -4.0%
Aug 14                  $28.15     -11.4% vs. pre-print
Aug 20 (intraday)       ~$32.63    +2.7% vs. pre-print

The tell: the stock drifted down 11% over four sessions on a revenue beat and a guidance raise, then recovered all of it and more over the following week. That pattern says the initial reaction was to the headline EPS miss and the gross margin line, and the recovery came once the market worked out that $81 million of the loss was non-recurring.

Grade rationale (B): Revenue re-accelerated, domestic growth improved to 16%, international went up seventeen-fold, subscribers grew 19%, marketing efficiency improved 500 basis points, adjusted EBITDA rose sequentially to an 8% margin, and full-year revenue guidance went up. Against that: gross margin fell 1,260 basis points year over year, GAAP EPS missed by a wide margin, free cash flow was negative, and the company took a $47.5 million legal accrual for an FTC complaint it intends to fight. This is a good operating quarter wearing an ugly GAAP mask - but the margin compression is real, permanent by management’s own admission, and not a one-off.

2. REVENUE COMPOSITION

SEGMENT               Q2'26 REV    YoY         NOTE
United States         $621.8M      +16%        Growth re-accelerated
International         $131.4M      +17x        From $7.5M LY
  of which Eucalyptus ~$40M        -           Closed June 2026
  organic intl.       ~$91M        +13% QoQ    Ex-acquisition
TOTAL                 $753.2M      +38%        -

Two things here matter more than the headline. First, the U.S. business - written off by many as saturated after the compounded-GLP-1 pivot - accelerated to 16% growth. Okupe credited the March decision to expand the assortment of branded weight loss products, which added subscribers fast enough to offset revenue-recognition headwinds from moving branded offerings to a monthly cadence.

Second, international is no longer a rounding error. The U.K., Australia and Germany each generate more than $100 million annualized, with Canada on track to join. Management guided international to at least $600 million for 2026 and was explicit that it will run at or near breakeven while it prioritizes scale - which is itself a margin decision.

Specialty milestones: Hers is on track to exceed $1 billion of revenue this year, five U.S. specialties already run above a $100 million annual run rate, and testosterone is expected to become the sixth in coming quarters.

3. THE MARGIN BRIDGE - THE MOST IMPORTANT SECTION

QUARTER    REVENUE     GROSS PROFIT   GROSS MARGIN   NET INCOME
Q2 2026    $753.2M     $480.8M        63.8%          $(86.3)M
Q1 2026    $608.1M     $396.8M        65.2%          $(92.1)M
Q4 2025    $617.8M     $444.4M        71.9%          $20.6M
Q3 2025    $599.0M     $442.1M        73.8%          $15.8M
Q2 2025    $544.8M     $416.2M        76.4%          $42.5M
Q1 2025    $586.0M     $430.7M        73.5%          $49.5M
Q4 2024    $481.1M     $369.5M        76.8%          $26.0M
Q3 2024    $401.6M     $317.9M        79.2%          $75.6M

This is the table that defines the investment case. Over eight quarters Hims has grown revenue 88% and lost more than fifteen points of gross margin. The business that earned 79.2% gross margins in Q3 2024 now earns 63.8%.

Okupe did not hedge on whether this reverses:

“As branded weight loss products and international revenue become a larger portion of the business, we expect gross margins will remain below the levels we have historically achieved.”

That is management telling you the old margin structure is gone. What they argue in exchange is that unit economics stay strong and the subscriber base expands: “What matters to us is that the underlying unit economics of the platform remains strong and that each of these investments expands the base of subscribers we can serve.”

The offsetting evidence is real. Marketing fell to 34% of revenue, down 5 points year over year and 2 points sequentially. Operations and support delivered three points of sequential leverage. Adjusted EBITDA margin rose a point sequentially to 8% despite the gross margin hit. Operating leverage is genuinely offsetting some of the gross margin decline - the question is whether it keeps pace as the mix shifts further.

4. THE $81 MILLION OF NON-RECURRING COSTS

ITEM                                AMOUNT     NOTE
Acquisition / transaction costs     -          Primarily Eucalyptus close
Restructuring costs                 -          Weight-loss strategic pivot
FTC legal contingency accrual       $47.5M     Complaint filed July 29
TOTAL NON-RECURRING                 ~$81M      Hit G&A, ops and support
GAAP net loss                       $(86.3)M   Largely explained by the above

Strip the $81 million out and the quarter looks materially different. That is the correct adjustment, with one caveat: the FTC matter is a live legal exposure, not a closed cost. Okupe was direct about the posture after nearly three years of cooperation and months of settlement talks:

“Ultimately, we were not prepared to accept the terms we do not believe reflect the facts or the law. We are confident in our position and intend to defend it vigorously.”

Fighting rather than settling is defensible, but it means the $47.5 million accrual estimates an outcome that has not happened yet, and litigation costs continue.

5. BALANCE SHEET AND LIQUIDITY

ITEM                        Q2'26        DETAIL
Cash + short-term inv.      $841M        Incl. $609.8M cash equivalents
Eucalyptus upfront payment  ~$225M       Paid at June close
Operating cash flow (Q2)    $(35.9)M     Branded weight-loss working capital
Operating cash flow (TTM)   >$260M       Underlying generation intact
Free cash flow (Q2)         $(68.2)M     vs. $(69.4)M LY
Receivables facility        $400M        Converts weight-loss receivables
Convertible debt offering   >$400M       Completed post-Q1
Buyback remaining           $225M        Available, not obligated

The negative quarterly cash flow is working capital, not deterioration - branded weight loss requires paying for inventory ahead of collecting from subscribers. The $400 million receivables facility exists precisely to convert that drag into cash, and management expects to resume free cash flow generation in the second half. The convertible offering is worth flagging as future dilution.

6. AI: THE EFFICIENCY ARGUMENT

METRIC                          RESULT
Nonclinical support task cut    -50%    (pilot)
Customer engagement             3x      messaging with care teams
Cancellations                   Lower   in participating cohorts
Ops & support leverage          +3pts   sequential
AI investment payback           12-18 months (management estimate)
AI R&D lab                      Menlo Park, newly opened

Hims is making an unusually concrete claim about AI: not that it will transform the business eventually, but that a running pilot cut nonclinical support workload in half while tripling engagement. CTO Mo ElShenawy framed the structural advantage as data ownership rather than model quality:

“Our AI is bound by clinical protocols and guidelines developed by our medical experts, not the open Internet. And every AI interaction is traceable and auditable by design... We believe we are the only company operating a closed loop that includes the intake, the treatment, the follow-up and the outcome at this scale. And the system learns from all of it.”

Dudum’s version of why that matters competitively: “Our advantage is being an everyday partner to customers rather than just a chatbot that lives on their home screen.”

The financially relevant part is what happens to the savings. Okupe said explicitly that AI efficiency gains get redeployed into lower prices and international expansion rather than dropping to margin - consistent with the gross margin trajectory, and a reason not to expect a margin recovery from AI.

7. WHAT MANAGEMENT ACTUALLY SAID

Dudum on the quarter’s theme

“I’ll end by saying that this quarter has been about a single word: proof. Proof that you can build a world-class health experience at global scale and a reasonable price, proof that deeper customer relationships and a profitable business aren’t in conflict; they strengthen each other.”

Dudum on the competitive posture

“Put more plainly, we’re playing offense, and we have the balance sheet and cost structure to sustain it in a way others cannot.”

Dudum on Novo Nordisk and the Wegovy pill

“The Wegovy pill has had a staggering launch. And I think we are the largest or one of the largest players that are helping people access that pill, which has really opened up the floodgates of people who had injection fears, price point fears and concerns.”

Dudum on the long-term product vision

“This is what most would consider concierge care today. And it costs anywhere from $50,000 to $150,000 annually. I think in the simplest way, my vision is to make that same level of proactive, preventative, always-on care affordable to everybody globally for a price that is universally accessible.”

He also flagged a coming business-model change worth watching: “I think this sets up very well for a natural membership offering for Hims & Hers, as you can imagine, coming out in the coming year or 2. I think this also sets up a really powerful freemium offering.”

Okupe on why margin compression is a choice

“This compression reflects deliberate strategic action to scale the specialties and markets that we believe will drive the long-term value of our platform. We do not believe anyone else in consumer health has the scale, infrastructure and balance sheet to invest in this way.”

8. THE Q&A THAT MATTERED

Maria Ripps, Canaccord - on branded cohort retention

Okupe: “Really, I think we’re seeing retention with our expectations. Part of the beat in Q2 or large part of the beat in Q2 was a result of really the weight category amongst others being much stronger than we thought. As we saw that, that’s what gave us conviction to elevate the guidance for the rest of the year.”

Why it matters: this is the direct link between the quarter’s outperformance and the guidance raise, and it locates both in weight loss. It also means the raise is levered to a category with regulatory and partner exposure.

Ryan MacDonald, Needham - on the international playbook

Dudum: “Eucalyptus has done a wonderful job building real dominance in the U.K., Australia and Germany and newly in Japan and Canada in the weight loss category... where we think there’s real opportunity is actually bringing some of the core specialties that the Hims & Hers business have gotten very good at here domestically overseas... testosterone replacement therapy, menopause therapy for women... men’s and women’s hair, men’s and women’s sexual health.”

Why it matters: Eucalyptus is a weight-loss business in five markets. The thesis is that Hims layers its higher-margin domestic specialties onto an acquired distribution footprint. If that works, it is the most credible path back toward better blended margins.

Ryan MacDonald, Needham - on the peptide catalyst

Dudum: “Our operating model today is that we are waiting on full and final rule making from the FDA... But we think it’s really important for them to complete that process before we bring this to market.”

Why it matters: Hims will not front-run the FDA on the six PCAC-recommended peptides. Right call on risk, but it pushes the peptide opportunity out, likely into 2027.

Mark Mahaney, Evercore - on whether Eucalyptus is diluting EBITDA guidance

Okupe: “We are expecting a great acceleration on both revenue and EBITDA in the U.S. in the back half of the year... given the switch, particularly on the branded products to a monthly cadence as you start to see more cohort stack due to greater tenure of the offering on the platform. That inherently results in both more revenue and EBITDA.”

Why it matters: Mahaney was pressing on a real tension - revenue guidance went up while EBITDA guidance implies a softer margin. The answer is that management is choosing to spend the upside on price cuts and an international land-grab rather than bank it.

Craig Hettenbach, Morgan Stanley - on AI rollout sequencing

Dudum: “Patients are interacting and messaging with their care teams 3x as much with the new care operating plan and operating system that we’ve rolled out. And at the same time, the actual tasks for humans has dropped 50% when you’re talking about nonclinical tasks... We started with the women’s weight loss business because that is one of the highest engagement parts of the business.”

Why it matters: the pilot ran in the hardest category first. If results hold there, extending to simpler specialties should be easier - the opposite of how most technology rollouts are sequenced, and a point in management’s favor.

9. PIPELINE AND PROGRAM STATUS

ITEM                    STATUS
Testosterone (TRT)      Fastest-growing specialty outside weight loss.
                        Three quarters post-launch. Injectable and oral TRT
                        targeted before end of 2026. Expected to become the
                        sixth U.S. specialty above a $100M run rate.
Peptides                Awaiting final FDA rule-making after 6 of 7 cleared
                        the July PCAC hearing. BPC-157 in stability and
                        validation testing at the Menlo Park API facility.
                        U.S.-manufactured APIs.
Near-term wellness      Sermorelin, glutathione and NAD+ targeted before
                        year-end using already-permitted supply chains.
Eucalyptus integration  Closed June 2026. Largest acquisition in company
                        history. Five international markets.
Leadership              Dr. Anant Vinjamoori joined as Chief Medical Officer
                        of Hims - internal medicine, primary care, longevity.
Membership / freemium   Flagged by Dudum as likely "in the coming year or 2."

10. VALUATION

METRIC                        VALUE
Share price (Aug 20 intraday) ~$32.63
Market capitalization         ~$7.52B
Shares outstanding            233.3M
Cash + short-term inv.        $841M
TTM revenue                   $2,578.1M
P/S (TTM)                     ~2.9x
EV (approx., incl. converts)  ~$7.1B
EV/Sales (TTM)                ~2.8x
EV / FY26 revenue midpoint    ~2.2x
EV / FY26 adj. EBITDA mid     ~24x
P/B                           22.4x
P/E (TTM)                     n/m - trailing net loss
52-week range                 $13.74 - $65.30
From 52-week high             -50%
From 52-week low              +137%

At roughly 2.2x forward revenue, Hims is not priced like a high-growth consumer platform - it is priced like a company the market is unsure about. That discount reflects three things: the gross margin trajectory, the FTC overhang, and lingering doubt about how durable weight loss revenue is.

The counter-argument is that a business growing revenue 32-41% with a stated path to $6.5 billion of revenue and $1.3 billion of adjusted EBITDA by 2030 should not trade at 2.2x sales. Whether that gap is opportunity or accurate skepticism depends almost entirely on the margin question.

11. RISKS AND WHAT TO WATCH

Near-term (next one to two quarters)

Gross margin. This is the number. Management said it stays below historical levels; watch whether it stabilizes near 64% or keeps sliding as branded weight loss and international grow.

Free cash flow. Management expects to resume generating it in the second half. If Q3 is negative again, the working capital explanation weakens.

Q3 revenue against the $880-900M guide, which implies 47-50% growth and leaves little room for slippage.

The FTC litigation. The $47.5 million accrual is an estimate, not a settlement.

Medium-term

Novo Nordisk dependency. The Wegovy pill is driving a meaningful share of the acceleration. That relationship is a commercial agreement, not a moat.

Peptide regulatory timing. Management has tied itself to final FDA rule-making, which could take quarters.

Eucalyptus integration. Largest acquisition in company history, five new markets, closed two months ago.

Convertible dilution. Over $400 million of converts issued this year.

Structural

The compounded-to-branded shift permanently changed the economics. Hims used to sell compounded generics at software-like margins; it now resells branded pharmaceuticals at distribution margins on a growing share of revenue.

Competition from incumbent pharmacy, payer and retail players entering direct-to-consumer telehealth.

Subscriber growth of 19% is running well below revenue growth of 38%, meaning revenue per subscriber is doing more of the work - a lever with a ceiling.

12. BOTTOM LINE

Hims delivered a better quarter than the headline EPS suggests and a worse one than the revenue line suggests, and reconciling those two is the whole exercise.

The operating case: revenue up 38% with U.S. growth re-accelerating to 16%, international up seventeen-fold to $131 million with three markets above $100 million annualized, 300,000 net subscriber adds, marketing efficiency up 500 basis points, three points of sequential leverage in operations, adjusted EBITDA up sequentially to an 8% margin, full-year revenue guidance raised to $3.1-3.3 billion, an AI pilot cutting nonclinical support tasks in half, and $841 million of liquidity plus a $400 million receivables facility.

The bear case: gross margin fell from 76.4% to 63.8% year over year and management says it is not coming back, GAAP EPS missed badly, free cash flow was negative $68 million, the company took a $47.5 million accrual on an FTC complaint it is choosing to litigate, the growth is concentrated in a weight loss category dependent on a Novo Nordisk relationship, and revenue per subscriber is carrying more of the growth than subscriber count.

The synthesis is that Hims has traded margin for scale on purpose, and the trade is being executed competently - operating leverage is offsetting roughly half the gross margin damage, and the subscriber base and geographic footprint are both meaningfully larger than a year ago. The strategy is coherent. The risk is that it is irreversible: you cannot un-shift a mix, and if the promised efficiencies do not compound, this ends up a lower-margin business without the scale to justify it.

For investors: at 2.2x forward revenue the market is pricing real doubt, which makes the setup asymmetric if gross margin stabilizes around here. Watch gross margin and free cash flow at the Q3 print - not revenue. Revenue is not the question anymore.

This is not financial advice - for informational and educational purposes only.

Decoding GEX: What Gamma Exposure Charts Are Really Telling You

Colby Hill — Thu, 20 Aug 2026 16:06:38 GMT

If you’ve spent any time in trading circles on X or Discord, you’ve probably seen charts covered in red and green bars labeled “GEX,” along with talk of “gamma flips,” “call walls,” and “put walls.” It looks intimidating, but the core idea behind gamma exposure is actually pretty approachable once you understand a couple of building blocks first. This article walks through what GEX is, why it exists, and how to read it without treating it as a magic prediction tool, because it isn’t one.

The Options Basics You Need First

Before GEX makes sense, you need two smaller concepts: delta and gamma. Both come from the world of options trading.

An option is a contract that gives its buyer the right, but not the obligation, to buy (a call option) or sell (a put option) a stock at a specific price before a specific date. Delta measures how much an option’s price is expected to move for every $1 move in the underlying stock. An option with a delta of 0.50, for example, would be expected to gain about $0.50 in value if the stock rises $1.

Gamma measures how quickly that delta changes as the stock price moves. It’s often described as the “delta of delta.” Gamma is highest for options that are at the money (meaning the strike price is close to the current stock price) and gets smaller as an option moves further in or out of the money.

Why does this matter to you as an investor who isn’t even trading options? Because the firms on the other side of most options trades, known as market makers or dealers, have to manage the risk from all the options they’ve sold. And how they manage that risk can actually influence the stock market itself.

What Is GEX, Really?

Gamma exposure, or GEX, is an estimate of the total gamma held by dealers across all the options contracts on a given stock or index, aggregated by strike price and expiration date. In plain terms, it’s an attempt to measure how much buying or selling pressure might show up in the market as dealers adjust their hedges when prices move.

Here’s the mechanism. Market makers who sell options generally want to stay “delta-neutral,” meaning they don’t want to have a directional bet on whether the stock goes up or down. To stay neutral, they buy or sell shares of the underlying stock to offset the delta risk created by the options they’ve sold. As the stock price moves, delta changes (that’s gamma at work), so dealers have to keep adjusting their hedge. GEX is an attempt to estimate, in aggregate, how large and in which direction those hedging flows might be.

It’s important to be upfront about something here: GEX is a model, not a directly observed fact. Public options data doesn’t tell you exactly who is holding which side of a trade, so anyone calculating GEX has to make assumptions about dealer positioning. Different data providers can and do publish different GEX numbers for the same stock on the same day, because they’re using different assumptions, different expirations, and sometimes different formulas entirely.

Positive Gamma vs. Negative Gamma

The sign of the aggregate GEX reading, positive or negative, is generally considered the most important piece of information it offers.

When modeled dealer gamma is positive, the theory goes that dealers need to sell stock as prices rise and buy stock as prices fall in order to stay hedged. That kind of counter-trend flow can act like a shock absorber, potentially compressing volatility and making the stock more prone to trading in a tighter range or reverting back toward a central price.

When modeled dealer gamma is negative, the dynamic flips. Dealers may need to buy stock as prices rise and sell as prices fall, which means their hedging flows move in the same direction as the market rather than against it. That can amplify moves rather than dampen them, contributing to wider, more volatile price swings, especially during selloffs.

It’s worth repeating that this describes a sensitivity to volatility, not a directional forecast. Positive gamma doesn’t guarantee a calm market, and negative gamma doesn’t guarantee a crash. Plenty of other forces, like earnings, economic data, and news, can easily overwhelm the modeled hedging effect.

Key GEX Levels to Know

GEX charts often highlight a few specific price levels that traders watch closely.

The zero gamma level, sometimes called the gamma flip, is the estimated stock price at which aggregate dealer gamma crosses from positive to negative (or vice versa). Because it marks a theoretical shift in how dealers might behave, some traders treat it as a rough dividing line between a “calmer” regime and a more volatile one.

The call wall refers to the strike price with the largest concentration of call option gamma. It’s sometimes treated as a level where upward momentum might slow down, since heavy call positioning at that strike could prompt more dealer selling as price approaches it. The put wall works the same way on the downside, marking the strike with the heaviest put gamma concentration, sometimes eyed as a potential area of support.

None of these levels are guaranteed to hold. They’re better thought of as scenario markers, places worth paying attention to, rather than hard lines the market is obligated to respect.

How to Read a GEX Chart

Most GEX charts you’ll come across follow a similar format: a bar chart plotted against strike prices, usually with green bars representing call-side gamma and red bars representing put-side gamma, layered on top of or alongside the current stock price.

A practical way to approach one of these charts is to first check whether the overall reading is positive or negative, since that sets the volatility context for the session. From there, look for where the bars cluster most heavily. Large concentrations often line up with the call wall and put wall levels described above. Finally, check where the zero gamma level sits relative to the current price, since price crossing that level is often treated as more significant than day-to-day noise elsewhere on the chart.

It’s also worth checking whether a chart is showing gamma exposure per $1 move in price or converted to a dollar notional figure for a 1% move, since providers present this differently and the raw numbers aren’t directly comparable across platforms.

Pros and Cons of Using GEX

Like most tools that try to translate options market activity into something readable, GEX comes with real strengths and real limitations.

On the positive side, GEX can offer a window into market structure that price charts alone don’t show. It can help explain why a stock has been unusually calm or unusually choppy, since dealer hedging flows are a real, mechanical force in the market, not just theory. For active traders, GEX levels can be a useful input alongside other tools when thinking about where volatility might expand or contract. It also encourages a more structural way of thinking about markets, beyond just technical patterns.

On the downside, GEX is fundamentally a model built on assumptions, not an exact measurement, which means it can be wrong and can vary meaningfully between data providers. It says nothing about market direction on its own, only about potential sensitivity to moves, so treating a positive or negative reading as a buy or sell signal is a misuse of the concept. GEX also tends to be most relevant for large, heavily optioned names like major indices and mega-cap stocks; the signal gets noisier and less reliable for smaller or less liquid names. And because it updates with new options flow throughout the day, a GEX chart you check in the morning can look meaningfully different by the afternoon.

Where I Get My GEX Data

Calculating gamma exposure yourself would require options chain data, real-time pricing, and a fair amount of math, which is why most people rely on a platform that does the modeling for them. I personally use BullFlow.io for this. It has real-time GEX charts and heatmaps that show gamma exposure across every strike and expiration for a given stock, layered against live price action, alongside its broader unusual options flow and dark pool tracking tools.

If you want to check it out, here’s my affiliate link, and you can use code CHILL for a discount. Using that link and code helps support this newsletter at no extra cost to you, and I only share tools I actually use myself.

Where This Fits in a Broader Strategy

GEX is a market-structure tool, not a standalone trading system. It doesn’t replace fundamental research, a sound risk management plan, or basic technical analysis; it’s another lens to look through, most useful when combined with everything else you already know about a stock or the broader market.

For long-term, buy-and-hold investors, GEX is probably more of a curiosity than a decision-making tool. It’s more relevant to active or short-term traders who are managing positions around specific price levels and want context on why volatility might be picking up or settling down. Even then, treating GEX levels as absolute lines in the sand, rather than probabilistic scenario markers, is a common and costly mistake.

Putting It All Together

GEX estimates how dealer hedging flows might behave as stock prices move, based on the gamma embedded in the options market. Positive gamma readings are associated with potentially calmer, more range-bound conditions, while negative gamma readings are associated with potentially more volatile, trending conditions. Key levels like the zero gamma flip, call wall, and put wall give traders scenario markers to watch, but none of them are guarantees. Because GEX is a model built on assumptions about dealer positioning rather than a directly observed fact, it’s best used as one input among many rather than a standalone signal. Tools like BullFlow.io make it far easier to actually see this data without doing the math yourself, but understanding what the chart is showing you, and what it isn’t, matters more than which platform you use to view it.

This article is for educational purposes only and isn’t personalized financial advice. Mentions of specific indicators, platforms, or techniques are for illustrative purposes only and aren’t recommendations to buy, sell, or trade any security. This post includes an affiliate link to BullFlow.io; I may earn a commission if you sign up through it, at no extra cost to you. Options trading involves substantial risk and isn’t suitable for every investor. Past performance and historical patterns don’t guarantee future results. Everyone’s financial situation, goals, and risk tolerance are different — consider talking with a licensed financial advisor before making investment decisions.

BULL: The Rule Change Nobody Priced In

Colby Hill — Thu, 20 Aug 2026 11:29:28 GMT

August 20, 2026 | Deep Dive by Colby

OPENING HOOK

Why this story matters right now

One thousand six hundred thirty-five thousand. That is how many daily average revenue trades Webull processed in the second quarter of 2026, against 1,008 thousand a year earlier. A 62% increase in the raw number of times a customer pressed the button.

Here is what makes that number worth sitting with. Webull did not launch a viral product to get it. It did not buy a competitor. It did not spend more on marketing — it spent $14.5 million less than the quarter before. The SEC deleted the Pattern Day Trader rule, the artificial ceiling that had capped accounts under $25,000 at three day trades per week, and roughly five million Webull customers with an average balance just under $5,000 discovered they could trade as often as they wanted.

Revenue hit a record $198.8 million, up 51% year over year. Adjusted operating profit hit $62.6 million, up 169%. The company swung from a $28.3 million loss to $24.4 million of GAAP net income. Adjusted earnings came in at $0.07 against a $0.03 consensus — a 133% beat.

And the stock closed Tuesday at $8.64, down 39.75% from the $14.34 it fetched a year ago. Webull Corporation is a $4.0 billion retail brokerage that just proved its platform economics work at scale, trading at roughly five times annualized revenue, and it spent Wednesday as the single most-discussed ticker on X. The question this post exists to answer is whether the quarter is a rule-change sugar high or the moment the market’s model of this company broke.

INVESTMENT THESIS

The core reasons to own this stock

1. The operating leverage is now proven, not projected. Webull’s adjusted operating margin went from 9.3% in Q1 2026 to 31.3% in Q2. That did not happen because revenue grew — it happened because revenue grew 24% sequentially while adjusted operating expenses fell. Marketing dropped from $49.1 million to $34.6 million. Management disclosed that excluding marketing entirely, operating margin has run consistently above 40%. That is the number that matters: the platform is structurally profitable, and marketing spend is a dial, not a fixed cost.

2. The PDT repeal is a permanent structural change, not a one-quarter event. Rules do not un-delete themselves. Webull’s average account sits just below $5,000 — precisely the cohort the $25,000 PDT threshold was designed to constrain. Management explicitly described the company as a net beneficiary of account consolidation across the industry, meaning small accounts that were previously spread across multiple brokers to work around the trade limit no longer need to be.

3. Institutions have been buying the entire decline. Two hundred fifty-six institutions now hold 198.87 million shares — roughly 43% of shares outstanding and 64% of the 311.7 million free float. That position has grown from 125.4 million shares last September to 198.9 million today, a 59% increase, accumulated while the stock fell by 40%. The last month alone added 15.9 million shares. Peak6, Susquehanna, BlackRock, Jane Street, Tiger Global, and Citadel are on that list.

4. The valuation embeds failure. At $8.64 the market pays about five times annualized Q2 revenue for a business that just posted 31% adjusted operating margins, holds 28.4 million registered users and $28.5 billion of customer assets, and is licensed in 35 markets. The published analyst consensus target sits at $16.50, nearly double the last close. You do not need the bull case to work to make money here. You need the base case not to collapse.

BUSINESS OVERVIEW

What Webull actually does

Webull Corporation is a self-directed brokerage and financial technology platform founded in 2016 by Anquan Wang and headquartered in St. Petersburg, Florida. It runs a commission-free trading app for equities, options, futures, crypto, fixed income, mutual funds, and — as of this year — prediction market contracts, aimed squarely at the active retail trader who wants depth: Level 2 order-flow data, paper trading, real charting, and a genuine options chain. Robinhood is the app you learn on. Webull is the app you graduate to.

The money comes from three places, and the mix is lopsided. Trading-related revenue was $147.7 million of the $198.8 million total in Q2, up 66% year over year. Within that, options generated $88.4 million, equities $33.5 million, and platform and trading fees $25.8 million. The rest is interest income on client bank deposits — $3.8 billion of them as of July — and margin lending, where balances ran $926 million. Crypto contributed about $2.25 million, barely 1% of revenue, which is worth remembering every time someone tells you Webull is a crypto story.

The scale numbers are real. 28.4 million registered users. 5.1 million funded accounts. $28.5 billion in customer assets, up 79% year over year. Trading operations in 18 markets, licenses in 35, roughly 810,000 international funded accounts, and APAC customer assets above $5.0 billion. The company employs 1,194 people, which works out to about $167,000 of quarterly revenue per employee — a genuinely lean operation for a regulated broker-dealer.

INDUSTRY OPPORTUNITY

A regulator just removed the retail trading speed limit

The Pattern Day Trader rule was written in 2001. It required any account executing four or more day trades in five business days to maintain $25,000 in equity, or face a ninety-day restriction. For twenty-five years it functioned as the single largest structural constraint on retail trading velocity in the United States, and it fell hardest on exactly the demographic that discount brokers serve — young traders with four-figure balances.

Its removal is a volume event with no obvious precedent. Webull’s own data quantifies it: equity notional volume grew 73% and options contract volume grew 68% year over year, with June setting a record at $105.8 billion of equity notional and 89.4 million options contracts. Options DARTs went from 354 thousand in Q2 2025 to 526 thousand in Q2 2026. When the constraint on trade frequency disappears for a customer base whose median account was built specifically around that constraint, the response is not marginal.

The second dynamic is consolidation. Traders who maintained accounts at three brokers to get nine day trades a week now have no reason to. Whoever offers the best tools wins the consolidated balance, and Webull’s entire product identity is built on offering better tools than the mass-market alternative. Management named this explicitly on the call. It is the most underappreciated line in the entire release.

COMPETITIVE ADVANTAGE

Why Webull might win

The moat is the toolset, and it shows up in a single ratio: options are 60% of Webull’s trading revenue. That is not an accident of pricing — it is what happens when your product is engineered for the sophisticated end of the retail spectrum. Webull claims a top-five position in U.S. retail options, and options traders are the most valuable retail cohort in existence because they trade more often, generate higher revenue per contract, and churn less once they have learned a platform’s order-entry muscle memory.

Against Robinhood, the comparison is honest rather than flattering. Robinhood is larger, more profitable, better known, and carries a market capitalization many multiples of Webull’s; it wins the beginner, the cash-sweep customer, and increasingly the tokenization narrative. But Robinhood’s simplicity is a ceiling as much as a feature, and the PDT repeal disproportionately rewards the platform whose users were already trying to trade more. Against the incumbents — Schwab, Fidelity, Interactive Brokers — Webull is far smaller and cannot match balance-sheet scale, but it is also not carrying their cost structure or their legacy customer expectations.

The genuine differentiator emerging now is distribution breadth. No other retail brokerage of Webull’s size operates licensed trading businesses in 35 markets. The company completed its first tokenized equity trade in Latin America this quarter, announced the acquisition of Pi Securities in Thailand, and launched MCP servers connected to leading AI models — letting users research, build tools, and execute trades in plain language. That last item is the closest thing in retail brokerage to a genuine platform bet on agentic trading, and Webull shipped it before anyone larger did.

FINANCIAL DEEP DIVE

The numbers: what they mean and what to watch

Metric                     Value
-------------------------  ----------------------------
Last Close (Aug 19, 2026)  $8.64 (+8.9%)
Pre-Market (Aug 20)        ~$9.61 (+11.2%)
52-Week Range              $4.50 - $16.04
Market Cap                 $3.98 billion
Trailing P/E               Negative (TTM GAAP loss)
Price / Book               4.13x
Dividend Yield             None
Q2 2026 Revenue            $198.8M (+51% YoY)
Q2 Adj. Operating Profit   $62.6M (+169% YoY)
Q2 Adj. Operating Margin   31.3%
Q2 Adj. EPS                $0.07 vs $0.03 est.
Customer Assets            $28.5 billion (+79% YoY)
Funded Accounts            5.1 million
Avg. Volume (30-day)       10.2 million shares
Aug 19 Volume              39.2 million (3.8x avg.)
Institutional Ownership    198.9M shares (64% of float)

Start with the expense line, because that is where the quarter was actually won. Adjusted operating expenses broke down as brokerage and transaction costs of $44.3 million, general and administrative of $36.8 million, technology and development of $20.6 million, and marketing of $34.6 million. Marketing peaked at $52.8 million in Q4 2025 and $49.1 million in Q1 2026 — the two quarters when adjusted operating profit collapsed to $21.6 million and $14.8 million. Webull was buying customers at a price that destroyed its own margin. In Q2 it stopped, revenue accelerated anyway, and profit tripled. That is the single most important fact in this filing.

Now the uncomfortable half. Funded accounts have been frozen at 5.1 million since January 2026 — eight straight months of no net growth. Net deposits have fallen from $1.2 billion in December to $0.4 billion in July. Customer assets peaked at $29.3 billion in May and slipped to $27.3 billion by July. Client bank deposits went from $4.47 billion in May to $3.82 billion in July, and July equity notional volume dropped 23% from June’s record. The Q2 blowout was driven by existing customers trading harder, not by new customers arriving or by asset growth compounding.

The trailing P/E is negative, which is not a red flag so much as a timing artifact — the trailing twelve months still carry heavy share-based compensation and the losses that preceded the SPAC listing. The relevant valuation frame is forward. Annualize Q2 and you get roughly $795 million of revenue and $250 million of adjusted operating profit against a $3.98 billion market cap. That is cheap if Q2 repeats. It is not cheap if Q2 was the peak.

MANAGEMENT TEAM

Who is running this?

Anquan Wang founded Webull in 2016 and serves as chief executive. He is a former Alibaba engineer who built the platform around the conviction that retail traders were being underserved by simplified interfaces — a thesis that looks considerably better this quarter than it did in April, when the stock bottomed at $4.50. He remains the controlling intellectual force behind the product, and the company’s willingness to ship an MCP-based agentic trading layer before larger competitors reflects an engineering-first culture that starts at the top.

Anthony Denier serves as President and Group President, and he is the operator investors actually hear from — he led the Q2 call alongside Carlos Questel, Head of Investor Relations. Denier’s tenure spans Webull’s entire U.S. build-out from a startup app to a licensed multi-market broker-dealer, and he was the executive who publicly positioned the company ahead of the PDT repeal rather than reacting to it.

The alignment question is the honest weak spot. Webull came public through a SPAC in 2025, and the share-based compensation load is heavy enough to keep GAAP trailing earnings negative even after a record quarter. That is the gap between the $62.6 million adjusted operating profit and the $24.4 million GAAP result. Investors should treat the adjusted numbers as directionally correct and the dilution as a real, ongoing cost — not a footnote.

BULL CASE

5 reasons this stock could dramatically outperform

1. The Q2 margin structure repeats and the market re-rates the multiple. If Webull holds adjusted operating margin anywhere near 31% through Q3 — reporting November 19 — the annualized adjusted operating profit run rate is roughly $250 million against a $4.0 billion market cap. Two consecutive quarters at that level converts the story from “rule-change anomaly” to “structurally profitable broker,” and that re-rating alone justifies a materially higher multiple.

2. Marketing discipline holds while volumes stay elevated. The entire margin swing came from cutting marketing $14.5 million and watching revenue rise anyway. If management has genuinely found that PDT repeal supplies organic volume that previously required paid acquisition, the ex-marketing 40%+ operating margin becomes the reported margin. That is the highest-torque line item in the model.

3. International scaling turns 810,000 funded accounts into a second growth engine. APAC customer assets already exceed $5.0 billion and the Pi Securities acquisition in Thailand closes Webull’s largest remaining Southeast Asian gap. Licensed in 35 markets against a U.S. base that is currently flat, international is where the next funded-account inflection has to come from — and it is the part of the business the U.S.-focused sell-side models least well.

4. Agentic trading arrives before competitors ship it. Webull’s MCP servers connect leading AI models directly to research, tool-building, and trade execution. If plain-language trading becomes the interface retail expects over the next twelve to eighteen months, Webull owns a genuine first-mover position in a category where Robinhood’s simplicity advantage inverts into a disadvantage.

5. The float math gets violent on any good news. Institutions hold 198.9 million shares against a 311.7 million share float — 64%. Add the 39.2 million shares that traded Wednesday against a 10.2 million average, and the tradable supply is thin. The consensus target of $16.50 sits 91% above the last close. In a stock this concentrated, a second earnings beat does not move the price linearly.

BEAR CASE

5 reasons this could disappoint

1. Funded accounts have not grown in eight months. Five point one million in January. Five point one million in July. A brokerage whose account base is flat is monetizing a fixed pool harder, and there is a mathematical limit to how much harder the same five million people can trade. This is the number that turns the bull thesis into a treadmill.

2. Customer assets and deposits are going the wrong direction. Assets peaked at $29.3 billion in May and fell to $27.3 billion in July. Net deposits collapsed from $1.2 billion in December to $0.4 billion in July — a 67% decline. Client bank deposits, which fund the interest-revenue line, dropped $654 million between May and July. Trading revenue can spike on volatility; interest revenue cannot recover from a shrinking deposit base.

3. July already shows the volume spike decaying. Equity notional volume fell from June’s record $105.8 billion to $81.9 billion in July, a 23% drop. Equity DARTs fell from 1,341 thousand to 1,069 thousand. If the PDT repeal was a burst of pent-up demand rather than a permanent level shift, Q3 will show it, and the November 19 report will be measured against a Q2 that flattered expectations.

4. The GAAP-to-adjusted gap is $38 million a quarter. Adjusted operating profit was $62.6 million; GAAP net income was $24.4 million. That difference is largely share-based compensation, and it is a real transfer from shareholders to employees. Trailing GAAP earnings remain negative. Every valuation argument in this post rests on non-GAAP figures that exclude an ongoing, structural cost.

5. Robinhood is bigger, better capitalized, and moving into the same ground. Robinhood gained ground on its tokenization push the same week Webull reported, carries a vastly larger market capitalization, and offers materially cheaper margin — 4.7% to 5.75% with Gold, against Webull’s 8.74% standard rate. In a business where the product difference is a matter of interface preference, the larger balance sheet usually wins the price war.

VALUATION

Building the model: Bear, Base, Bull

Scenario  Key Assumption                          Target  Prob.
--------  --------------------------------------  ------  -----
Bear      Volumes normalize, accounts stay flat   $5.50   30%
Base      ~25% adj. margin holds, modest growth   $11.00  45%
Bull      PDT gain durable, international scales  $18.00  25%

The probability-weighted expected value works out to $11.10 — thirty percent above Tuesday’s $8.64 close, and roughly fifteen percent above the pre-market print near $9.61. That spread is real but it is not enormous, which is the honest read on this setup: the easy money was made by whoever bought at $4.50 in April, and buyers today are paying up for a quarter that is already public information.

What makes the distribution interesting is its asymmetry. The bear case at $5.50 assumes volumes fully revert and the account base stays frozen, and it still sits above the April low — because the company is now demonstrably profitable on an adjusted basis, which it was not in April. The bull case at $18.00 requires only that Q2’s margin structure survives two more quarters and that the $16.50 consensus target proves conservative. The downside is bounded by proven profitability; the upside is bounded by execution. Those are the terms of the trade.

KEY METRICS TO WATCH

The five signals that tell you if the thesis is working

01 — Funded accounts. Green flag: any month that breaks above 5.2 million after eight months at 5.1. Yellow flag: a print below 5.0 million, which would mean the PDT-driven trading surge is coming from a shrinking base.

02 — Monthly equity notional volume. June set the record at $105.8 billion, July fell to $81.9 billion. Green flag: August and September stabilizing above $85 billion. Yellow flag: anything under $75 billion, which would confirm the repeal was a one-time burst.

03 — Net deposits. The single cleanest read on whether customers are committing capital. Green flag: a return above $0.8 billion in any month. Yellow flag: two consecutive months below $0.4 billion.

04 — Marketing expense in Q3. Reported November 19. Green flag: marketing at or below $35 million with revenue flat-to-up, proving the Q2 margin was structural. Yellow flag: a snap back above $45 million, which would mean the account base only holds with paid acquisition.

05 — Institutional share count. Currently 198.9 million shares across 256 holders, up 15.9 million in the last month. Green flag: continued accumulation through the next 13F cycle. Yellow flag: any sequential decline, which would signal the smart money used the earnings pop as an exit.

MY TAKE

The bottom line

I have been skeptical of the discount-brokerage trade for most of this year, and Webull’s Q2 is the first release that made me revisit the position honestly. The reason is not the revenue beat — revenue beats on a volume spike are the least interesting thing a broker can produce. The reason is the expense line. Webull cut marketing by $14.5 million and grew revenue 24% sequentially in the same quarter. That is a company discovering that it had been paying for demand it did not need to buy.

But I am not going to pretend the operating metrics are clean, because they are not. Funded accounts have been flat for eight months, net deposits are down 67% from December, customer assets peaked in May, and July’s equity volume was already 23% below June’s record. Strip away the PDT narrative and what you have is a fixed customer base trading more frequently for reasons that may or may not persist. That is a fragile foundation for a growth multiple, and anyone buying this at $9.61 pre-market is paying for a quarter that has already been reported.

What tips the balance is the institutional data. Two hundred fifty-six institutions have taken their combined position from 125.4 million shares to 198.9 million over eleven months — a 59% increase — while the stock fell 40%. They added 15.9 million shares in the last month alone, and they now control 64% of a 311.7 million share float. Peak6, Susquehanna, Jane Street, and Citadel are market-structure firms; they understand retail order flow economics better than anyone reading this post, and they have been accumulating into weakness for the better part of a year.

The setup I see is a genuinely profitable platform, priced at five times annualized revenue, with a proven margin structure, a flat account base, a decaying volume tailwind, and a float that is 64% spoken for. That is not a clean compounder. It is a well-structured speculation where the downside is cushioned by demonstrated profitability and the upside requires only that management repeat what it just did.

Webull spent two years buying growth it could not afford; last quarter it stopped paying and grew anyway — the only question left is whether the customers stay when the tailwind stops.

ANALYST RATING

⚡ SPECULATIVE BUY

A 2–3% starter position with a 12-month horizon, sized to survive a Q3 volume miss on November 19. Add only on confirmation that funded accounts break 5.2 million or that marketing stays below $35 million — not on price momentum.

Sources & Further Reading

Webull Corp Q2 2026 Results — SEC Form 6-K

Transcript: Webull Q2 2026 Earnings Conference Call — Benzinga

Webull Q2 2026 Earnings: Record Trading Volumes Drive 51% Revenue Growth — TradingKey

Webull posts record $199M Q2 revenue, swings to profit — StockTitan

BULL Institutional Ownership Data — Fintel.io

Top X (Twitter) Stocks — Trending Tickers — AltIndex

This is not financial advice. Always do your own research.

Rocket Lab ($RKLB) Q2 FY2026 Earnings Call Review

Colby Hill — Thu, 20 Aug 2026 00:07:43 GMT

Reported: Monday, August 10, 2026, after the close | Call: 5:00 p.m. ET
Speakers: Peter Beck (Founder & CEO), Adam Spice (CFO), Murielle Baker (IR)
Overall Grade: B

1. THE SCORECARD

Headline vs. Consensus

METRIC                    ACTUAL      CONSENSUS/GUIDE    RESULT
Revenue                   $234.1M     $231.4M            BEAT (+1.2%)
Revenue growth YoY        +62.0%      --                 Record quarter
Revenue growth QoQ        +16.8%      --                 5th straight record
GAAP EPS                  $(0.08)     $(0.06)            MISS by $0.02
GAAP gross margin         36.1%       33-35% (co.)       BEAT (+110bps)
Non-GAAP gross margin     41.5%       38-40% (co.)       BEAT (+150bps)
Adjusted EBITDA           $(8.8)M     $(20)-(26)M (co.)  BIG BEAT (+$11.2M)
GAAP opex                 $142.1M     $138-144M (co.)    In range
Non-GAAP opex             $115.7M     $120-126M (co.)    BEAT (-$4.3M)
Total backlog             $2.36B      --                 RECORD, +137% YoY
Free cash flow            $(110.1)M   prior $(77.4)M     MISS - burn widened

Q3 2026 Guidance — The Reason the Stock Fell

METRIC                 Q3'26 GUIDE    Q2'26 ACTUAL   STREET      READ
Revenue                $250-265M      $234.1M        ~$252M      In line (+10% QoQ)
GAAP gross margin      29-31%         36.1%          ~37.6%      ~700bps BELOW STREET
Non-GAAP gross margin  35-37%         41.5%          --          -500bps sequentially
GAAP opex              $143-149M      $142.1M        --          Rising on Neutron
Non-GAAP opex          $121-127M      $115.7M        --          Rising
Adjusted EBITDA        $(17)-(23)M    $(8.8)M        --          DETERIORATING
Net interest income    $21M           --             --          On $2.4B cash
Basic WA shares        ~641M          ~598M o/s      --          ~7% dilution

Stock Reaction

SESSION                      PRICE      MOVE
Aug 7 close (pre-print)      $82.83     --
Aug 10 close (report day)    $80.04     -3.37%
Aug 11 pre-market low        ~$75       -8% to -9%
Aug 11 intraday low          $75.46     -13.1% from Aug 10 high of $86.83
Aug 11 close                 $80.01     -0.04%  FULL INTRADAY RECOVERY
Aug 18 close                 $79.16     -4.4% vs. pre-print

The tell: the market gapped RKLB down 8–9% on the margin guide, then bought every share of it back inside one session. That is not a market rejecting the thesis. It is a market repricing the timing of the thesis.

Grade rationale (B): Record revenue, a 137% backlog increase, over $1 billion in new bookings, and an adjusted EBITDA loss less than half the low end of guidance. Offset by a two-cent EPS miss, a Q3 gross margin guide roughly 700 bps below the Street, free cash burn accelerating to $110 million, and a Neutron milestone that quietly shifted from “launch” to “delivery to the pad.” Excellent business momentum, deteriorating near-term financial optics.

2. SEGMENT DETAIL

SEGMENT           Q2'26 REV    QoQ        % OF TOTAL
Space Systems     $189.5M      +38.6%     81.0%
Launch Services   $44.6M       -30.0%     19.0%
TOTAL             $234.1M      +16.8%     100%

The Launch decline is an accounting artifact, not a demand signal, and Spice was explicit about why:

“This decline is primarily attributable to a shift in the revenue mix between our point-in-time Electron business and our overtime HASTE business. This quarter was somewhat atypical for our launch operations as we launched HASTE missions for which a significant portion of revenue had already been recognized in prior periods under the overtime accounting method. In contrast, revenue from Electron missions is recognized at the point of time of launch.”

In plain English: HASTE revenue gets booked as the work is performed, so the launches that flew in Q2 had already been substantially recognized in Q1 and Q4. Rocket Lab flew a comparable number of missions and booked 30% less revenue for them. Do not extrapolate the Launch line.

Spice also framed the diversification benefit directly: Space Systems “provides comforting diversification alongside our robust, but at times lumpy launch business.” At 81% of revenue, Rocket Lab is now a satellite manufacturer that also happens to own a rocket company.

3. MARGIN AND OPEX BRIDGE

Gross margin drivers, Q2

DRIVER                              DIRECTION   DETAIL
Space Systems mix shift             NEGATIVE    Toward lower-margin sat platforms
Mynaric initial contribution        NEGATIVE    Pre-integration, pre-synergy
Tariff refund (non-recurring)       POSITIVE    One-time benefit
Neutron Flight 2 inventory reserve  NEGATIVE    Largely offset the tariff refund
NET RESULT                          36.1% GAAP  Still beat the 33-35% guide

Note the quality flag: the tariff refund was a one-time benefit, and management stated it was “largely offset by an inventory reserve against our Neutron Flight II launch vehicle.” The two roughly cancel, so the beat is mostly clean — but a Flight 2 inventory reserve on a rocket that has not flown Flight 1 is worth filing away.

Operating expense and headcount

LINE            Q2'26 GAAP   QoQ CHG    Q2'26 NON-GAAP   QoQ CHG
R&D             --           +$1.9M     --               +$0.8M
SG&A            --           +$7.7M     --               +$9.8M
TOTAL OPEX      $142.1M      --         $115.7M          --

HEADCOUNT       Q2'26        QoQ CHG
Production      1,688        +240
R&D             1,087        +138
SG&A            442          +61
TOTAL           3,217        +439

A 439-person increase in a single quarter — roughly 16% sequential headcount growth — is the clearest evidence of the Neutron and Space Systems ramp. Most of the SG&A increase is Mynaric consolidation rather than organic bloat.

4. BALANCE SHEET AND CASH FLOW

METRIC                          Q2'26        Q1'26        CHANGE
Operating cash flow             $(84.1)M     $(50.3)M     Worse by $33.8M
Capex (PP&E + cap. software)    $26.0M       $27.1M       Better by $1.1M
FREE CASH FLOW                  $(110.1)M    $(77.4)M     WORSE BY $32.7M
Cash + equiv + marketable sec.  ~$2.4B       --           Up on ATM proceeds
ATM equity raised in quarter    $1.08B       --           Program TERMINATED

Two things matter here. First, Rocket Lab raised $1.08 billion in equity during the quarter through an at-the-market program and then shut the program down — a sequence that reads as “we took the window while it was open, and we are done for now.” Second, the $2.4 billion cash position is explicitly earmarked: Spice said the funds “are intended to support acquisitions such as the recently announced Iridium acquisition as well as other targets in a robust M&A pipeline.”

Morgan Stanley’s Kristine Liwag pushed directly on the burn, and Spice’s answer was unusually specific about what is driving it:

Liwag: “Can you provide more color on the higher expected cash burn in the quarter? How much of this was driven by higher-than-expected Neutron development cost versus acceleration of inventory to support future launches?”
Spice: “You’ve pretty much hit the nail on the head, Kristine. So a significant amount of the spend or cash flow consumption in the quarter was driven by building out the subsequent tails for Neutron... for the parts of the rocket that we think are at low risk to needing some kind of a change as a result of the first test launch later this year, hopefully. That really is kind of informing what we’re building ahead on. And then I would say also the Mynaric acquisition... had some supply chain challenges and so forth. And so we basically had to replenish that supply chain, and that was part of the step-up in the quarter as well.”

This is the most important disclosure on the call. Rocket Lab is not burning cash on R&D overruns — it is pre-building Neutron airframes before Flight 1 has flown, betting that the design is stable enough that those units will not need rework. If the bet pays, the flight-2-through-10 ramp compresses dramatically. If Flight 1 reveals a design change, some of that inventory gets written down. It is a calculated, disclosed, and genuinely aggressive capital allocation decision.

5. EIGHT-QUARTER TREND

QUARTER    PERIOD END   REVENUE   YoY      QoQ      GROSS PROFIT   GM%     NET INCOME   NET MARGIN
Q2 2026    6/30/26      $234.1M   +62.0%   +16.8%   $84.6M         36.1%   $(49.3)M     (21.0)%
Q1 2026    3/31/26      $200.3M   +63.5%   +11.5%   $76.5M         38.2%   $(45.0)M     (22.5)%
Q4 2025    12/31/25     $179.7M   +35.7%   +15.8%   $68.2M         38.0%   $(52.9)M     (29.5)%
Q3 2025    9/30/25      $155.1M   +48.0%    +7.3%   $57.3M         37.0%   $(18.3)M     (11.8)%
Q2 2025    6/30/25      $144.5M      --    +17.9%   $46.4M         32.1%   $(66.4)M     (46.0)%
Q1 2025    3/31/25      $122.6M      --     (7.4)%  $35.2M         28.8%   $(60.6)M     (49.5)%
Q4 2024    12/31/24     $132.4M      --    +26.3%   $36.8M         27.8%   $(52.3)M     (39.5)%
Q3 2024    9/30/24      $104.8M      --    +26.3%   $28.0M         26.7%   $(51.9)M     (49.6)%

What the table shows that a single quarter cannot:

Revenue has more than doubled in eight quarters, from $104.8M to $234.1M, with only one sequential decline (Q1 2025).
Gross margin has expanded roughly 940 basis points over that span, from 26.7% to 36.1% — this is a business with demonstrated operating leverage, not a perpetual science project.
Net margin has improved from (49.6)% to (21.0)%. Losses are shrinking as a percentage of a rapidly growing revenue base, which is exactly the right shape.
But: Q2 2026 gross margin of 36.1% is down from 38.2% in Q1, and the Q3 guide of 29–31% would take margin back to roughly Q1 2025 levels. That is the two-quarter reversal the Street is reacting to.

TTM figures: revenue of $769.1M; net loss of $165.5M.

6. BACKLOG AND BOOKINGS

METRIC                            Q2'26              DETAIL
Total backlog                     $2.36B             +137% YoY
Launch share                      ~40% (~$944M)      90+ launches, company record
Space Systems share               ~60% (~$1.42B)     SDA Tranche 2 & 3 largest
12-month conversion               45.5% (~$1.07B)    Plus quick-turns business
New bookings, Q2 + post-quarter   >$1B               $437M launch, $581M Space Sys
New launches booked               26                 --

Named contract wins

CONTRACT                          VALUE     NOTES
Space Force suborbital (HASTE)    $266M     Up to 18 missions. LARGEST LAUNCH
                                            CONTRACT IN COMPANY HISTORY. Won vs.
                                            legacy primes. Flies from Kodiak, AK.
Space Force SBAMTI (Flatellite)   $397M     Multiple Flatellite spacecraft.
                                            NEUTRON WILL LAUNCH THIS MISSION.
GEO satellites (two deals)        $160M+    Three GEO birds incl. prime contract
                                            with Space Systems Command for two
                                            space-domain-awareness sats. Uses
                                            Heimdall payload from GEOST deal.
Kepler Communications             n/d       First dedicated Neutron booking by a
                                            commercial operator (prev. rideshare).
NASA                              n/d       Three Electron launches, two missions.
QPS                               n/d       Three more launches. Third bulk buy in
                                            under a year. 18 total.
Confidential defense prime        n/d       Two HASTE launches in 2027.

A backlog growing 137% while revenue grows 62% means bookings are outrunning recognition by better than two to one. Spice flagged the caveat honestly — “bookings across Space Systems and launch can be inherently lumpy due to the timing of increasingly larger high-impact program opportunities” — but the direction is unambiguous.

Operationally: 13 launches year to date with 100% mission success, on track to beat last year’s tally, across six pads at three launch sites.

7. WHAT MANAGEMENT ACTUALLY SAID

Beck on the Iridium acquisition and the strategic thesis

“It was also a milestone quarter for strategic acquisitions, having closed Mynaric and Motiv and of course, announcing our intentions to acquire Iridium, which will accelerate our future in space applications and evolve Rocket Lab into a fully integrated space powerhouse. So with that, welcome to the beginning of our new space applications era.”
“For years, we’ve talked about the space value chain in three key verticals. The first is access to space, and we have that with launch. The second is the hardware to do things in space once you’re there. That’s the satellites and their components, and we have this too. The third and final vertical is space applications, the entire reason for going to space in the first place... With Iridium, Rocket Lab will have all three of these verticals.”
“I’ve long since said that the most successful space companies will be the ones that have the keys to space, i.e., can build and launch their own satellites. Rocket Lab is one of only two companies capable of this now. By acquiring Iridium, we accelerate our entrance into the space applications market. It will take a decade or more to build out a constellation from scratch. With Iridium, we are starting with a constellation of 66 satellites relied upon by more than 2.5 million subscribers and delivering more than $870 million in annual revenue this past year.”

Deal timing: expected to close mid-2027, pending Iridium stockholder approval and regulatory review.

Beck on the launch supply shortage — the core macro argument

“Launch has never been so constrained. The Pentagon is accelerating procurement for its top priority programs, like we’ve seen with the NSSL Lane 1 tripling its ceiling from $5.6 billion to $17 billion. Military spending in Europe has increased and sovereign launch remains a hot topic globally, too. The truth is that if you want to book a launch now or especially after 2029, the options are extremely limited.”
“I’ve personally never seen launch so constrained in pretty much ever. Other launch providers are backing off and focusing on their own needs as well. So the amount of launch that’s left in the industry is really, really tight.”

Beck on the components business as a moat

“I think people think of launch is a big moat and launch. Yes, launch is a really, really big moat. But equally well, spacecraft components is a big moat because if you just turn up in the industry and say, I want 1,000 reaction wheels, then your chances of getting them are almost 0. And unless you come to perhaps us.”
“When we first announced like an old solar company in Albuquerque, you are thinking what on earth is Pete on. But you can see now that we’ve scaled that to being the largest space-grade solar manufacturer in the world. And every single satellite needs solar and tremendous amounts of it.”

Spice added the financial mechanism, which is the part most investors miss:

“Being a components provider on a merchant level is interesting. It’s a growing business for us. Most of those products have very nice margins associated with them. It’s more strategic than that for us because really... that’s really enabled to a large part because we’re able to eliminate margin stacking and just be that much more competitive in going after these large strategic programs.”

Beck on VICTUS HAZE — the execution proof point

“The mission from Space Force was clear, launch an Electron to orbit in just 24 hours, we did this in 16 hours and 42 minutes, a new record. We also designed and built the satellite that it launched. The Space Force gave us 72 hours to commission that spacecraft on orbit. We did it in 38. Then we had 84 hours to track, chase and photograph a noncooperative satellite, we did it in less than 59.”
“VICTUS HAZE was the first time that the Space Force had ever seen a single prime contractor deliver the rocket, the spacecraft and orbit operations for the same tactically responsive space mission.”

Beck on Flatellite — the underappreciated announcement

“Most of the stuff that I see in the future that Rocket Lab will do for itself will be built on the back of a Flatellite platform.”

8. THE Q&A THAT MATTERED

Thirteen analysts asked questions. These are the exchanges that moved the analytical picture.

Andres Sheppard-Slinger, Cantor Fitzgerald — on scaling past Flight 1

Q: “I want to maybe move past the first launch and talk about scale... how quickly do we think we could potentially get to 10 launches and additional scale?”
Beck: “Obviously, the first flight is extremely important, but what about flight 10 because that’s probably... apart from first flight, the second most important thing... it’s not just to get to the pad quickly for flight 1. Of course, we all want that no one more than I. But it’s really about how do we get to flight 10 in the shortest time possible. So for us, that’s really all about reusability and we’re constantly trading the timelines and the qualification criteria for the various systems and subsystems to get to the pad for flight 1, but also making sure that when it comes to flight 10 that we don’t have to go back and requalify things... our ramp that we’ve sort of explained to everyone is a 1, 3, 5 ramp.”

Why it matters: This reframes the entire Neutron schedule debate. Beck is explicitly trading first-flight speed for ramp speed. That is the correct engineering decision and the wrong optics decision — it explains why “delivery to the pad in Q4” replaced “launch in Q4,” and why the stock gapped down.

Andres Sheppard-Slinger, Cantor Fitzgerald — on Neutron pricing power

Spice: “We brought Neutron to market with a $50 million to $55 million ASP with a commitment really not to do any significant discounting for early launches. And we’ve stuck to that. We feel very good, though, about where the market is from a supply versus demand perspective. And I think right now, the view is that we see more upside to ASPs and certainly anything that are down or sideways.”

Why it matters: A pre-revenue launch vehicle with pricing power is unusual. Management is signaling ASP expansion, not the discounting normally required to fill an unproven rocket’s manifest.

Kristine Liwag, Morgan Stanley — on the path to cash flow positive

Spice: “It’s still very much a function of the timing of the first successful test launch of Neutron... We’ve been pretty consistent in pointing towards — that will be the real turning point where we go to adjusted EBITDA positivity in the quarter after that, that event happens. But then we’ve also been pretty clear that it’s going to probably be 18 to 24 months after that pivot that we get to cash flow positivity because we’ll have to continue to invest in a fleet of tails to build out for Neutrons.”
“Now, of course, pending the closing of the Iridium acquisition, the table gets reset pretty significantly, right? So as we’ve talked about that, that business generates pretty significant free cash flow.”

Why it matters: This is the single most useful modeling disclosure on the call. Management has now put a framework on the record — adjusted EBITDA positive one quarter after a successful Neutron flight; free cash flow positive 18–24 months after that — with Iridium as a potential accelerant. If Neutron flies in early 2027, that implies FCF breakeven somewhere in 2029 on a standalone basis. Anyone modeling sooner is modeling against management’s own guidance.

Ryan Koontz, Needham & Company — on the Q3 margin guide

Spice: “There was a little bit less of the, I would say, the more mature merchant products within the mix in kind of, I would say, a little bit in Q2 and actually pointing towards some of the weakness in margin in Q3, we have a pretty wide range of margins in our Space Systems business. We have some component solutions such as solar that are always going to be more towards the lower end... and then if you look towards some of the product areas, they can be more kind of north of 70 points of gross margin... And you mentioned the biggest pieces of our backlog today are for SDA Tranche 2 and Tranche 3. And again, those are more kind of in the mid-30s.”

Why it matters: This is the actual answer to “why did the margin guide collapse,” and it is not a cost problem. Rocket Lab’s Space Systems portfolio spans roughly mid-30s gross margin (SDA satellite platform programs) to north of 70% (certain components). The Q3 guide-down is a mix outcome: large, low-margin, high-visibility government platform programs are ramping faster than the high-margin merchant components business. That is a materially better problem than pricing pressure or cost overruns — but it also means margin is now hostage to program timing, and management said only that they “expect a beneficial remixing impact on gross margins as we look beyond Q3.”

Gautam Khanna, TD Securities — on competitive capacity risk

Q: “Given your large competitor is launching a much larger vehicle. Does that give you any concern on industry capacity maybe 3 or 4 years from now?”
Beck: “Pre-test flight, post-test flight, I would say Neutron demand is just not a concern. Now naturally, I think everybody will be more comfortable with buying Neutron’s post-test flight, but we’ve had absolutely 0 issues in selling full price Neutrons pre-test flights... even as new capacity comes on market from some competitors, a lot of that capacity is already spoken for, for their own internal programs, whether it be Internet or AI data centers or whatever... So I see this constrained launch market persisting for quite some time.”

Why it matters: The bear case on Neutron has always been that a much larger competitor vehicle makes medium-lift irrelevant. Beck’s counter is that competitor capacity is captive to competitor constellations, so merchant launch supply stays scarce regardless. Worth testing against actual third-party manifest data, but it is a coherent rebuttal.

Also worth noting

Ben Johnson (Piper Sandler) asked what the highest-risk remaining Neutron milestone is. Beck: “the stage testing is always the thing that gets your adrenaline running because you have fully fueled vehicles on the pad. And you’re igniting the engines for the first time. And I think you can also see for another space company that when it doesn’t go well, it really doesn’t go well.”
Erik Rasmussen (Stifel) on Mynaric margins. Spice benchmarked it to the SolAero integration, where margins started “high single digits” and were brought in line: “This won’t take nearly that long.”
Michael Leshock (KeyBanc) on M&A after Iridium. Beck: “people should not assume that Iridium is a one and done.”
Edison Yu (Deutsche Bank) on Iridium’s spectrum. Beck defended L-band over broadband: “it’s the wrong kind of spectrum for broadband... the 2 most wealthiest people in the world are going after that, and that’s quite a challenge to go and compete.”
Edison Yu (Deutsche Bank) on government program revenue shape. Spice gave a modeling gift: government programs run roughly four years on a “10, 40, 40, 10 kind of curve.”
Jeff Van Rhee (Craig-Hallum) pressed on how Rocket Lab reaccelerates Iridium’s slow top line. Beck pointed to PNT and IoT: “with a relatively modest tweak to a constellation add a whole bunch more capability... our focus is rightly so with the acquisition from day 1 will be growth, growth, growth.”
Suji Desilva (ROTH Capital) on when Neutron ordering feels “normal.” Beck: “it almost feels like that now, to be honest with you... we have a very limited supply of Neutrons coming out of the gate, and we have to be very, very careful with where we put those.”
9. NEUTRON: PROGRAM STATUS
```
COMPONENT           STATUS
Stage 1             New barrels and domes pulled from AFP; moving into assembly
Interstage          New panels into flight assembly after qual load tests done
Stage 2             Flight avionics + fluid systems install completing; then ships
                    to LC-3 Virginia for integrated fluids testing before
                    Archimedes vacuum engine mate
Thrust module       Aux tanks installed; fluids/avionics integration finishing;
                    integrated systems test ON THE PAD before mating to Stage 1
Archimedes engines  400+ HOT FIRES completed across Stage 1 and Stage 2 vacuum
                    engines. Now in durability/restart-cycle testing. Full flight
                    engine set already in production.
Hungry Hippo        Control surfaces, avionics, fluids, thermal protection
                    integrated. Preflight testing next.
TARGET              DELIVERY TO THE PAD IN Q4 2026
```
The critical language, verbatim from Beck:
“As with any complex development program, we’ve had to tweak a few things along the way, but we’re moving now into final checkout and assembly of all of our flight hardware before integrating them at the pad... Production currently lines up with the target delivery of Neutron to the pad in Q4 2026. While the window for an end year launch is narrowing, the work we’re doing now is about risk trading, balancing the timing of our first launch against how quickly and seamlessly we can scale at tenth launch.”
That clause is the real news of the quarter. Rocket Lab did not formally delay Neutron — but it moved the Q4 goalpost from “launch” to “delivery to the pad” and told you the launch window is narrowing. Treat a Q4 2026 first flight as unlikely and early-to-mid 2027 as the working assumption.
10. VALUATION
```
METRIC                                    VALUE
Share price (Aug 19 intraday)             ~$74-77
Market capitalization                     ~$45.8B
Shares outstanding                        598.4M
Guided Q3 basic WA shares                 ~641M  (~7% DILUTION)
Cash + equiv + restricted + marketable    ~$2.4B
TTM revenue                               $769.1M
P/S (TTM)                                 ~60x
EV/Sales (TTM, net-cash basis, ex-debt)   ~56x
Annualized Q3 guide midpoint              ~$1.03B
EV/SALES (FORWARD RUN-RATE)               ~42x
P/B                                       13.6x
P/E                                       n/m (TTM net loss $165.5M)
52-week range                             $37.57 (11/21/25) - $151.00 (5/27/26)
From 52-week high                         -49%
From 52-week low                          +102%
```
Rocket Lab trades at roughly 42x forward run-rate revenue with negative earnings, negative free cash flow, and a rocket that has not flown. There is no valuation framework in which that is cheap. The stock is priced on the option value of three things landing simultaneously: Neutron reaching cadence, the Iridium acquisition closing and being “supercharged,” and Space Systems compounding at 40%+ into a $2.36 billion backlog.
The 49% drawdown from the May high is worth sitting with. The market has already removed half the enterprise value from the peak while the fundamentals improved in every quarter of that period. That is a sentiment and duration repricing, not an earnings repricing.
11. RISKS AND WHAT TO WATCH
Near-term (next one to two quarters)
1. Q3 gross margin actual vs. the 29–31% guide. If it comes in at the low end or below, the mix explanation weakens and the cost explanation gains credibility.
2. Neutron stage testing. Beck named it as the highest-risk remaining milestone. A static-fire anomaly is the single most likely source of a sharp drawdown.
3. Free cash burn. Q2 was $(110.1)M and management guided to “elevated levels” continuing. Watch whether Q2 was the peak.
4. Launch segment revenue normalization. The -30% sequential decline should reverse as Electron point-in-time recognition returns to the mix. If it does not, the accounting explanation was incomplete.
Medium-term
1. Neutron Flight 2+ inventory risk. Rocket Lab is pre-building airframes before Flight 1 flies. A design change post-flight means write-downs.
2. Iridium closing risk and integration. Mid-2027 close requires Iridium stockholder approval plus regulatory review. Iridium is a slow-growth business — Jeff Van Rhee pressed on exactly this, and Beck’s answer was aspiration (”growth, growth, growth”) rather than a quantified plan.
3. Dilution. $1.08B ATM in one quarter and a guided share count 7% above current shares outstanding. The ATM is terminated, but the M&A pipeline Beck described implies more capital raises.
4. Mynaric integration. Acquired out of an insolvency process with supply chain damage. Spice declined to disclose current margins.
Structural
1. Program concentration. SDA Tranche 2 and Tranche 3 are the largest backlog components and carry mid-30s margins. Government program timing now materially drives consolidated gross margin.
2. Competitive launch capacity post-2029. Beck’s argument that competitor capacity is captive is plausible but unproven.
12. BOTTOM LINE
Rocket Lab delivered the best operating quarter in its history and the stock fell 8% pre-market on it. Both facts are defensible.
The operating case: record revenue up 62%, gross margin above guidance, adjusted EBITDA loss less than half the low end of guidance, non-GAAP opex below guidance, backlog up 137% to $2.36 billion, over $1 billion in new bookings including the largest launch contract the company has ever signed, 13 launches at 100% success, and a pending acquisition that would add $870 million of high-margin recurring revenue and make Rocket Lab one of two companies on earth that can build, launch, and operate its own constellation.
The bear case: a two-cent EPS miss, a Q3 gross margin guide roughly 700 basis points below the Street, adjusted EBITDA guided to worsen sequentially, free cash burn accelerating 42% quarter over quarter to $110 million, 7% embedded dilution, and a Neutron first flight that quietly slipped out of 2026 in all but name.
The synthesis is that Rocket Lab’s revenue quality is improving while its revenue mix is temporarily working against reported margin, and that the entire equity story still hinges on a single vehicle reaching a launch pad. Management was unusually candid about all of it — Spice put a dated framework on the path to profitability, and Beck said out loud that the end-of-year launch window is narrowing rather than letting investors find out in November.
For investors: nothing in this quarter broke the thesis, and the intraday recovery on August 11 says the market agrees. But the timeline extended, the near-term margin optics got worse, and at ~42x forward run-rate revenue there is no valuation cushion for a static-fire anomaly. This is a story where the next stock-moving event is an engineering test, not an earnings print.
Grade: B
This is not financial advice — for informational and educational purposes only.

CIFR: Institutions Bought 114 Million Shares in a Single Month While the Stock Fell in Half

Colby Hill — Wed, 19 Aug 2026 11:21:17 GMT

August 19, 2026 | Deep Dive by Colby

OPENING HOOK

Why this story matters right now

Here is a number worth sitting with. Between July and August of 2026, institutional holders of Cipher Digital went from owning 342.6 million shares to owning 456.5 million shares. That is 113.9 million shares added in thirty days — a 33% increase in institutional position size in a single reporting month, and the largest one-month accumulation in the company's five-year history as a public company.

The stock, over roughly that same stretch, went from $25.82 to $16.10. Down 38%. Down 47% from its June 22 high of $30.14.

So the smart money bought the largest block of shares it has ever bought into the worst drawdown the stock has had all year. On Tuesday, August 18, CIFR fell another 13% on 53.4 million shares — nearly a million and a half above its 30-day average — as rising Treasury yields punished every capital-intensive AI infrastructure name on the board. It closed at $16.095. It also became one of the five most-discussed tickers on X, with sentiment running 57% bullish even as the price collapsed.

Cipher Digital Inc. — formerly Cipher Mining, renamed in February 2026 — is a New York-based company with sixty-six employees that is in the middle of the most aggressive business-model transplant in the public markets. It is killing a Bitcoin mining operation and replacing it with contracted, investment-grade data center leases signed with Amazon Web Services and Google. The market is pricing the death of the old business. The filings say the new one already started paying rent.

INVESTMENT THESIS

The core reasons to own this stock

One — the revenue is already contracted, not projected. Cipher has signed roughly $11.4 billion in contracted revenue across its hyperscaler leases, including a 15-year, 300 MW agreement with AWS at the Black Pearl campus worth about $5.5 billion, and a 10-year, 300 MW agreement at Barber Lake backed by Google through Fluidstack, with an additional 39 MW layered on top. These are not letters of intent or framework agreements. They are executed leases with contractual completion dates, and management projects roughly $793 million in average annualized net operating income from October 2026 through September 2036. That is a decade of visible cash flow against a $6.7 billion market cap.

Two — the first rent check already cleared, early. In early August, Cipher delivered initial capacity at Black Pearl two months ahead of schedule at the tenant's request and began charging rent. This is the single most underrated fact in the story. Every bear thesis on the Bitcoin-to-AI conversion trade rests on the assumption that these companies cannot actually build and energize on time. Cipher just built and energized early, and got paid for it.

Three — the enterprise value is trading below the contract book. With roughly $4.25 billion in cash and restricted cash against $5.21 billion in principal long-term borrowings as of Q1, net debt runs near $1 billion, putting enterprise value around $7.6 billion against $793 million of forward NOI. That is under 10x EV/NOI for a portfolio of decade-long leases signed with two of the most creditworthy tenants on earth. Digital Realty and Equinix trade at multiples roughly double that. Compass Point has made exactly this argument publicly — that Cipher and its peers are trading below the value of contracts already signed.

Four — the pipeline is larger than the contracted book. Cipher carries a 5.3 gigawatt development portfolio, including the 1 GW Colchis site in West Texas and a newly optioned 900 MW site near San Antonio called Apollo spanning about 288 acres. Only 700 MW of that is currently under contract. Every incremental megawatt leased is upside that nobody is paying for at today's price.

BUSINESS OVERVIEW

What Cipher Digital actually does

Cipher Digital is, functionally, a power landlord. The company acquires land with grid interconnection rights in Texas and elsewhere, builds industrial-scale data center campuses on that land, and then leases the finished capacity to tenants under long-duration contracts. The historical business was Bitcoin mining — Cipher bought power cheaply and converted it into hashrate. The new business converts that same power into rented compute capacity for artificial intelligence workloads, which is a dramatically better trade because the counterparty is Amazon rather than a volatile commodity.

The mechanical difference matters enormously. Bitcoin mining revenue is a function of hashprice, network difficulty, and the price of an asset that can drop 40% in a quarter. Data center lease revenue is a function of a signed contract with a fixed rent schedule and an investment-grade tenant on the other side. Cipher earned $24.8 million in Q2 2026, down from $43.6 million a year earlier, precisely because the mining business is being deliberately wound down and the lease business had not yet started collecting. That gap — the trough between the old revenue and the new revenue — is what the market has spent the summer selling.

Management has stated it intends to fully exit Bitcoin by the end of 2026, redeploying the proceeds from its remaining holdings into infrastructure development. The company ended 2025 with roughly $754 million in liquidity, including $628 million in cash and $125 million in Bitcoin. The transition is not a pivot in the marketing sense. It is a full replacement of the income statement.

INDUSTRY OPPORTUNITY

Electricity, not capital, is now the binding constraint on AI

The Big Five hyperscalers — Amazon, Alphabet, Meta, Microsoft, and Oracle — are on pace to spend roughly $725 billion on AI infrastructure in 2026 alone, with some estimates running as high as $800 billion. That number is staggering, but it is not the interesting one. The interesting one is that they cannot spend it fast enough because there is nowhere to plug the machines in.

US data center power demand is projected to reach approximately 76 gigawatts in 2026, up from around 50 gigawatts in 2024, and forecasts put it near 176 gigawatts by 2035. Against that demand curve sits an estimated 9.3 gigawatt structural capacity shortfall in the US market right now. Microsoft has publicly described an $80 billion Azure backlog driven not by weak demand but by power availability. When the largest software company in the world tells you it has money it physically cannot deploy, the scarce asset is not capital. It is an energized megawatt.

This is why a Bitcoin miner with grid interconnection rights in West Texas became a strategically valuable company. Cipher spent five years assembling exactly the thing that is now in critically short supply — sited, permitted, powered industrial land — for reasons that had nothing to do with artificial intelligence. Bernstein initiated coverage on the group with the phrase "the power landlords of AI" and projected a ninefold increase in AI revenue by 2030. That framing is correct. These companies do not sell compute. They sell the right to exist near a substation.

COMPETITIVE ADVANTAGE

Why Cipher might win

The moat is tenant quality, and it is not close. Cipher's contracted book is anchored by Amazon Web Services and Google — two of the highest-credit counterparties available anywhere in the world. Compare that to TeraWulf, whose contracted revenue is roughly 91% concentrated in Fluidstack, a private neocloud whose credit ultimately depends on backstop arrangements rather than a balance sheet. Cipher's AWS exposure sits near 48% of contracted revenue, meaningful but materially more diversified, and the other anchor is Google-backed. When the AI capex cycle eventually cools — and it will — the question that decides who survives is whether your tenant can keep paying rent through a downturn. Amazon can.

The second advantage is demonstrated execution speed. Six publicly traded miners have announced HPC contracts: Core Scientific, Cipher, TeraWulf, Applied Digital, Galaxy Digital, and IREN. Announcing a contract is easy. Energizing capacity on a contractual deadline in the middle of a global shortage of transformers, switchgear, and skilled electricians is not. Cipher secured roughly 95 to 100% of its long-lead equipment and workforce requirements ahead of schedule and then delivered Black Pearl capacity two months early. That is the proof point that separates a real developer from a press release.

Where Cipher is genuinely behind is scale. IREN commands a roughly 5 gigawatt secured power portfolio spanning North America, Spain, and Australia. TeraWulf carries a 2.9 gigawatt platform with more than $13 billion in contracted revenue. Cipher's 5.3 GW development pipeline is competitive on paper, but only 700 MW of it is contracted today, and roughly 2 gigawatts of the pipeline still lacks grid approval. The company is not the biggest player. It is the one with the best tenants and, so far, the best on-time record.

FINANCIAL DEEP DIVE

The numbers: what they mean and what to watch

Metric                     Value
-------------------------  ---------------------------
Last Close (Aug 18, 2026)  $16.095 (-13.0%)
Day Range                  $15.77 - $17.90
52-Week Range              $5.32 - $30.14
Market Cap                 $6.68B
Shares Out / Float         415.0M / 340.2M
P/E Ratio                  -5.84 (unprofitable)
P/B Ratio                  11.86
Dividend Yield             None
Q2 2026 Revenue            $24.84M (est. $31.9M)
Q2 2026 EPS                -$0.65 (est. -$0.24)
Q2 Adj. EBITDA             -$29.99M
FY2026 Revenue Guide       ~$220.8M (cut from $232.3M)
Contracted Revenue         ~$11.4B
Avg. Annualized NOI        ~$793M (Oct 26 - Sep 36)
Avg. Volume (30d)          35.5M

The income statement is genuinely ugly, and pretending otherwise would be dishonest. Q2 revenue of $24.8 million missed consensus of $31.9 million badly and fell from $43.6 million a year prior. The loss of $0.65 per share was nearly triple the expected $0.24. GAAP net loss reached $267.5 million, though the overwhelming majority of that is non-cash — impairments and fair-value marks associated with winding down the mining fleet. Adjusted EBITDA of negative $30 million is the number that actually reflects operations, and it is negative because Cipher is currently running a construction company with almost no revenue attached to it. Analysts have responded by cutting the 2026 revenue forecast to $220.8 million and widening the expected loss to $1.34 per share.

The balance sheet is where the real story lives. As of Q1, cash and restricted cash totaled roughly $4.25 billion against principal long-term borrowings of $5.21 billion — 2030 and 2031 convertible notes plus senior secured notes tied specifically to the Barber Lake and Black Pearl facilities. That is roughly $1 billion of net debt, and critically, both flagship campuses are fully funded through completion via high-yield issuance already done. Cipher also completed a bond offering fully funding the Stingray development and reimbursed itself $56.7 million of prior project spending. The company does not need to come back to the equity market to finish what it has already sold.

What to watch is the October energization date. Both Black Pearl and Barber Lake target October 2026 lease commencement, and Black Pearl is already early. If both hit, Cipher goes from roughly $25 million of quarterly revenue to a run-rate that supports $793 million of annual NOI within four quarters. The negative P/E and the negative EBITDA both become historical artifacts. The next two earnings reports are the entire thesis.

MANAGEMENT TEAM

Who is running this?

Tyler Page has been CEO since August 2021 — five years, which in this sector counts as institutional memory. His background is finance rather than engineering: more than twenty years across institutional finance and fintech, including the management committee and Head of Client Strategies at New York Digital Investment Group, Head of Institutional Sales at Stone Ridge Asset Management, Global Head of Business Development for Fund Solutions at Guggenheim Partners in New York and London, and derivatives roles at Goldman Sachs and Lehman Brothers. He started his career as an attorney at Davis Polk & Wardwell, and ran business development for digital asset infrastructure at Bitfury before taking the Cipher job.

That resume is worth reading carefully, because it explains the company's actual competence. Cipher's edge has not been operational brilliance in mining — it has been structuring. Negotiating a 15-year investment-grade lease, financing two 300 MW campuses simultaneously through the high-yield market, and getting bondholders comfortable enough to fully fund construction before a single dollar of rent arrives is a capital-markets exercise, not an engineering one. Page is a derivatives-and-structuring guy running a company whose central problem is derivatives-and-structuring.

Alignment is real. Page holds roughly 8.26 million shares, worth about $133 million at Tuesday's close and considerably more before the drawdown. His compensation has shifted meaningfully toward performance: stock awards rose from $2.6 million in 2023 to $14.86 million in 2024 while cash bonuses fell from $2.49 million to $1.65 million, and 50% of annual long-term incentive is now delivered in PSUs tied to relative total shareholder return with an absolute TSR modifier. He is paid in the outcome, not the effort. Notably, Morgan Stanley disclosed a 6.3% stake in the company — a sell-side firm's asset management arm taking a position of that size in a $6.7 billion company is a genuine vote of confidence.

BULL CASE

5 reasons this stock could dramatically outperform

One — October energization converts the story into cash. Both flagship campuses target October 2026 commencement, and Black Pearl already started collecting rent in early August. If Barber Lake follows on schedule, Cipher exits 2026 with a revenue run-rate that makes the current $6.7 billion market cap look absurd against $793 million of annualized NOI. The re-rating would not require any new contracts. It would require only that the existing ones start.

Two — the pipeline is 4.6 gigawatts of unpriced optionality. Of a 5.3 GW development portfolio, only 700 MW is contracted. The Colchis site in West Texas carries 1 GW; the newly optioned Apollo site near San Antonio supports up to 900 MW. Every lease signed against that pipeline is incremental NOI against a fixed share count. Morgan Stanley's $76.50 bull case explicitly assumes further lease signings, additional Bitcoin-to-AI site conversions, and continued power pipeline growth.

Three — institutional accumulation is running at unprecedented scale. Institutions added 113.9 million shares in a single month and now hold 458.7 million shares across 766 filers, with BlackRock, Vanguard, UBS, D. E. Shaw, Jane Street, Wells Fargo, and Susquehanna all on the register alongside V3 Holding as the largest position. Institutional holdings are up roughly 64% year over year from 277.9 million shares. That is not index rebalancing. That is deliberate, size-driven accumulation into weakness.

Four — the valuation gap versus data center comps is structural, not cyclical. At roughly 10x EV/NOI, Cipher trades at half the multiple of established data center REITs with comparable tenant credit and comparable lease duration. As the mining business fully exits the financials by year-end 2026 and the reported numbers start looking like a REIT rather than a crypto company, the comparison set changes — and the multiple assigned to it should change with it.

Five — the power shortage is worsening, not easing. A 9.3 gigawatt structural US capacity shortfall against $725 billion of annual hyperscaler capex means rent for energized capacity is a seller's market. Management has publicly discussed pricing power in lease negotiations, and every quarter the shortage persists is a quarter in which the next contract gets signed at a better rate than the last one.

BEAR CASE

5 reasons this could disappoint

One — the October date is the whole company, and construction slips. Cipher currently generates negative $30 million of quarterly adjusted EBITDA against $5.21 billion of borrowings. The entire equity value assumes two 300 MW campuses energize roughly on schedule. A transformer delay, a grid interconnection dispute, or an electrical labor shortage that pushes commencement two quarters right does not reduce the value of the leases — but it does force the company to service $5.21 billion of debt for six more months with essentially no revenue. That is the scenario that actually breaks the stock.

Two — counterparty concentration is real. AWS represents roughly 48% of contracted revenue. Amazon is about as good a credit as exists, but concentration is concentration. A single tenant deciding to renegotiate, delay a phase, or slow its AI buildout would put nearly half the contract book in question. Hyperscaler capex plans are announced annually and revised quarterly.

Three — roughly 2 gigawatts of the pipeline lacks grid approval. The bull case leans heavily on converting the 4.6 GW of uncontracted pipeline into signed leases. But a meaningful chunk of that pipeline is not yet approved for interconnection, and ERCOT queue timelines have been extending, not compressing, as data center load requests flood the system. Pipeline megawatts without grid approval are land, not capacity.

Four — rate sensitivity is brutal and getting worse. Tuesday's 13% decline had nothing to do with Cipher's business. It was a Treasury yield move. A company financing gigawatt-scale construction with high-yield debt and valued on ten-year cash flows is a long-duration bond with a ticker. If yields keep grinding higher, the discount rate applied to $793 million of 2036 NOI rises mechanically, and the stock falls regardless of execution.

Five — the AI capex cycle is the only reason any of this works. Strip out the assumption of perpetually rising hyperscaler spending and Cipher is a company with 66 employees, no profits, $5.2 billion of debt, and a dying Bitcoin business. If the AI infrastructure buildout decelerates — through model efficiency gains, disappointing enterprise adoption, or simple capex fatigue after $725 billion — the 4.6 GW pipeline becomes stranded land and the contracted leases become the ceiling rather than the floor.

VALUATION

Building the model: Bear, Base, Bull

Scenario  Key Assumption                     Price Target  Probability
--------  ---------------------------------  ------------  -----------
Bear      Energization slips 2+ quarters     $9.00         25%
Base      Oct 2026 start on time; ~12x NOI   $32.00        50%
Bull      Colchis + Apollo leased; ~18x NOI  $62.00        25%

The probability-weighted expected value works out to $33.75 — roughly 110% above Tuesday's $16.095 close. That figure sits comfortably between the sell-side consensus 12-month target of $32.18 across 17 analysts and the three-firm average of $37.83 from Rosenblatt, BTIG, and Morgan Stanley, and well below Morgan Stanley's $76.50 bull case. The distribution is not symmetric, and that asymmetry is the point: the bear case is a timing problem that costs you 44%, while the base case alone doubles your money on nothing more than the leases starting when the contracts say they will. What makes this compelling is not the magnitude of the upside but the fact that the base case requires no new information — only the execution of documents already signed.

KEY METRICS TO WATCH

The five signals that tell you if the thesis is working

01 — Barber Lake energization date. Green flag: capacity delivered and rent commencing in October 2026 or earlier, matching the Black Pearl precedent. Yellow flag: any language in the Q3 update about "phased delivery" or "revised commencement" without a specific date attached.

02 — Quarterly HPC lease revenue. Green flag: Q3 2026 shows a clean, separately disclosed HPC revenue line growing sequentially toward the $198 million quarterly figure implied by $793 million of annualized NOI. Yellow flag: HPC revenue buried inside a blended total, or a Q4 figure materially below one-quarter of run-rate.

03 — New lease signings against the 4.6 GW uncontracted pipeline. Green flag: any executed agreement at Colchis or Apollo before mid-2027, particularly with a third distinct hyperscaler that reduces AWS concentration below 40%. Yellow flag: two consecutive quarters with no new contracted megawatts announced.

04 — Institutional share count in subsequent 13F cycles. Green flag: holdings sustaining above 450 million shares or continuing to build, confirming the August accumulation was accumulation rather than a one-time index or swap artifact. Yellow flag: a reversal back toward the 340 million share level, which would suggest the spike was mechanical.

05 — Net debt and any equity issuance. Green flag: net debt stable near $1 billion with no new share issuance, confirming both campuses genuinely are funded through completion. Yellow flag: any at-the-market equity program or convertible issuance before October, which would signal the construction budget is running over.

MY TAKE

The bottom line

The market is currently valuing Cipher Digital on its trailing income statement, and its trailing income statement is a picture of a company that no longer exists. Revenue fell in half because management deliberately killed the business that was generating it. The $267 million net loss is overwhelmingly non-cash write-downs on mining equipment being retired on purpose. Judging this company on Q2 2026 results is like valuing a restaurant during its renovation.

What actually matters is that on August 4, Cipher announced it had delivered capacity at Black Pearl two months early and started charging a hyperscaler rent. That is the first time in the company's history it has collected contracted HPC revenue from an investment-grade tenant, and it happened ahead of the contractual deadline in a construction environment where nothing happens ahead of deadline. The stock fell anyway, because the same press release contained an earnings miss, and because two weeks later Treasury yields moved and every long-duration infrastructure name got hit indiscriminately.

Meanwhile, institutions added 114 million shares. I do not treat 13F data as a signal to follow blindly — a lot of it is swaps, lending, and market-making inventory, and Jane Street and Susquehanna sitting on the top-ten list should temper anyone's enthusiasm about interpreting it as pure conviction. But a 33% jump in a single month, into a 47% drawdown, with BlackRock, Vanguard, D. E. Shaw, and a 6.3% Morgan Stanley position on the register, is not noise. Somebody with a research budget looked at the same October energization date I am looking at and decided the timing risk was worth taking.

The risk here is honest and I want to name it plainly: this is a leveraged construction bet with a hard deadline. If October slips meaningfully, $5.21 billion of debt gets serviced by a company with negative EBITDA, and the stock trades to single digits before the leases ever start. That is a real scenario and it deserves the 25% weight I gave it. But the payoff for being right is that a decade of Amazon- and Google-backed rent gets repriced from ten times NOI to something closer to what actual data center landlords fetch, and you do not need a single new contract for that to happen.

You are being asked to underwrite roughly six weeks of construction risk in exchange for ten years of investment-grade rent — and the market is currently pricing the six weeks as though the ten years might not exist.

ANALYST RATING

⚡ SPECULATIVE BUY

Speculative buy with a 12-to-24 month horizon, sized at 1-2% of portfolio and no more — the asymmetry is genuinely attractive but the leverage and single-date execution risk make this position-sizing discipline non-negotiable. Scale in rather than committing at once, and treat the Q3 update as the moment the thesis either confirms or breaks.

Sources & Further Reading

This is not financial advice. Always do your own research.

Z: Institutions Bought 38 Million Shares Into a 60% Crash — Six Days Before the Trial

Colby Hill — Tue, 18 Aug 2026 11:31:00 GMT

August 18, 2026 | Deep Dive by Colby

OPENING HOOK

Why this story matters right now

Here is a number worth sitting with. Between May and August of this year, institutional holders of Zillow Group added 38.2 million shares — a 21% increase in three months, taking total institutional ownership to 218.4 million shares, the highest level in the company's recorded filing history.

They did it while the stock was trading within four dollars of its 52-week low.

They did it while the company was reporting a GAAP net loss. They did it while four separate Wall Street desks were cutting price targets in the same week. And they did it knowing that on August 24 — six days from now — a federal judge in the Western District of Washington begins a two-week bench trial that will decide whether Zillow's dominant position in online rental listings was built on an illegal agreement.

That is the setup. A company growing revenue 18% year over year, guiding to 25% EBITDA margins, trading at roughly ten times forward EBITDA — and carrying a binary legal event on the calendar. Zillow was the single most-discussed ticker on FinTwit yesterday, with mentions up 150% in twenty-four hours. Retail is arguing about the lawsuit. Institutions appear to have already voted.

INVESTMENT THESIS

The core reasons to own this stock

1. The business is compounding while the stock is collapsing. Q2 2026 revenue hit $772 million, up 18% year over year, against a stock that is down 60.09% over the same twelve months. Zillow closed at $33.72 on August 17; a year ago it was $84.50. Revenue went one direction and the multiple went the other. That gap is the entire opportunity, and it exists because of legal overhang, not operational failure.

2. The preferred agent transition is a genuine margin machine, and it is working. Zillow now routes 61% of connections through its integrated preferred-agent experience, up from 44% at the end of 2025, with management accelerating the target past 75% by year-end. Preferred connections generated 23% more revenue per connection in 2025, and management expects that to reach 35% by the end of 2026. This is the same customer, monetized harder, at almost no incremental acquisition cost.

3. Mortgage and rentals are outgrowing the core by a mile. Rentals revenue grew 31% to $209 million and mortgage revenue grew 75% to $84 million in Q2, against for-sale revenue growth of 14%. Management guided rentals to high-20% growth in Q3 and mortgage to over 50% growth for the full year. The parts of Zillow that are least dependent on housing transaction volume are the parts growing fastest.

4. The valuation already prices in a loss. At a $7.6 billion market capitalization against full-year 2026 EBITDA guidance of $730–760 million, Zillow trades near 10 times forward EBITDA for a business with 72.8% gross margins growing mid-to-high teens. Comparable internet marketplaces with those characteristics do not trade at ten times. The market is applying a litigation discount, and litigation discounts unwind.

BUSINESS OVERVIEW

What Zillow actually does

Zillow Group operates the largest real estate marketplace in the United States. Founded in 2004 by Rich Barton and Lloyd Frink — Barton having previously built Expedia and Glassdoor — the Seattle company runs Zillow, Trulia, StreetEasy, HotPads, Zillow Rentals, Follow Up Boss and Zillow Home Loans, employing roughly 7,068 people. Zillow, Trulia and StreetEasy together account for something on the order of half of all U.S. online real estate traffic.

The revenue model is not the listings. It is the leads. Zillow does not charge consumers to browse homes; it charges real estate agents for the introduction to a consumer who is ready to transact, and it charges property managers for rental listing placement. In the newer preferred-agent construction, Zillow does more than hand off a lead — it wraps the connection in its own software stack, its own tour scheduling, its own follow-up CRM, and increasingly its own mortgage origination through Zillow Home Loans. Each additional layer captures more of the transaction economics without requiring another dollar of traffic acquisition.

That layering is the entire 2026 story. Management has been explicit that the preferred model shifts roughly 700 to 800 basis points of revenue out of the residential category and into mortgages by the fourth quarter. What looks like a slowdown in one line item is a deliberate migration into a higher-value one, and reading the segment table without that context produces exactly the wrong conclusion.

INDUSTRY OPPORTUNITY

The housing market froze, and Zillow got paid anyway

The U.S. housing transaction market has been miserable for three years. Mortgage rates have climbed again since their early-March lows, affordability has deteriorated, and Zillow's own leading indicator — newly pending listings — rose just 0.3% year over year in the most recent read and fell 7.7% from June. Zillow grew revenue 18% into that.

The reason is that Zillow's growth is decoupling from transaction volume. Rentals is a structurally different market: renters churn annually regardless of mortgage rates, and multifamily operators advertise through cycles because vacancy costs them daily. That business grew 31% last quarter. Mortgage is a share-gain story rather than a market-growth story — Zillow is originating loans for buyers it already owns the relationship with, which is why the segment can grow over 50% in a year when purchase originations are guided to decline.

The broader online real estate advertising market remains one of the largest under-digitized commission pools in the country. Roughly $100 billion flows annually to U.S. real estate agents in commissions, and the portals capture only a low single-digit percentage of it. Every basis point Zillow moves toward capturing more of that transaction — through preferred agents, through Showcase listings, through Follow Up Boss, through its own mortgage desk — is worth more than any incremental traffic gain. The market opportunity is not more eyeballs. Zillow already has the eyeballs. It is more dollars per eyeball.

COMPETITIVE ADVANTAGE

Why Zillow might win

The moat is the verb. Americans do not say they are going to search for a home online — they say they are going to Zillow something, and that single linguistic fact is worth more than any feature set. Zillow's traffic arrives direct and unpaid, which means its customer acquisition cost for the consumer side of the marketplace is structurally near zero while every competitor pays for reach.

That advantage is under the most serious assault of its existence. CoStar's Homes.com has spent aggressively to climb from near-invisibility to over 100 million monthly visitors in about two years, growing traffic roughly 200% year over year — a genuinely impressive land grab funded by CoStar's commercial data monopoly. Compass, which has been openly hostile to Zillow's listing standards, announced a major partnership with Redfin and Rocket Companies in February, creating a rival pre-market listing pipeline. Zillow's response was a counter-pact with Realtor.com on pre-market listings, which tells you the company takes the threat seriously. Meanwhile Google's AI-driven search changes hang over every content-dependent marketplace, though management stated on the Q2 call that it has seen no measurable impact.

Here is what separates Zillow anyway: none of those competitors owns the full stack. Homes.com has traffic but no mortgage arm, no agent CRM, and no rentals monetization at Zillow's scale. Compass is a brokerage picking a fight with the distribution channel its own agents depend on. Redfin has been absorbed into Rocket and is now optimizing for loan volume rather than portal dominance. Zillow is the only player that can take a consumer from a Tuesday-night browse through a 3D tour — now on 30% of for-sale listings in top-ten markets — through an agent introduction, through a CRM-managed follow-up, and into a mortgage it underwrites itself. That vertical integration is what the preferred model monetizes, and it cannot be replicated by buying traffic.

FINANCIAL DEEP DIVE

The numbers: what they mean and what to watch

Metric                     Value
-------------------------  -------------------
Last Close (Aug 17, 2026)  $33.72
52-Week Range              $29.23 - $93.88
One-Year Return            -60.09%
Market Cap                 $7.60B
P/E (Trailing)             154.4x
Price/Book                 1.82x
Dividend Yield             None
Q2'26 Revenue              $772M (+18% YoY)
Q2'26 Adj. EPS             $0.52 vs $0.38 est.
FY26 Revenue Guide         $2.92B - $2.96B
FY26 EBITDA Guide          $730M - $760M
Avg. Volume (30-day)       4.00M shares
Institutional Holders      755 (218.4M shares)

The tension in this table is the P/E line, and it deserves an honest explanation rather than a dodge. Zillow's trailing GAAP P/E of 154x looks absurd because GAAP net income is nearly zero — the company earned $55 million on a trailing-twelve-month basis on $2.81 billion of revenue, and posted a $4 million net loss in Q2 itself. That gap between a $0.52 adjusted EPS and a GAAP loss is stock-based compensation, and it is real dilution that shareholders bear. But it is also why the P/E is a useless valuation tool here. The number that matters is EBITDA: $745 million at the midpoint of guidance against a $7.6 billion market cap is roughly 10x, and that is the multiple the institutions buying in June, July and August were almost certainly underwriting.

The earnings quality itself has been excellent and consistently underappreciated. Zillow has beaten consensus EPS in six consecutive quarters, and the last four beats were by 43%, 57%, 63% and 37% respectively. Gross margin sits at 72.8%. Revenue has grown from $554 million in Q4 2024 to $772 million in Q2 2026 — a 39% increase in six quarters, through the worst housing transaction environment in a generation. This is not a broken business.

What to watch next is the Q3 print on October 29. Guidance calls for $745–760 million of revenue, implying roughly 11% growth — a deceleration from 18%, and the single most important number in the report. Management attributed the slowdown to a revised expectation that purchase mortgage originations will decline rather than stay flat, which is a macro call rather than a company-specific stumble. If Zillow prints at the high end and holds its 25% EBITDA margin target, the deceleration narrative dies. If it guides Q4 below, the bears get their proof that this is a cyclical business wearing a growth-stock costume.

MANAGEMENT TEAM

Who is running this?

Jeremy Wacksman, 48, became CEO on August 7, 2024 — two years and eleven days ago. He is not an outside fixer; he joined Zillow in 2009 and worked through CMO, President and COO before taking the top job, with a prior background running Xbox and mobile product at Microsoft. He holds an engineering degree from Purdue and an MBA from Kellogg. His total compensation as CEO runs approximately $7.1 million, weighted heavily toward equity — which, given the stock is down 60% in a year, means his own paper has been shredded alongside shareholders'.

The more important governance fact is that Rich Barton and Lloyd Frink remain co-executive chairmen. Barton founded Expedia inside Microsoft, spun it out, founded Glassdoor, and then founded Zillow — a track record of building category-defining internet marketplaces that very few operators can match. He also presided over the Zillow Offers disaster in 2021, which cost the company hundreds of millions and remains the clearest evidence that this board will take large, expensive swings. Investors should read the founder presence as both an asset and a risk.

One data point worth noting honestly: CTO David Beitel sold 1,798 shares on August 13. The filing indicates it was to cover tax withholding on vesting equity, which is administrative rather than a sentiment signal, and it would be dishonest to characterize it as anything more.

BULL CASE

5 reasons this stock could dramatically outperform

1. Zillow wins the FTC trial and the discount evaporates overnight. The bench trial starting August 24 runs roughly two weeks. Zillow and Redfin have already beaten the FTC's attempt to narrow the trial and defeated the request to presume the deal unlawful outright — the judge ruled key facts remained genuinely disputed. If the court rules for Zillow, a stock trading at 10x EBITDA on litigation fear re-rates toward the peer group with no operational change required. Analyst consensus sits at $55.83, roughly 66% above the current price.

2. The preferred agent flywheel hits its stride in Q4. Management is targeting 75%+ of connections in the preferred model by December, with 35% more revenue per connection. Do that arithmetic against a roughly flat connection count and it produces meaningful revenue growth from monetization alone, independent of housing volume. This is the highest-confidence catalyst on the calendar because it is entirely within the company's control.

3. Rate cuts unlock the pent-up transaction wall. Zillow's guidance now assumes purchase originations decline. That is a low bar. Millions of American households are sitting on sub-4% mortgages they will not surrender, and the moment the spread narrows enough to break the lock-in, transaction volume snaps back. Zillow is the maximum-torque way to own that recovery, and the Q3 guide has already de-risked expectations.

4. Mortgage becomes a real profit center rather than an accessory. Growing over 50% in a declining origination market means Zillow Home Loans is taking share from incumbents by owning the relationship earlier in the funnel. Every basis point of attach rate on Zillow's own traffic is nearly pure incremental margin. This segment has been a rounding error for years; at $84 million a quarter and compounding, it stops being one.

5. The stock is a coiled setup at the low end of its range. Zillow trades at $33.72 against a 52-week low of $29.23 and a high of $93.88. Institutions added 38 million shares into that trough. Sentiment on FinTwit is bullish and mention volume is up 150%. A favorable trial verdict landing into that positioning — with the buy-side already long and the sell-side price targets sitting 40–125% above spot — is the kind of asymmetry that produces violent upside.

BEAR CASE

5 reasons this could disappoint

1. The FTC wins and the rentals business gets restructured. The government alleges Zillow paid Redfin $100 million to exit the internet listing services market, calling it an unlawful agreement to eliminate competition. If the court agrees, remedies could include unwinding the exclusive arrangement, injunctive relief on future deals, and a template for follow-on private antitrust suits. Rentals is the fastest-growing segment at 31% growth; an adverse ruling attacks the exact engine the bull case depends on. This is a genuine binary and it resolves within weeks.

2. The securities class action compounds the legal exposure. A separate Sections 10(b) and 20(a) fraud action is pending in the Western District of Washington, alleging Zillow misled investors by characterizing the Redfin transaction as a "partnership" when it functioned as an acquisition, and by understating the antitrust risk. An FTC loss would hand the securities plaintiffs a near-complete evidentiary record. The financial exposure is unquantifiable today, and unquantifiable is worse than large.

3. Growth is decelerating in plain sight. Revenue grew 18% in Q2. Q3 guidance implies 11%. That is a seven-point deceleration in a single quarter, and management's explanation — that purchase originations will decline rather than stay flat — is an admission that the macro assumption underpinning the full-year plan has already been cut once. If rates rise further, it gets cut again.

4. GAAP profitability keeps not arriving. Zillow posted a $4 million net loss in Q2 on $772 million of revenue, and trailing-twelve-month net income of $55 million on $2.81 billion. Twenty-two years after founding, on 72.8% gross margins, the company still cannot reliably convert to GAAP earnings because stock-based compensation absorbs the operating profit. Shareholders are funding employee wealth creation, and the second round of layoffs in 2026 suggests management knows the cost base is the problem.

5. Homes.com and the Compass-Redfin-Rocket axis are real, funded, and hostile. CoStar has the balance sheet to spend for years and has already taken Homes.com past 100 million monthly visitors. Compass is trying to route listings around Zillow entirely. Zillow's counter-alliance with Realtor.com is a defensive move, not an offensive one. If the industry successfully fragments listing distribution, Zillow's traffic monopoly — the entire basis of its pricing power with agents — erodes, and no amount of preferred-agent monetization saves a marketplace that has lost its supply.

VALUATION

Building the model: Bear, Base, Bull

Scenario  Key Assumption                             Price Target  Probability
--------  -----------------------------------------  ------------  -----------
Bear      FTC wins; rentals restructured; 8x EBITDA  $22           30%
Base      Narrow ruling; $745M EBITDA at 13x         $43           45%
Bull      Clean win; $950M FY27 EBITDA at 16x        $67           25%

Run the probability-weighted math and you get $42.70 — roughly 27% above the $33.72 close. That is a respectable expected value, but the distribution matters more than the mean. The bear case costs you 35% and the bull case pays you 99%, which is a genuinely favorable asymmetry, and it is being offered precisely because the resolution date is on the calendar and most investors do not want to hold a position through a verdict they cannot handicap. Note also that the base case at $43 sits well below the $55.83 sell-side consensus — I am being deliberately more conservative than Wall Street here, because the analyst targets appear to embed a clean legal outcome that is far from assured.

KEY METRICS TO WATCH

The five signals that tell you if the thesis is working

01 — The FTC verdict. Trial begins August 24 and runs approximately two weeks in the Western District of Washington. A ruling for Zillow or a narrow remedy is the green flag and the single largest value unlock available. A broad injunction against the Redfin arrangement plus divestiture-style remedies is the yellow flag, and it invalidates the rentals growth assumption in the base case.

02 — Preferred connection penetration. Zillow reported 61% in Q2, up from 44% at the end of 2025, targeting 75%+ by year-end. Anything at or above 70% in the Q3 report is the green flag — it confirms the transition is ahead of schedule. Below 65% signals agent resistance to the new model, and agent resistance is the quietest way this thesis fails.

03 — Q3 revenue against the $745–760 million guide. Printing at or above the high end with reaffirmed full-year revenue of $2.92–2.96 billion is the green flag. Coming in at or below the low end, or trimming the full-year range, confirms the deceleration is structural rather than a conservative guide, and the 10x EBITDA multiple stops looking cheap.

04 — Rentals revenue growth rate. Q2 delivered 31%; Q3 guidance calls for high-20% growth. Holding above 25% is the green flag and demonstrates the segment compounds independently of the legal outcome. A drop below 20% — particularly if it coincides with an adverse ruling — is the yellow flag that the rental moat was more legal than structural.

05 — Institutional share count in the next Fintel update. Institutions hold 218.4 million shares across 755 filers as of the August data, up 38.2 million from May. A further increase through the trial is the green flag: it means the buy-side is underwriting a favorable outcome with fresh capital. A sharp reversal back toward the 180 million level would be the clearest possible signal that professional money has re-handicapped the verdict, and retail should not be the last to know.

MY TAKE

The bottom line

I keep coming back to the ownership chart. From May to August, institutions took their position from 180 million shares to 218 million — the largest three-month accumulation in the recorded series, executed with the stock pinned near a 52-week low and a federal antitrust trial six days out. Professional money does not accidentally add 38 million shares of a litigation-clouded name. Somebody with better lawyers than mine read the docket, watched the FTC lose its motion to presume the deal unlawful, watched the judge decline to narrow the trial, and concluded the government's case is weaker than the 60% drawdown implies.

That does not make them right. It makes the trade legible. What you are buying at $33.72 is a marketplace with 72.8% gross margins growing revenue 18%, guiding to 25% EBITDA margins, trading at roughly ten times forward EBITDA — priced as though it is going to lose. If the market's implied probability of an adverse ruling is materially higher than the actual probability, this is cheap. If it is not, this is a value trap with a court date.

The honest counterweight is that Zillow has been unable to produce meaningful GAAP earnings for two decades on some of the best gross margins in software, and the competitive picture is deteriorating on two fronts at once. Homes.com has real traffic now. Compass and Rocket are actively building around Zillow. Neither of those problems is solved by a favorable verdict — a win just removes the thing that made you look at this at all.

So I size this like what it is: a legally binary position on an operationally sound business at a genuinely low multiple. Not a core holding. Not something to average down into with conviction before August 24. A defined-risk allocation on an asymmetry that the calendar created and that will not exist in a month.

You are not being paid to predict Zillow's next quarter — you are being paid to take the other side of a courtroom outcome that institutions just bought 38 million shares of confidence in, at ten times EBITDA, six days before the gavel.

ANALYST RATING

⚡ SPECULATIVE BUY

A 2–3% starter position sized to survive a total adverse ruling, with the discipline to add only after the August 24 verdict is known — 12 to 18 month horizon, and no averaging down into the trial.

Sources & Further Reading

Zillow Group (ZG) Q2 2026 Earnings Call Transcript — The Motley Fool

Zillow beats Q2 forecasts with 18% spike in revenue — Real Estate News

Zillow, Redfin must face FTC's antitrust lawsuit — Multifamily Dive

FTC's challenge to Zillow-Redfin partnership headed to trial — The Real Deal

Zillow Group (Z) Institutional Ownership — Fintel.io

Zillow Group (Z) Stock Forecast & Analyst Price Targets — StockAnalysis

This is not financial advice. Always do your own research.

DOW: Institutions Just Bought Back Five Years of Selling in Ninety Days

Colby Hill — Mon, 17 Aug 2026 11:30:07 GMT

August 17, 2026 | Deep Dive by Colby

OPENING HOOK

Why this story matters right now

Here is a number worth sitting with. In May 2026, institutions held 493 million shares of Dow Inc. — the lowest level in more than five years, the end of a slow, grinding, quarter-after-quarter exit that had run since late 2021. Three months later, they hold 706 million shares.

That is 212 million shares added in a single quarter. It is not a rounding error, it is not index rebalancing noise, and it did not happen while the stock was cheap and forgotten. It happened while Dow was climbing off a $20.65 low and while every strategist on the Street was still telling you the global petrochemical downcycle would not bottom until 2028.

1,785 institutions now hold 712 million of Dow’s 722 million shares outstanding. Round it however you like — the float is essentially fully owned by professionals. Retail is not driving this stock. Somebody with a spreadsheet is.

And on Friday, August 14, Dow became the single most-discussed ticker on X, with mentions surging 96% in 24 hours — an unusual thing for a 129-year-old Michigan chemical company that pays a dividend and makes polyethylene pellets. Dow Inc. closed at $31.07, up 2.3% on the day, up roughly 69% year-to-date, and still 27% below its March high. This is what a cyclical looks like when the market cannot decide whether it is early or late.

INVESTMENT THESIS

The core reasons to own this stock

1. The self-help is real and it is compounding. Dow finished its $1 billion 2025 cost program and immediately raised the bar. Full-year 2026 self-help is now guided above $1.3 billion, and the Transform to Outperform program was lifted from $500 million to $700 million of 2026 benefit on the way to a $2 billion EBITDA opportunity by the end of 2028. Roughly 55% of planned role reductions are already implemented, worth more than $200 million of EBITDA in the second half alone. None of this requires polyethylene prices to cooperate.

2. Capacity is finally leaving the industry, and Dow is helping it leave. Dow shut its high-cost siloxanes unit in Barry, U.K. — about 25% of European siloxane industry capacity — for roughly $60 million of second-half EBITDA. Seven European crackers have been closed or are slated to close by 2027 across ExxonMobil, Shell, SABIC, TotalEnergies, Versalis and Dow. Over 8 million metric tons of global capacity has been pulled. Supply discipline is the only thing that ever ends a chemical downcycle, and it has started.

3. The balance sheet was defused before the market noticed. The Macquarie partnership around Diamond Infrastructure Solutions delivered roughly $3 billion in total proceeds, including a $540 million second tranche when Macquarie lifted its stake from 40% to 49%. The $10 billion Path2Zero project in Alberta was pushed to a late-2029 Phase One commissioning. Dow chose liquidity over ambition at exactly the right moment.

4. You are paid to wait. At $0.35 quarterly, Dow yields 4.51% — the 460th consecutive dividend paid by Dow or its predecessors since 1912. It was halved in July 2025, which is precisely why it is now credible. The cut already happened. What is left is a payout the company can actually cover through a trough.

BUSINESS OVERVIEW

What Dow actually does

Dow is a materials science company built on three segments. Packaging & Specialty Plastics is the engine — hydrocarbons, energy, and the polyethylene franchise that turns ethane and naphtha into the resin inside food packaging, films, pipe and wire insulation. This is the most commoditized part of the business and the most levered to the cycle. Industrial Intermediates & Infrastructure covers industrial solutions, polyurethanes and construction chemicals, the products that sit inside mattresses, insulation, coatings and adhesives. Performance Materials & Coatings is the specialty end — silicones, performance monomers, and consumer solutions, where Dow sells into personal care, electronics and architectural coatings at margins that behave less like a commodity.

The economics are simple and brutal. Dow buys hydrocarbon feedstock, cracks it, and sells derivatives. When the spread between feedstock cost and resin price is wide, Dow prints cash. When new Chinese capacity floods the market, that spread compresses toward zero and Dow’s fixed-cost base — 34,600 employees, dozens of plants, and one of the largest integrated footprints on the U.S. Gulf Coast — becomes the problem instead of the moat.

The company was founded in 1897 in Midland, Michigan, and separated from DowDuPont in 2019 as a standalone materials science pure play. It has now spent the majority of its short independent life inside the worst petrochemical downcycle in a generation.

INDUSTRY OPPORTUNITY

The oversupply that built the bottom

Between 2020 and 2025, global ethylene capacity expanded by more than 40 million tons. Demand grew by roughly 27 million tons. Around 70% of that new capacity was built in China. That single mismatch is the entire story of the last four years — global plant utilization fell to roughly 80%, and integrated polyethylene cash margins have been negative since mid-2022. Not thin. Negative.

The correction is finally arriving from the supply side. More than 8 million metric tons of capacity closures have been announced, concentrated in Northeast Asia and Europe where energy costs made those assets structurally uncompetitive. Wood Mackenzie and ICIS both frame the same picture: the capacity build-up slows meaningfully after 2026, at which point the industry pivots to a demand-led recovery. ICIS puts the true cycle bottom in 2028 or 2029, and warns that a return to “satisfactory” industry profitability may not arrive until the early 2030s.

That is the honest macro. It is also why this setup is interesting rather than obvious. The stock does not need the cycle to be fixed. It needs the market to stop pricing the cycle as if it will never be fixed — and every closed cracker in Europe moves that needle. New U.S. and Qatari ethylene capacity starting in 2026 is the counterweight, and it is a real one.

COMPETITIVE ADVANTAGE

Why Dow might win

The single most important moat is feedstock position. Dow’s U.S. Gulf Coast complex runs on cheap North American ethane, and ethane-based cracking sits at the bottom of the global cost curve. When European naphtha crackers are losing money at the same resin price where Dow’s Gulf Coast assets are profitable, market share transfers by physics, not by strategy. That is why the crackers being permanently closed are in Europe and Northeast Asia and not in Texas and Louisiana.

The second advantage is integration breadth. LyondellBasell is the closest structural comparison and trades around 7.2x LTM EV/EBITDA versus Dow near 8.5x — a premium Dow has to justify, and the justification is the silicones, polyurethanes and coatings portfolio that LYB does not have. Those non-polyethylene segments have underperformed peers for years, which is exactly Evercore’s argument for owning it: they are the coiled spring if industrial and housing demand stabilizes. Westlake Chemical is the more levered PVC and building-products play and gives you sharper torque to a construction recovery with less diversification. BASF carries the European cost problem Dow is actively shedding.

Where Dow is not advantaged is scale of ambition. Path2Zero was supposed to be the differentiator — a net-zero cracker in Alberta that would have separated Dow on carbon intensity and cost simultaneously. It is delayed to 2029 at the earliest. The moat today is operational, not visionary.

FINANCIAL DEEP DIVE

The numbers: what they mean and what to watch

Metric                     Value
-------------------------  ---------------------
Last Close (Aug 14, 2026)  $31.07 (+2.3%)
52-Week Range              $20.65 - $42.74
Market Cap                 $22.45B
P/E (TTM)                  -17.05 (negative EPS)
Price / Book               1.42x
Dividend / Yield           $0.35 qtr / 4.51%
Ex-Dividend Date           Aug 31, 2026
Q2'26 Net Sales            $12.1B (+20% YoY)
Q2'26 Operating EBITDA     $2.3B
Q2'26 Adjusted EPS         $1.44 (beat 12.6%)
Q3'26 EBITDA Guide         ~$1.7B
2026 Self-Help Target      >$1.3B
Avg Volume (30-day)        9.60M
Shares Outstanding         722.3M

Start with the contradiction. Dow just posted $12.1 billion in quarterly revenue, up 20% year over year, with $2.3 billion of operating EBITDA and adjusted EPS of $1.44 that beat consensus by nearly 13% — and the trailing P/E is still negative 17. That gap between the reported quarter and the trailing twelve months is the entire investment debate in one line. The last four quarters contain the trough. The next four may not.

The P/B of 1.42x is the number that actually matters for a cyclical. You are paying a modest premium to replacement-adjusted book for a business whose assets are irreplaceable at anything close to carrying value in the current cost environment. Meanwhile the market cap is $22.4 billion against a company generating $2.3 billion of EBITDA in a single quarter of a bad cycle.

Now the cold water. Management guided Q3 operating EBITDA to roughly $1.7 billion, a $600 million sequential decline. That is not a stumble — it is the June polyethylene price decline flowing through the averaging effect, heavier planned maintenance, and the absence of a one-time land sale gain that flattered Q2. Self-help adds about $130 million sequentially and does not come close to covering it. What to watch is whether Q4 stabilizes near $1.7B or slips further; the first is a plateau, the second is a new leg down.

MANAGEMENT TEAM

Who is running this?

Karen S. Carter became Chief Executive Officer on July 1, 2026, the first woman to lead Dow in the company’s 126-year history. She is six weeks into the job as of this writing. She is also not an outsider — Carter spent more than three decades at Dow, most recently as Chief Operating Officer, and before that ran human resources and the packaging business. The board framed the handoff as the culmination of a multi-year succession process, which is corporate language for “we are not changing strategy.”

Jim Fitterling, CEO since 2018 and the architect of both the DowDuPont separation and the current cost program, moved to Executive Chair on the same date. That structure cuts both ways. It preserves continuity through a cycle trough, which is genuinely valuable when your main job is executing $2 billion of cost takeout. It also means the person who cut the dividend and delayed Path2Zero is still in the room, and a new CEO inheriting an executive chair rarely gets to make a clean break.

The alignment question is straightforward: Carter is an operator being handed an operating problem. Transform to Outperform is her scoreboard. If the $2 billion by 2028 lands, she will have earned the seat. If it slips, the market will conclude that Dow’s problem was never execution.

BULL CASE

5 reasons this stock could dramatically outperform

1. The institutional bid is already in the tape. Institutions went from 493 million shares in May to 706 million in August — a 43% increase in three months that took ownership back to 2021 levels. That is not a quiet accumulation, it is a repositioning. Money that size does not chase a 4.5% yield; it buys a cycle turn. If they are early, you are early with them.

2. Self-help alone can carry earnings without any price recovery. More than $1.3 billion of 2026 benefit, $700 million from Transform to Outperform, and $200 million-plus of role-reduction EBITDA landing in the second half. Against a $22 billion market cap, executing the full $2 billion by 2028 target is a structural repricing of the earnings base independent of polyethylene.

3. European capacity closures are permanent, not cyclical. Seven crackers gone by 2027. The Barry siloxanes shutdown removed a quarter of European industry capacity in that product. When demand does return, it returns to a smaller supply base — and Dow’s Gulf Coast ethane position captures a disproportionate share of that recovery.

4. Analyst targets imply the Street is still behind the story. Eighteen analysts average a Buy with a $35.69 twelve-month target, about 21% above the current price. Evercore’s Outperform call carries a $56 target — 80% upside — on the thesis that silicones, polyurethanes and coatings mean-revert. Even UBS, at Neutral, raised its target to $32 from $29. The distribution is skewed upward.

5. The dividend cut is behind you and the balance sheet is funded. Roughly $3 billion from the Macquarie infrastructure partnership, Path2Zero pushed to 2029, and a payout already halved to a level the company can defend through a trough. The forced-selling scenario that haunts leveraged cyclicals at the bottom has been taken off the table.

BEAR CASE

5 reasons this could disappoint

1. Q3 guidance is a 26% sequential EBITDA decline and management told you so. From $2.3 billion to roughly $1.7 billion. June polyethylene prices fell, maintenance is heavier, and the Q2 land-sale gain does not repeat. If you buy here you are buying into a quarter the company has already warned will be materially worse.

2. The cycle bottom may be two to three years away, not two quarters. ICIS puts the ethylene and polyethylene trough in 2028 or 2029 and warns satisfactory industry profitability may not arrive until the early 2030s. Integrated PE cash margins have been negative since mid-2022. Buying a cyclical three years before the trough is how investors lose money slowly and then all at once.

3. New capacity is still coming online in 2026. Fresh ethylene and polyethylene plants are starting in the United States and Qatar this year, in low-cost feedstock regions. Every European closure is partially offset by a new Gulf Coast or Middle East line. Net supply discipline is far less clean than the closure headlines suggest.

4. Trailing earnings are still negative and the valuation premium is real. A -17 P/E means there is no earnings floor under this stock, only asset value and hope. Dow trades near 8.5x LTM EV/EBITDA against LyondellBasell at 7.2x. You are paying a premium multiple on depressed EBITDA in a commodity business — the most dangerous combination in cyclical investing.

5. The stock has already run 69% year-to-date and sits 27% below its own March high. That is not a forgotten value stock; it is a volatile one. From $20.65 in October to $42.74 in March to $31.07 today, Dow has round-tripped a fortune in ten months. Anyone buying now is buying in the middle of that range with a negative earnings print and a guided-down quarter directly ahead.

VALUATION

Building the model: Bear, Base, Bull

Scenario  Key Assumption                               Target  Prob.
--------  -------------------------------------------  ------  -----
Bear      Trough slips to 2028-29; Q4 below $1.7B      $22     30%
Base      EBITDA plateaus; $1.3B self-help lands       $36     50%
Bull      PE margins inflect 2027; full $2B Transform  $52     20%

The probability-weighted expected value works out to roughly $35 per share — about 13% above Friday’s close of $31.07, and closer to 17% once you add the 4.5% dividend. That is a respectable but not extraordinary risk-adjusted return, and the shape of it matters more than the number. The bear case costs you 29%. The bull case pays you 67%. You are being offered slightly better than two-to-one on a business where the downside is a further leg of a downcycle you can measure, and the upside is a mean reversion in a commodity spread that has been negative for four years and cannot stay negative forever.

Note that the base case of $36 sits almost exactly on the eighteen-analyst consensus of $35.69, and the bull case sits well below Evercore’s $56. This is not an aggressive model. It is what the numbers give you.

KEY METRICS TO WATCH

The five signals that tell you if the thesis is working

01 — Quarterly operating EBITDA versus the $1.7 billion guide. Green flag: Q3 comes in at or above $1.7B and Q4 holds flat or better, confirming a plateau. Yellow flag: Q3 misses or Q4 guides below $1.6B, which means the trough is still deepening and the self-help is being consumed by price.

02 — Transform to Outperform run-rate disclosure. Green flag: management reiterates or raises the $700 million 2026 benefit and stays on track for $2 billion by 2028. Yellow flag: any language about “timing” or “phasing” of savings — that is how cost programs die quietly.

03 — U.S. integrated polyethylene cash margins. Green flag: margins turn positive on a sustained basis for two consecutive quarters, the first time since mid-2022. Yellow flag: margins stay negative through 2027 while new U.S. and Qatari capacity ramps.

04 — Institutional share count in subsequent 13F cycles. Green flag: the 706 million share position holds or grows into the November filings, confirming the June-to-August buying was conviction rather than a one-quarter trade. Yellow flag: a retrace back toward 550-600 million, which would mark the recent surge as noise.

05 — The dividend and free cash flow coverage. Green flag: operating cash flow comfortably covers the roughly $1.0 billion annual payout with room for capex. Yellow flag: any quarter where the dividend is funded from asset sales or the balance sheet — the 460-consecutive-dividend streak is a source of pride, and pride is expensive when cash is tight.

MY TAKE

The bottom line

I keep coming back to the ownership chart. From September 2021 through May 2026, institutions sold Dow relentlessly — 703 million shares down to 493 million, a slow-motion exit that tracked the death of the petrochemical cycle almost perfectly. Then in three months they bought back every share of that decline and more. You do not get a cleaner behavioral signal than that in a large-cap industrial.

But I am not going to pretend the fundamentals have caught up. Trailing earnings are negative. Management has explicitly guided the current quarter down 26% sequentially. The most credible industry research says the cycle does not bottom until 2028. Everything the bears say about this stock is true, verifiable, and comes from Dow’s own investor deck. This is not a case where one side is misreading the data.

What tips it for me is the asymmetry of what has already been done versus what still needs to happen. The dividend cut — done. The Path2Zero deferral — done. The $3 billion infrastructure monetization — done. The European high-cost shutdowns — done. More than half the role reductions — done. Everything painful is in the rearview mirror, priced in, and behind a stock that still trades 27% below its March high. What remains is execution on cost and patience on price. Karen Carter’s job is the first one, and the second one is not hers to control.

This is a position you build slowly and hold through ugly quarters, funded by a 4.5% dividend that is now credible precisely because it was already halved. It is not a trade. If you need this to work by December, do not own it.

You are being paid 4.5% to wait for a cycle that everyone agrees will turn and nobody agrees on when — and the institutions who spent five years selling just bought it all back in ninety days.

ANALYST RATING

🟡 ACCUMULATE ON WEAKNESS

A 2-3% starter position with a 24-36 month horizon, adding on any move toward the $26-28 zone; the Q3 guide-down is a known event, so let it print before sizing up.

Sources & Further Reading

Dow Inc. Q2 2026 Earnings Call Transcript — Investing.com

Dow Q2 2026 slides: self-help drives margin surge, $3B target eyed — Investing.com

Karen S. Carter Appointed Chief Executive Officer — Dow Investor Relations

DOW Institutional Ownership and Shareholders — Fintel.io

Global ethylene and polyethylene cycle to bottom by 2028/2029 — ICIS

Petrochemicals in peril: oversupply crisis and energy transition — Wood Mackenzie

Top X (Twitter) Stocks — Trending Tickers — AltIndex

This is not financial advice. Always do your own research.

Week in Review: August 10–14 — What's Next for Investors

Colby Hill — Mon, 17 Aug 2026 01:55:28 GMT

In This Issue

Full Recap of Last Week
Upcoming Earnings Week
Macro Event Calendar
Macro Trends: Rates, Inflation & the Business Cycle
Market Volatility
Analyst Upgrades & Downgrades
Top 5 Big News Events
Capital Flows: Insiders, Whales & ETF Money

1. Full Recap of Last Week

Market Performance

S&P 500: +0.4% for the week to 7,785.76 — a third straight weekly gain, after a record close above 7,800 on Wednesday
Nasdaq Composite: +0.1% to 26,729.16 — also a third consecutive weekly gain
Dow Jones Industrial Average: -0.6% to 53,732.41 — snapping a two-week win streak
Russell 2000: +3.1% to 3,068.42 — small caps notched a fresh record and led the tape

Underneath the headline numbers it was a week of rotation. The Nasdaq 100 gained roughly 2.3% while the Dow slipped, and precious metals ripped — gold up about 3.0% and silver up 4.7%. Within the Magnificent 7, Tesla (+7.1%) and Nvidia (+2.8%) led, while Amazon (-3.5%), Alphabet (-3.3%) and Apple (-2.1%) lagged.

Earnings Results

Q2 reporting season is nearly wrapped, and it has been extraordinary. Per FactSet, with 88% of the S&P 500 reported, 86% beat EPS estimates — the highest rate since Q2 2021 — and blended earnings growth is running at 50.4% (32.0% excluding Alphabet and Amazon).

The headline report was Applied Materials on Thursday. EPS of $3.50 beat the $3.39 estimate, revenue rose 25% year over year to $9.12 billion, and Q4 guidance of $3.82-$4.22 landed well above the roughly $3.71 consensus. The stock still fell about 5% after hours — a textbook sell-the-news reaction after a huge run.

Elsewhere, Intel dropped roughly 5% after announcing a $15 billion common stock offering, and MarineMax soared 46% on a $1.5 billion buyout by Blackstone Infrastructure’s Safe Harbor Marinas at $53 a share.

Macro Event Results

July CPI (Wednesday): +0.1% monthly, 3.4% annual, down from 3.5%. Core CPI +0.2% monthly, 2.5% annual. Shelter alone drove about two-thirds of the headline gain.
July PPI (Thursday): flat on the month; 4.7% annual, down sharply from 5.5% in June.
July retail sales (Friday): -0.6% versus expectations for +0.2% — the biggest monthly drop since May 2025.
August consumer sentiment (Friday): the University of Michigan preliminary index fell to 51.0 from 55.2, versus 54.2 expected.

The message from the data was consistent: inflation is cooling, but so is the consumer.

2. Upcoming Earnings Week (Aug 17-21)

The focus shifts hard from AI infrastructure to the American shopper. Roughly 100 notable companies report.

Monday, Aug 17: A quiet start — no major S&P 500 names.
Tuesday, Aug 18: Home Depot reports before the open (call at 9:00 a.m. ET) — the cleanest read on housing-linked and big-ticket spending. Baidu and Klarna also report.
Wednesday, Aug 19: The heaviest day. Target (call 8:00 a.m.), Lowe’s (call 9:00 a.m.), TJX, Estee Lauder, and Analog Devices, which guided to about $3.9 billion in revenue and $3.30 adjusted EPS.
Thursday, Aug 20: Walmart — the single biggest tell on U.S. consumer resilience. Deere and Ross Stores also report.
Friday, Aug 21: Light on the corporate calendar.

What to watch: same-store sales, foot traffic, and full-year guidance. After Friday’s retail sales miss, the guidance will matter more than the quarter itself.

3. Macro Event Calendar (Aug 17-21)

Monday, Aug 17: August Empire State Manufacturing Index; NAHB Housing Market Index.
Tuesday, Aug 18: July housing starts and building permits; industrial production and capacity utilization; import and export prices; pending home sales. Reddit also officially joins the S&P 500 before the open.
Wednesday, Aug 19: FOMC minutes from the July 28-29 meeting, 2:00 p.m. ET. The main event of the week.
Thursday, Aug 20: Initial jobless claims; August Philadelphia Fed Manufacturing Index; Conference Board Leading Economic Index for July (June fell 0.2%).
Friday, Aug 21: S&P Global flash August PMIs (July: manufacturing 53.9, services 54.6); BLS state employment data.

Looking one week further out, the Jackson Hole symposium runs August 27-29. It will be Chair Kevin Warsh’s first as Fed chair, on the theme “Financial Innovation: Implications for Payments and Policy.”

4. Macro Trends: Rates, Inflation & the Business Cycle

Rates. The federal funds rate sits at 3.50%-3.75%. The Fed held in July on a 9-3 vote, but here is the twist: all three dissents wanted to raise rates, not cut them. After last week’s soft CPI and PPI, futures put roughly 68% odds on a hold in September and about 32% on a hike. Cuts are barely in the conversation.

Inflation. Headline CPI at 3.4% and core at 2.5% are both drifting lower, and producer prices decelerated meaningfully. But 3.4% is still well above the Fed’s 2% target, and energy costs tied to Middle East supply disruptions remain a live upside risk.

The business cycle. The economy is showing a genuine split. Corporate profits are booming on AI capital spending, with 50%-plus earnings growth and PMIs comfortably in expansion. Households are not: July payrolls fell 23,000, unemployment is 4.1%, retail sales dropped 0.6%, and sentiment just printed 51.0. That looks late-cycle for the consumer and mid-cycle for the capex boom. Which one converges toward the other is the question that defines the next six months.

5. Market Volatility

VIX: roughly 14.85, near multi-month lows and down about 4% on the week.
CNN Fear & Greed Index: 65, firmly in Greed territory.
Options positioning: subdued put demand alongside the low VIX, consistent with a market pricing very little near-term risk.

A 14-handle VIX alongside a Greed reading tells you positioning is comfortable. Low volatility does not predict a selloff, but it does mean there is little cushion if Wednesday’s FOMC minutes read more hawkish than expected.

6. Analyst Upgrades & Downgrades

Notable Upgrades

Abbott Laboratories (ABT) — Wolfe Research to Outperform from Peer Perform, $130 target
AbbVie (ABBV) — Wolfe Research to Outperform from Peer Perform, $300 target
Intuitive Surgical (ISRG) — Oppenheimer to Outperform from Perform, $500 target
Five Below (FIVE) — Jefferies to Buy from Hold, target raised to $350 from $210
Hilton Worldwide (HLT) — Deutsche Bank to Buy from Hold, $365 target
Wayfair (W) — Bernstein to Outperform from Market Perform, target raised to $125 from $100
Fox Corporation (FOXA) — JPMorgan to Overweight from Neutral, target raised to $82 from $70

Notable Downgrades

T-Mobile US (TMUS) — Wolfe Research to Peer Perform from Outperform
Alcon (ALC) — BNP Paribas to Neutral from Outperform, $79 target
CareTrust REIT (CTRE) — RBC Capital to Sector Perform from Outperform, target trimmed to $43 from $44
Tencent Music (TME) — China Renaissance to Hold from Buy, $9.30 target

Worth noting: two of the week’s biggest upgrades landed in healthcare, a sector that has badly lagged the AI-driven rally.

7. Top 5 Big News Events

The S&P 500 cleared 7,800 for the first time. Wednesday’s record close capped a third consecutive up week for the index.
The consumer cracked. Retail sales fell 0.6% and consumer sentiment dropped to 51.0 — both far worse than expected, and the clearest sign yet that high prices are biting.
Inflation cooled and September hike odds collapsed. CPI at 3.4% and a flat PPI knocked the probability of a September rate hike from above 50% down to roughly 32%.
Reddit was added to the S&P 500. Shares jumped 11% on the announcement; inclusion takes effect before Tuesday’s open, replacing AvalonBay Communities.
13F season revealed big repositioning. David Tepper’s Appaloosa concentrated into Magnificent 7 names, while Berkshire Hathaway added housing exposure and Alphabet.

8. Capital Flows: Insiders, Whales & ETF Money

ETF flows. Money is still pouring in — roughly $43.7 billion of net ETF issuance in the week ended August 5, after $46.5 billion the prior week. But the composition is rotating. On August 11 alone, SPY took in $3.56 billion while QQQ lost $3.28 billion and the semiconductor ETF SOXX shed $1.54 billion. Value and dividend strategies are the year’s quiet winners, with CGDV and AVLV each pulling in roughly $7 billion year to date.

Insider buying. Lumen Technologies CEO Kathleen Johnson bought 100,000 shares for about $613,000 on August 6. Coastal Financial CEO Eric Sprink bought 10,000 shares for $444,500 the same day. OppFi CEO Todd Schwartz picked up 60,000 shares for roughly $428,000 on August 13.

Insider selling. The sales dwarfed the buys. Arista Networks CEO Jayshree Ullal sold 763,029 shares for approximately $154.3 million on August 5. SharkNinja CEO Mark Barrocas sold 500,000 shares for $88.75 million on August 5 and 6.

The pattern is a familiar one: small, conviction-sized buying in beaten-down value names, and very large scheduled selling in stocks that have already run hard.

The Bottom Line

Markets are sitting near record highs, volatility is near lows, and corporate earnings are the strongest in five years. At the same time, the consumer is weakening, payrolls went negative, and the Fed’s internal debate is about whether to hike rather than cut. That is an unusual combination. It means Wednesday’s FOMC minutes and Thursday’s Walmart results carry more weight than a typical late-August week would suggest. Watch whether the retailers confirm what Friday’s data implied.

This newsletter is for informational purposes only and is not financial advice. Always do your own research before making investment decisions.

AMAT: The Best Quarter in Company History Got Sold Anyway

Colby Hill — Sun, 16 Aug 2026 11:18:31 GMT

August 16, 2026 | Deep Dive by Colby

OPENING HOOK

Why this story matters right now

Here is a number worth sitting with. $10.25 billion.

That is what Applied Materials told Wall Street it expects to do in revenue next quarter — a company that had never printed a $10 billion quarter in its 59-year history, guiding to one with a straight face. On the same call, it reported the highest sequential revenue growth in company history. Non-GAAP EPS of $3.50, up 41% year over year. Gross margin expansion for the 13th consecutive quarter.

The stock fell 5.12%.

Applied closed Friday at $507.18 on 13.1 million shares — roughly 1.7 times its normal volume — after opening the session down near $499. It is now 31% below its June 30 high of $739.67, and still 228% above the $154.47 it traded at last September. Both of those facts are true at the same time, which is the whole problem with this stock right now.

And here is the part almost nobody on X was talking about while the ticker trended Friday: institutions did not sell this quarter. According to Fintel, institutional holders went from 630.4 million shares in May to 868.3 million in August — an increase of roughly 238 million shares, the single largest accumulation in the dataset's five-year history. The people who file 13Fs and the people who post charts came to opposite conclusions about the same company.

INVESTMENT THESIS

The core reasons to own this stock

1. The margin story is structural, not cyclical. Thirteen straight quarters of year-over-year gross margin expansion is not a cycle — it is a mix shift. Semiconductor Systems non-GAAP operating margin hit 38.0% this quarter versus 33.2% a year ago, a 4.8-point improvement while revenue grew 27%. Applied is selling harder tools into harder problems, and pricing accordingly.

2. The Q4 guide implies acceleration, not deceleration. Guidance of $10.25 billion +/- $500 million against a $9.115 billion Q3 is another 12% sequential step, on top of a 15% sequential step. Companies at the top of a cycle do not guide like that. CFO Brice Hill specifically flagged DRAM, leading-edge foundry-logic, and advanced packaging as the drivers, and management raised its calendar-2026 Semiconductor Systems expectations mid-year.

3. The China overhang is already priced and already shrinking. China fell to 28% of revenue from 35% a year ago, and Applied has guided to a roughly $600 million fiscal-2026 revenue hit from expanded export controls. That is the bear case — and Applied grew 25% anyway. Meanwhile U.S. revenue nearly doubled to $1.367 billion, or 15% of the total, up from 9%. The customer base is re-shoring faster than the restrictions are biting.

4. The balance sheet gives management room to be aggressive. Applied holds $7.04 billion in cash, $2.20 billion in short-term investments and $5.27 billion in long-term investments against $6.54 billion of total debt — comfortably net cash. It generated record operating cash flow of $3.04 billion in a single quarter and still returned $860 million to shareholders. It is funding a $500 million Singapore campus and expanding manufacturing capacity to support projected demand through the end of the decade without touching the credit markets.

BUSINESS OVERVIEW

What Applied Materials actually does

Applied Materials makes the machines that make chips. Not the chips — the machines. When TSMC, Samsung, Intel, SK Hynix or Micron builds a fab, Applied sells them the deposition, etch, chemical-mechanical planarization, ion implantation and metrology tools that physically construct a transistor layer by layer. Founded in 1967 and headquartered in Santa Clara, it employs roughly 38,900 people and is the largest semiconductor equipment company in the world by revenue.

The business runs in two segments that behave very differently. Semiconductor Systems is the tool business — $7.04 billion this quarter, up 27% year over year, carrying a 55.4% non-GAAP gross margin. It is lumpy, capex-driven and tied directly to how many fabs the world is building. Applied Global Services (AGS) is the annuity — $1.78 billion, up 22%, selling spare parts, service contracts and factory automation software into an installed base that only grows. AGS margin expanded to 30.1% from 27.3%, and it is the reason Applied's earnings do not collapse in a downturn the way pure-play tool makers' do.

Within Semiconductor Systems, the mix is where the AI story lives. Foundry and logic remain 67% of segment revenue, but DRAM climbed to 26% from 22% — a direct read on high-bandwidth memory demand. Flash memory shrank to 7% from 9%. Applied's six new DRAM and advanced-packaging systems launched this quarter, including the Producer Avila 2 PECVD for stabilizing ultra-thin DRAM dies in 12- and 16-high HBM stacks, exist because HBM is the bottleneck in every AI accelerator being built today.

INDUSTRY OPPORTUNITY

A $150 billion equipment market where 80% of the growth is concentrated in three places

Wafer fab equipment spending is tracking toward the low $150 billion range in calendar 2026, and the composition of that spend matters more than the headline. Industry analysis points to leading-edge foundry-logic, DRAM and advanced packaging driving roughly 80% of WFE growth in 2026 and 2027. Those are precisely the three areas Applied called out as its Q4 strength.

This is a meaningful change from prior cycles. For a decade, WFE growth was driven by node transitions and capacity additions in mature logic — a market where Chinese domestic fabs were an enormous incremental buyer and where competition on price was real. The AI buildout has shifted the center of gravity to a small number of customers spending enormous sums on the hardest processes in the industry, where there are two or three qualified suppliers and switching costs are measured in years of requalification.

Advanced packaging is the newest and least appreciated leg. Hybrid bonding, through-silicon vias, chiplet architectures and glass substrates all require process control that did not exist five years ago. Applied introduced the Opta Quad CMP for hybrid bonding uniformity and the SEMVision G7AP for defect analysis across silicon, organic and glass substrates this quarter — products with essentially no incumbent competition because the application itself is new. Broadcom joined Applied's EPIC Center as an innovation partner specifically to co-develop advanced packaging technology, which tells you who is worried about this bottleneck.

COMPETITIVE ADVANTAGE

Why Applied might win

The single most important moat is breadth inside a fab where breadth compounds. ASML owns lithography outright and nobody is taking it; KLA owns process control and metrology; Lam Research is the closest true competitor in etch and deposition; Tokyo Electron and SCREEN cover broad portfolios regionally. Applied is the only company that plays across deposition, etch, planarization, implantation, metrology and services at scale — which means when a customer is co-optimizing a new transistor architecture, Applied is the only vendor that can tune five process steps together rather than one.

That advantage is getting formalized. The EPIC Center — Applied's collaborative R&D facility — now has 11 engagements, expanded this quarter to include Broadcom, UC Berkeley and SCREEN Semiconductor Solutions. SCREEN is nominally a competitor in wafer cleaning; it joined anyway, because co-optimizing cleaning with Applied's deposition beats optimizing either alone. When your rivals are paying to work inside your R&D facility, you have a structural position, not a product lead.

The services business deepens it further. AGS revenue of $1.78 billion grows with the installed base regardless of whether anyone is buying new tools, and its 30.1% operating margin is now expanding faster than the tool business's. Lam and Tokyo Electron have service arms too, but neither carries Applied's installed-base scale. This is the line item that turns a cyclical equipment maker into something closer to an industrial compounder — and it is the reason the market is willing to pay 40 times earnings for a company that was a 15-multiple stock a decade ago.

FINANCIAL DEEP DIVE

The numbers: what they mean and what to watch

Metric                     Value
-------------------------  ------------------
Last Close (Aug 14, 2026)  $507.18 (-5.12%)
52-Week Range              $154.47 - $739.67
Off 52-Week High           -31.4%
Market Cap                 $402.7B
P/E (trailing)             43.7x
P/B                        16.8x
Dividend Yield             0.78%
Q3 FY26 Revenue            $9.115B (+25% YoY)
Q3 Non-GAAP EPS            $3.50 (+41% YoY)
Q3 GAAP EPS                $3.17 (+43% YoY)
Non-GAAP Gross Margin      50.4% (+1.5 pts)
Non-GAAP Op Margin         34.0% (+3.3 pts)
Q3 Free Cash Flow          $2.33B
Q4 FY26 Revenue Guide      $10.25B +/- $0.5B
Q4 FY26 EPS Guide          $4.02 +/- $0.20
Avg. Daily Volume (30d)    7.72M
Aug 14 Volume              13.13M (1.7x)

The gap between GAAP EPS of $3.17 and non-GAAP EPS of $3.50 is the entire reason this stock traded down, and it deserves precision. The difference is almost entirely a $220 million unrealized loss on strategic investments — a $0.31 per-share mark-to-market hit on Applied's venture portfolio. It is not an operating result. It is the same line item that added $0.39 per share in the year-ago quarter and has contributed a $1.01 per-share gain across the first nine months of fiscal 2026. Traders who screened on GAAP EPS saw a miss. Traders who read the reconciliation saw noise.

The number that actually deserves scrutiny is on the balance sheet. Accounts receivable jumped to $7.69 billion from $5.19 billion at the October fiscal-year close — a 48% increase against nine-month revenue growth of 11%. Some of that is timing on a record quarter that back-loaded, and contract liabilities rose to $3.27 billion from $2.57 billion, which is a genuinely good sign about future shipments. But receivables growing four times faster than revenue is the kind of thing that resolves quietly or does not resolve at all. Watch it next quarter.

Everything else is clean. Operating cash flow of $3.04 billion was a record. Free cash flow of $2.33 billion covered the $440 million buyback and $420 million dividend nearly three times over. Inventory rose modestly to $6.56 billion — consistent with building for a guided $10.25 billion quarter, not with channel stuffing. R&D spending grew to $1.10 billion from $901 million, up 22%, roughly in line with revenue. The effective non-GAAP tax rate of 11.0% is low and worth noting as a tailwind that may not persist.

MANAGEMENT TEAM

Who is running this?

Gary Dickerson has been President and CEO since September 2013 — thirteen years, which makes him one of the longest-tenured chief executives in large-cap semiconductors. He took over a company with roughly $7 billion in annual revenue; fiscal 2026 will land north of $34 billion. He has navigated the 2015 collapse of the Tokyo Electron merger under antitrust pressure, the 2018-19 memory downturn, the COVID supply shock, and now the most aggressive export-control regime in the industry's history. His public framing of China has been unusually blunt for a CEO with a quarter of revenue there — he has stated plainly that Applied can no longer supply China's memory and mature-node markets, and that foreign competitors are reaching customers Applied cannot.

Brice Hill serves as Senior Vice President and CFO, and his fingerprints are on the margin story. The 13-consecutive-quarter gross margin expansion streak is the kind of metric a CFO builds a compensation framework around, and Applied explicitly uses non-GAAP measures as performance benchmarks in its executive comp program — which is worth knowing when evaluating how hard the company leans on the non-GAAP presentation.

One governance item belongs in any honest write-up. Applied took a $253 million charge in fiscal 2026 to settle with the U.S. Commerce Department's Bureau of Industry and Security over a previously disclosed export-controls compliance matter. That is a real cost and a real black mark. It is also, at roughly 0.7% of annual revenue, resolved and behind them — but it is the second export-related event in as many years, and it suggests the compliance surface area at this company is genuinely difficult to manage.

BULL CASE

5 reasons this stock could dramatically outperform

1. The Q4 print confirms the acceleration and the multiple re-rates. Applied reports fiscal Q4 in mid-November. If it delivers $10.25 billion and $4.02, fiscal 2026 lands at roughly $34.3 billion in revenue and $12.75 in non-GAAP EPS. At Friday's close that is 39.8x — expensive, but on a number that grew 41% year over year with management guiding to another strong growth year in 2027. A single quarter of confirmation flips this from a multiple debate to an earnings-power debate.

2. DRAM is early, not late. DRAM went from 22% to 26% of Semiconductor Systems revenue in twelve months, and every HBM roadmap in the industry calls for 12-high and 16-high stacks that require exactly the deposition and metrology tools Applied just launched. Memory capex cycles historically run three to four years once they start. This one started roughly a year ago.

3. Advanced packaging is a land grab with no incumbent. Hybrid bonding, TSV fill and glass-substrate defect analysis are new enough that market share is being set right now. Applied shipped six new systems into this category this quarter and has Broadcom co-developing inside its EPIC Center. Share won here is share held for a decade.

4. Institutions are accumulating aggressively into the weakness. The 238-million-share increase in institutional holdings between May and August — from 630.4 million to 868.3 million shares across 4,900 filers — is the largest single-period accumulation in five years of Fintel data. Large funds do not build positions that size on momentum; they build them on models. Somebody with a research budget did the work and decided the June-to-August drawdown was an entry.

5. The gap to the Street's targets is unusually wide. The three-month average analyst price target sits near $677, roughly 33% above Friday's close, and even Morgan Stanley's Equal Weight rating carries a $646 target. When the most cautious major bank on the name has a target 27% above spot, the sell-side is telling you the stock is dislocated from its own models — not that the models are wrong.

BEAR CASE

5 reasons this could disappoint

1. You are paying 40 times earnings for a cyclical. This is the core risk and no amount of AI narrative erases it. Applied has spent most of its history trading between 12 and 20 times earnings because equipment demand is violently cyclical. At 43.7x trailing and 16.8x book, the stock is priced as though the cycle has been abolished. It has not been. The 31% drawdown from the June high is what a mild change of mind looks like; a real one looks worse.

2. China is 28% of revenue and the policy direction is one-way. Applied has already guided to a $600 million fiscal-2026 revenue hit from expanded export controls, and Dickerson has confirmed the company is now locked out of China's memory and mature-node markets entirely. China revenue was $2.51 billion this quarter — down only slightly in dollars but down seven points in mix. If Beijing retaliates, or if Washington tightens again, there is no version of that where the number goes up.

3. Customer concentration is extreme and getting worse. Taiwan (22%), Korea (17%) and China (28%) account for 67% of revenue, and within those geographies the buyers are a handful of names — TSMC, Samsung, SK Hynix. A single customer pushing a fab construction schedule out two quarters moves Applied's revenue by more than most companies earn in a year. Applied acknowledges the concentrated nature of its customer base as a risk factor in its own filings.

4. Receivables are growing four times faster than revenue. AR at $7.69 billion versus $5.19 billion nine months earlier is a 48% increase against 11% nine-month revenue growth. The benign explanation is a back-loaded record quarter. The unfriendly explanation is that revenue is being recognized on shipments that customers are slow to pay for — which, at a moment when Applied is also building inventory for a guided $10.25 billion quarter, would be the classic setup for a working-capital squeeze if demand hesitates.

5. Domestic Chinese competition is being actively subsidized to replace Applied. Every export restriction that locks Applied out of a Chinese fab also hands Naura, AMEC and their peers a captive customer and a decade of state-funded runway. Dickerson has said foreign competitors are already reaching customers Applied cannot. Mature-node equipment is the easiest place to build a domestic alternative, and once a Chinese fab qualifies a domestic tool for a process step, that step is gone permanently — even if policy reverses.

VALUATION

Building the model: Bear, Base, Bull

Scenario  Key Assumption           Price Target  Probability
--------  -----------------------  ------------  -----------
Bear      FY27 EPS ~$12.00 at 26x  $315          25%
Base      FY27 EPS ~$15.75 at 36x  $570          50%
Bull      FY27 EPS ~$18.60 at 42x  $780          25%

Run the probability weighting and you get an expected value of $558.75, roughly 10% above Friday's close of $507.18. That is a positive number, and it is a much smaller positive number than the bull case rhetoric around this stock implies. The distribution is what matters: a 25% chance of losing 38% against a 25% chance of making 54%. The math is favorable but the variance is enormous, and anyone sizing this position as though it were a defensive industrial is making a category error.

The base case deserves a word on the multiple. Thirty-six times forward earnings is not a heroic assumption for a company compounding EPS above 40% with expanding margins and a net cash balance sheet — but it is an assumption that the market continues to treat semiconductor equipment as a secular AI play rather than a cyclical. That single judgment call is worth more to the outcome than any operating variable in the model. If the market's frame shifts back to cyclical, the bear case is not a 25% outcome. It is the outcome.

KEY METRICS TO WATCH

The five signals that tell you if the thesis is working

01 — The fiscal Q4 revenue print in mid-November. Green flag is $10.25 billion or better with non-GAAP EPS at or above $4.02. Yellow flag is anything at the low end of the $500 million guidance band, because a company guiding to record growth that lands at the bottom of its own range is a company that lost visibility during the quarter.

02 — Accounts receivable as a percentage of quarterly revenue. At $7.69 billion against $9.115 billion in quarterly revenue, AR now represents 84% of a quarter's sales. Green flag is that ratio falling toward 70% next quarter as the back-loaded shipments collect. Yellow flag is it holding above 80% or climbing, which would mean the collection problem is structural rather than seasonal.

03 — DRAM as a share of Semiconductor Systems revenue. It moved from 22% to 26% year over year. Green flag is 28% or higher next quarter, confirming HBM capacity additions are still accelerating. Yellow flag is a reversal back toward 22-23%, which would signal memory customers are pausing to digest.

04 — China revenue in absolute dollars, not percentage. China was $2.506 billion this quarter versus $2.548 billion a year ago — nearly flat in dollars while shrinking seven points in mix. Green flag is China holding above $2 billion while total revenue grows, meaning the mix shift is dilution rather than decline. Yellow flag is China dropping below $1.8 billion, which would mean the export controls are now taking real dollars off the table.

05 — Applied Global Services operating margin. AGS hit 30.1% versus 27.3% a year ago, and this is the metric that determines whether Applied deserves a premium multiple through a downturn. Green flag is 30% or better sustained across the next two quarters. Yellow flag is any retreat toward 28%, because a services business losing margin leverage while the installed base grows means pricing power is weaker than the story requires.

MY TAKE

The bottom line

I keep coming back to the divergence. On Friday, retail traders looked at a GAAP EPS number distorted by a mark-to-market loss on a venture portfolio and sold Applied Materials down 5% on 1.7 times normal volume. Over the preceding three months, institutions bought 238 million shares. Both groups had access to the same 10-Q. One of them read the reconciliation table.

That said, I am not going to pretend the bears have no case, because they have a good one. This is a company that has traded at 12 to 20 times earnings for most of its existence, and it is asking you to pay 40. The entire bull thesis rests on a single judgment — that AI infrastructure spending has changed the character of the wafer fab equipment cycle rather than merely extending it. If that is right, $507 will look cheap in three years. If it is wrong, the June-to-August drawdown from $739 to $507 was the appetizer.

What tilts me positive is the quality of the underlying operating result, which is not ambiguous in any way. Thirteen consecutive quarters of gross margin expansion. Record operating cash flow. A services segment expanding margin faster than the tool segment. Net cash. A CEO thirteen years into the job who has been publicly, uncomfortably honest about the China problem rather than managing the narrative. Companies that are quietly deteriorating do not produce that pattern of evidence.

The expected value on my numbers is roughly 10% upside, with a 25% chance of a 38% drawdown. That is not a table-pounding setup. It is a good business at a demanding price, in a stock that has already given back a third of its gains and where the largest, best-resourced investors chose the drawdown as their entry point.

You are not buying a cheap stock. You are buying the best-positioned company in a $150 billion market at 40 times earnings, and betting that the AI capex cycle has three more years in it rather than one.

ANALYST RATING

🟡 ACCUMULATE ON WEAKNESS

A 2-3% starter position sized to add on any retest of the $450-470 zone, with an 18-24 month horizon and a hard reassessment if fiscal Q4 revenue lands at the low end of the $10.25 billion guide. This is a conviction business at a non-conviction price — let the volatility work for you rather than paying up on strength.

Sources & Further Reading

Applied Materials Announces Third Quarter 2026 Results

Fintel — AMAT Institutional Ownership

Applied Materials Stock Fell 5% After a Record Q3 2026 Quarter (TIKR)

Applied Materials sees weaker China spending in 2026 on tighter U.S. curbs (Reuters via Investing.com)

Stocks making the biggest moves midday: Sandisk, Applied Materials, AeroVironment, Cisco, Reddit (CNBC)

Applied Materials (AMAT) Stock Forecast and Price Targets (MarketBeat)

This is not financial advice. Always do your own research.

VOR: The Biotech That Fired 95% of Its Staff and Came Back as a $1.4 Billion Bet

Colby Hill — Sat, 15 Aug 2026 11:23:58 GMT

August 15, 2026 | Deep Dive by Colby

OPENING HOOK

Why this story matters right now

In May 2025, Vor Bio told the world it was winding down. Clinical trials halted. Manufacturing stopped. Roughly 95% of the workforce gone. The board hired bankers to explore a sale, a merger, a liquidation — whatever could be salvaged from a cell-therapy company that had run out of road.

Fourteen months later, Vor Bio is the single most-discussed ticker on X.

On Friday, August 14, VOR logged 453 mentions on X, roughly 1.8 times the volume of the second-place name and more than double the chatter around NVIDIA, a company 4,000 times its size. The sentiment skew was bearish. The volume was thin — just 417,565 shares against a 30-day average north of 813,000. And the stock closed at $23.03, down 1.1% on the day but up 27% in a month.

That combination — enormous attention, modest volume, a stock quietly grinding higher while the crowd argues — is what a genuine controversy looks like before it resolves.

Here is what everyone is arguing about. Vor Bio no longer develops the drugs it spent nine years and hundreds of millions of dollars inventing. It develops telitacicept, a dual-target fusion protein it licensed from Chinese biotech RemeGen in June 2025 for $45 million in cash, $80 million in penny warrants, and up to $4.105 billion in future milestones. The drug already has five commercial approvals in China. It has never been approved anywhere else. The trial that decides whether it ever will be reads out in the first half of 2027.

INVESTMENT THESIS

The core reasons to own this stock

1. You are buying de-risked Phase 3 data at a Phase 2 price. Telitacicept has already cleared registrational trials in China across five indications — systemic lupus, rheumatoid arthritis, generalized myasthenia gravis, Sjögren's disease, and IgA nephropathy. Two of those datasets were good enough for The New England Journal of Medicine, including the TELIGAN IgA nephropathy trial published in May 2026. The scientific risk that normally defines a clinical-stage biotech — does the molecule do anything at all — has largely been answered. What remains is a narrower question: does it replicate in Western patients under FDA-grade scrutiny.

2. The balance sheet buys time the market is not pricing. Vor Bio ended June with $466.1 million in cash and marketable securities, and after $48.4 million of July at-the-market sales carries a pro-forma balance near $514.5 million — funding operations into early 2029. Against a $1.36 billion market cap, that means the entire drug franchise, the global rights, and every milestone in the RemeGen agreement are being valued at roughly $850 million of enterprise value. For a molecule with five approvals in the world's second-largest pharmaceutical market, that is a strikingly small number.

3. Institutions have rebuilt the shareholder base from scratch. After the 2025 collapse and the 1-for-20 reverse split, split-adjusted institutional holdings bottomed near 4.1 million shares. As of the latest Fintel data they sit at 50.2 million shares across 144 filers — roughly 85% of the 59.2 million shares outstanding. TCG Crossover, Fidelity, BlackRock, RA Capital, Frazier Life Sciences, and Caligan Partners are all on the register. This is not a retail meme with a story attached. It is a professionally owned asset that retail has only recently discovered.

4. Sjögren's disease is the free option nobody is modeling. Telitacicept is the first and only approved therapy for Sjögren's disease in China, having hit statistically significant ESSDAI reductions at both 160mg and 80mg sustained through Week 48. In the United States there is still no approved disease-modifying treatment for primary Sjögren's. If the global UPSTREAM SjD trial replicates the Chinese result, Vor Bio owns a first-in-class approval in a disease with zero incumbents.

BUSINESS OVERVIEW

What Vor Bio actually does

Vor Bio today is a 76-person company in Boston whose entire reason for existing is one molecule. Telitacicept is a recombinant fusion protein that simultaneously blocks BLyS and APRIL, the two cytokines that keep autoreactive B cells and plasma cells alive. Shut both down and you starve the machinery that produces the autoantibodies driving lupus, myasthenia gravis, Sjögren's, IgA nephropathy, and rheumatoid arthritis. It is a mechanism with real biological pedigree — Benlysta validated BLyS blockade alone; telitacicept's argument is that hitting both targets is meaningfully better than hitting one.

The business model is licensing arbitrage. RemeGen developed and commercialized telitacicept in China. Vor Bio bought the rest of the planet — everywhere except China, Hong Kong, Macau, and Taiwan — for an upfront that would not fund a single Phase 3 trial from scratch. In exchange, RemeGen collects tiered royalties from high-single-digit to low-double-digit percentages of net sales, plus up to $4.105 billion in development and commercial milestones. Vor Bio's job is narrow and expensive: run Western registrational trials, file with the FDA and EMA, and build a commercial organization.

That job is currently two trials. UPSTREAM MG is the Phase 3 in generalized myasthenia gravis, enrollment ongoing, topline expected in 1H 2027. UPSTREAM SjD is the Phase 3 in primary Sjögren's disease, roughly 250 patients, first patient dosed earlier this year. Everything else — the lupus franchise, the IgA nephropathy franchise, rheumatoid arthritis — sits behind those two readouts as optionality.

INDUSTRY OPPORTUNITY

Autoimmune is where the money went when oncology got crowded

The generalized myasthenia gravis market was worth roughly $3.78 billion in 2024 and is projected to reach $15.57 billion by 2033, a 17.1% compound annual growth rate. That is not a niche. It is one of the fastest-repricing rare-disease categories in pharma, and the reason is simple: for decades gMG patients got steroids and plasma exchange, and in the last five years they got FcRn blockers, complement inhibitors, and B-cell depleters that actually work.

Sjögren's is the larger and stranger opportunity. The disease affects far more people than myasthenia gravis, the market was already valued near $1.9 billion on symptomatic care alone, and yet in the United States there is still no approved therapy that modifies the underlying disease. Amgen's dazodalibep, J&J's nipocalimab, Bristol's deucravacitinib, and Novartis's ianalumab are all racing toward that finish line. Telitacicept is the only one of them that has already crossed it somewhere.

The macro tailwind underneath both is the industry's pivot into immunology. Large-cap pharma spent 2024 and 2025 paying enormous premiums for late-stage autoimmune assets because the oncology aisle got expensive and crowded. A dual-target B-cell modulator with five approvals and NEJM-grade data is precisely the profile that gets bought — which is why the acquisition tail on this stock is not a fantasy.

COMPETITIVE ADVANTAGE

Why Vor Bio might win

The moat here is not the molecule. It is the data that already exists.

Every competitor entering generalized myasthenia gravis or Sjögren's is running Phase 3 trials against an unknown. Vor Bio is running Phase 3 trials against a known — RemeGen has already reported a 98.1% MG-ADL response rate in its Chinese Phase 3 in myasthenia gravis, and the Sjögren's registrational data was strong enough to earn China's first-ever approval in the indication. Vor Bio is not asking whether telitacicept works. It is asking whether it works the same way in Cleveland as it does in Chengdu. That is a materially different — and cheaper — question to answer.

Compare that to the field. argenx is the incumbent and the standard, with Vyvgart generating $573 million in a single quarter as far back as Q3 2024, an FDA label expanded in May 2026 to cover all adult gMG serotypes, and a prefilled syringe that lets patients dose at home. That is a genuinely formidable position. Johnson & Johnson has Imaavy. Amgen got Uplizna approved for gMG in December 2025. UCB fields Rystiggo and Zilbrysq. Behind them sit Immunovant's IMVT-1402, Cartesian's Descartes-08, Novartis's remibrutinib, AstraZeneca's gefurulimab, and Regeneron's pozelimab-cemdisiran combination.

Vor Bio will not out-commercialize argenx. It does not need to. FcRn blockers work by clearing existing antibodies — a maintenance mechanism requiring chronic dosing. Telitacicept works upstream, on the cells that make the antibodies in the first place. Those are complementary rather than substitutable mechanisms, and in a $15 billion market growing at 17% a year, the third or fourth product with a differentiated mechanism still builds a very large business.

FINANCIAL DEEP DIVE

The numbers: what they mean and what to watch

Metric                     Value
-------------------------  ---------------------------
Last Close (Aug 14, 2026)  $23.03
Day Change                 -1.1%
52-Week Range              $6.50 - $50.20
1-Year Return              -46.2%
1-Month Return             +27%
Market Cap                 $1.36B
Shares Outstanding         59.2M
Public Float               31.5M
Enterprise Value (est.)    ~$850M
Cash (pro-forma)           $514.5M
Cash Runway                Into early 2029
Q2 2026 Net Loss           $62.8M
P/E Ratio                  Not meaningful
P/B Ratio                  -6.26
Avg. Volume (30d)          813,626
Volume (Aug 14)            417,565
Institutional Ownership    50.2M shares / 144 filers
Wedbush Price Target       $27 (Neutral, Aug 12, 2026)

Start with the number that looks broken. Price-to-book is negative 6.26, which means Vor Bio reports negative shareholders' equity despite holding half a billion dollars in cash. That is not distress — it is accounting. The $80 million of warrants issued to RemeGen carry an exercise price of $0.0001, which makes them economically equivalent to shares and, under the relevant accounting treatment, a liability that gets marked to market every quarter. As VOR's share price climbed, that liability inflated and drove book value below zero. Understand this before you let a screener talk you out of the stock.

The burn is the number that actually matters. A $62.8 million quarterly net loss against $514.5 million of pro-forma cash is roughly eight quarters of runway on a static basis, which reconciles cleanly with management's "into early 2029" guidance. That is enough to fund UPSTREAM MG through topline in 1H 2027 and UPSTREAM SjD behind it without a dilutive raise — but only if the burn does not accelerate. It will accelerate. Building a US commercial organization ahead of a potential 2028 launch costs real money, and the $48.4 million of at-the-market sales in July is the tell that management intends to keep the balance sheet fat rather than thin.

Watch the float. Only 31.5 million of the 59.2 million shares outstanding trade freely, while institutions hold 50.2 million. That gap is the mechanical reason a stock with a $1.36 billion market cap can move 5% on 400,000 shares. It cuts both ways — thin floats amplify good news and gap down on bad.

MANAGEMENT TEAM

Who is running this?

Jean-Paul Kress, M.D. took over as CEO on June 25, 2025 — the same day the RemeGen deal was announced. That timing is not a coincidence; the license and the leadership change were a single act of corporate reinvention, replacing Robert Ang and the cell-therapy strategy in one motion.

Kress's résumé is the strongest argument for the plan. He was most recently CEO of MorphoSys, which he steered into a $2.9 billion buyout by Novartis. Before that he ran Syntimmune through its acquisition by Alexion. He sits on the board of Sanofi. This is an executive with a demonstrated pattern of taking a single-asset biotech, running the clinical and regulatory playbook, and selling the company to large-cap pharma at a premium. If you are underwriting Vor Bio as an eventual acquisition target rather than a standalone commercial operator, Kress is the reason that thesis is credible.

The board has been rebuilt around the same idea. David Zaccardelli joined in July 2026, following the additions of Alexander (Bo) Cumbo and Michel Detheux, Ph.D. — a slate weighted toward commercial-stage biotech operators and dealmakers rather than the discovery scientists who populated the old Vor.

The alignment question is the honest caveat. The company has 76 employees and one asset. There is no diversification, no second shot, and no revenue. Management's incentives are heavily equity-weighted, which means they are pointed at exactly the same outcome shareholders want — and exposed to exactly the same binary.

BULL CASE

5 reasons this stock could dramatically outperform

1. UPSTREAM MG replicates in 1H 2027 and the stock reprices in a single session. RemeGen's Chinese Phase 3 reported a 98.1% MG-ADL response rate. If the global trial lands anywhere near that, Vor Bio goes overnight from a speculative license play to a company with a Western registrational dataset in a market growing 17% annually toward $15.57 billion. Biotechs that clear this bar routinely double.

2. Sjögren's delivers a first-in-class US approval with no incumbent to displace. Telitacicept is already the first and only approved Sjögren's therapy in China, with significant ESSDAI reductions at both doses sustained through Week 48. The US has nothing approved. Winning an empty category is worth far more than taking share in a crowded one, and virtually no sell-side model gives Vor Bio credit for it today.

3. The acquisition tail is unusually live. A $1.36 billion market cap with $514.5 million of cash means an acquirer is paying roughly $850 million for global ex-China rights to a five-indication autoimmune franchise. Kress sold MorphoSys to Novartis for $2.9 billion. Large-cap pharma is actively hunting late-stage immunology. The math is not hard for a strategic buyer to run.

4. Institutional accumulation is still in progress, not finished. Split-adjusted institutional holdings went from 4.1 million shares in late 2025 to 50.2 million in the latest filing — a twelvefold rebuild in under a year, and the most recent quarter added roughly 9 million shares on its own. That is a trend line, not a plateau. With only 31.5 million shares in the public float, continued institutional buying has to come out of a very small pool.

5. The IgA nephropathy and lupus franchises are entirely unpriced optionality. TELIGAN showed substantial proteinuria reductions with eGFR holding roughly stable through 39 weeks while placebo declined — data good enough for NEJM. The Chinese lupus Phase 3 also landed in NEJM. Neither indication has a Western trial running yet, which means neither carries cost today and both carry upside if the MG readout de-risks the platform enough to fund expansion.

BEAR CASE

5 reasons this could disappoint

1. Chinese trial data has a documented history of not replicating in Western populations. This is the entire bear thesis in one sentence. Differences in patient genetics, standard of care, concomitant medication, endpoint administration, and trial conduct have repeatedly turned spectacular Chinese results into ambiguous global ones. A 98.1% response rate is, frankly, a number that invites skepticism. If UPSTREAM MG delivers a statistically real but clinically modest effect, the stock does not go to zero — it goes to cash value, which is roughly $8.70 per share.

2. It is one molecule, one company, no plan B. Vor Bio deliberately destroyed its own optionality — trem-cel and VCAR33 discontinued, intellectual property sold, 95% of staff gone. There is no second program to pivot to if telitacicept stumbles. Seventy-six employees and a single asset is the most concentrated risk profile in the mid-cap biotech universe.

3. argenx is not standing still, and the gMG window may be closing. Vyvgart was already doing $573 million a quarter two years ago. The May 2026 label expansion covers all adult serotypes. The prefilled syringe moves patients to at-home self-administration. By the time telitacicept could plausibly launch in 2028, physicians will have four to six years of Vyvgart experience, established prescribing habits, and payer contracts locked in. Being fourth to market with a differentiated mechanism is a real business; it is not the business the current narrative implies.

4. Dilution is a live and recurring risk. Management sold $48.4 million of stock through the at-the-market program in July alone, on top of a $175 million private placement at the RemeGen deal, a subsequent $75 million placement, and a $119.7 million shelf prospectus. The pattern is unambiguous: this management team raises capital opportunistically into strength. Every rally is a potential supply event, and the current share count of 59.2 million is not a ceiling.

5. The timeline is long and the news flow between now and then is thin. Topline UPSTREAM MG is 1H 2027. That is potentially twelve months away with no company-changing catalyst in between — just enrollment updates, quarterly burn, and whatever RemeGen does in China. Meanwhile Wedbush maintains a Neutral rating even after raising its target to $27, and the broader analyst group is split with roughly as many hold and sell ratings as buys. A stock that has already run 27% in a month into a catalyst-free stretch is vulnerable to simple boredom.

VALUATION

Building the model: Bear, Base, Bull

Scenario  Key Assumption                             Target  Prob.
--------  -----------------------------------------  ------  -----
Bear      MG readout misses; stock trades to cash    $9      30%
Base      MG hits; US/EU path clear; gMG value only  $32     45%
Bull      MG + SjD both hit; franchise or takeout    $70     25%

Run the arithmetic and the probability-weighted expected value lands at roughly $34.60 per share — about 50% above Friday's close of $23.03. That spread is what a properly priced binary looks like: enough upside to justify the position, enough downside to keep it small.

The bear case is unusually well-defined here, and that is a feature. With $514.5 million of pro-forma cash against 59.2 million shares, the floor is approximately $8.70 per share in net cash — call it $9 accounting for burn between now and a 2027 readout. Failure costs you roughly 61%. Success in the base case pays roughly 39%, and the bull case pays 204%. The asymmetry is not enormous, but it is positive, and it improves materially if you believe the Chinese data is more predictive than the market's discount implies.

What the market is effectively pricing at $23 is somewhere between the bear and base cases — meaning consensus assigns real weight to the replication risk and essentially zero weight to Sjögren's, IgA nephropathy, or an acquisition. Those are the three places where this thesis makes money.

KEY METRICS TO WATCH

The five signals that tell you if the thesis is working

01 — UPSTREAM MG enrollment pace. Management has said enrollment "remains on track" for a 1H 2027 topline. Green flag: the company confirms completion of enrollment on the next two earnings calls. Yellow flag: the phrase "on track" quietly disappears, or the topline guidance slips to 2H 2027. Enrollment slippage in a single-asset biotech is the earliest and most reliable warning sign there is.

02 — Quarterly cash burn versus the $62.8 million baseline. Green flag: burn holds between $60 and $70 million per quarter, keeping the early-2029 runway intact. Yellow flag: burn climbs past $80 million as commercial buildout accelerates, which pulls the runway forward and makes a dilutive raise more likely before the MG readout rather than after it.

03 — At-the-market issuance in the quarterly filings. July's $48.4 million of ATM sales is the number to benchmark against. Green flag: ATM activity slows materially, signaling management believes the stock is undervalued. Yellow flag: another $40 million-plus quarter of issuance, which tells you insiders think $23 is a fine price to sell at.

04 — Institutional share count in the next 13F cycle. The latest Fintel data shows 50.2 million shares held by 144 filers, up roughly 9 million shares quarter over quarter. Green flag: the count keeps climbing and new tier-one crossover funds appear on the register. Yellow flag: the count flattens or the specialist biotech names — RA Capital, Frazier, TCG Crossover — begin trimming.

05 — RemeGen's China commercial trajectory. Vor Bio does not book China revenue, but RemeGen's telitacicept sales are the only real-world evidence of how the drug performs in actual clinical practice across five approved indications. Green flag: accelerating China uptake, particularly in Sjögren's, where the drug is the sole approved option. Yellow flag: decelerating scripts or pricing pressure, which would suggest the clinical benefit is less compelling in practice than in trials.

MY TAKE

The bottom line

I find this setup genuinely interesting, and I want to be precise about why. Most speculative biotech asks you to underwrite whether a molecule works. Vor Bio asks something narrower — whether a molecule that demonstrably works in one population works the same way in another. That is still a real risk. It is a smaller one.

The market is not paying for that distinction. At $23.03, you are paying roughly $850 million of enterprise value for global rights outside China to a drug with five approvals, two NEJM publications, and Phase 3 programs running in two indications, one of which has no approved competitor in the United States. Meanwhile the balance sheet carries $514.5 million and funds the business into 2029, and the shareholder register has been rebuilt from 4.1 million split-adjusted institutional shares to 50.2 million in under a year by exactly the funds that specialize in this kind of asset.

The bear case is equally clear-eyed, and I take it seriously. Chinese efficacy data has burned Western investors before, a 98.1% response rate strains credulity, argenx is a genuinely excellent competitor with a four-year head start, and this management team has shown it will sell stock into every rally. There is no second asset. If UPSTREAM MG disappoints in 1H 2027, you are holding a cash shell at roughly $9 and hoping someone buys the corporate structure.

What tips me positive rather than neutral is Kress. A CEO who sold MorphoSys to Novartis for $2.9 billion did not join a 76-person company to run a commercial launch in 2029. He joined to de-risk one asset and sell it. That is a shorter path than the standalone model implies, and it is not reflected in the current price.

You are risking about 61% to make about 204%, on a drug that already works somewhere else — the only question left is geography, and the market is pricing that question as if it were biology.

ANALYST RATING

⚡ SPECULATIVE BUY

A 1–2% position for investors who can tolerate a single-catalyst binary and hold through a quiet stretch into the 1H 2027 UPSTREAM MG readout. Size it so a move to cash value at $9 is survivable, because that is a live outcome — and do not add on strength given management's demonstrated appetite for at-the-market issuance.

Sources & Further Reading

Vor Bio Reports Second Quarter 2026 Financial Results and Provides Corporate Update — BioSpace

Vor Bio Enters into Exclusive Global License Agreement with RemeGen — GlobeNewswire

Telitacicept Receives NMPA Approval for the Treatment of Sjögren's Disease in China — Vor Bio IR

Vor Bio Announces Publication of TELIGAN Interim Analysis in The New England Journal of Medicine

Vor Biopharma Institutional Ownership — Fintel.io

Vor, with new CEO, changes course to target autoimmune disease — BioPharma Dive

Top X (Twitter) Stocks — Trending Tickers — AltIndex

This is not financial advice. Always do your own research.

UMAC: The Tariff Wall That Just Repriced America's Drone Supply Chain

Colby Hill — Fri, 14 Aug 2026 15:46:50 GMT

August 14, 2026 | Deep Dive by Colby

OPENING HOOK

Why this story matters right now

Here is a number worth sitting with. Institutions owned 17,000 shares of Unusual Machines in June 2024. Today they own 37.1 million.

That is not a rotation. That is a discovery.

Yesterday, President Trump signed a proclamation imposing a 100% tariff on imported drones weighing more than 25 kilograms or carrying thermal imaging, along with their docking stations and critical components — and a 25% tariff on smaller drones and remaining parts. The stock opened this morning and ran to $34.93, a fresh 52-week high, on nearly 9.6 million shares against a 30-day average of 3.7 million. It is up 25% today alone and 370% from its November low of $7.25.

Eight days earlier the company had already posted the number that mattered: Q2 revenue of $16.7 million, up 687% year over year and 107% sequentially, against a consensus of $9.19 million. An 82% beat.

So here is the question this piece exists to answer. The business inflection is unambiguously real. The policy tailwind is unambiguously real. Unusual Machines now carries a $2.0 billion market cap against $31.9 million of trailing twelve-month revenue. Is that a company the market has finally found, or a company the market has run past?

INVESTMENT THESIS

The core reasons to own this stock

1. The tariff is a structural moat, not a headline. Chinese drone components have undercut American manufacturers on price for a decade. A 100% duty on sensitive categories and 25% on the rest does not make UMAC's motors better — it makes them competitive on cost for the first time. Tariffs take effect within 21 days, with the less-sensitive component levies phasing in after 180 days. This is a repricing of an entire supply chain, and UMAC sits on the right side of it.

2. The revenue curve has genuinely inflected. Four quarters ago this company did $2.1 million in a quarter. Last quarter it did $16.7 million, with enterprise customers representing roughly 95% of sales. That is the difference between a consumer hobby-drone reseller and a defense supply-chain vendor, and the transition happened faster than anyone modeled.

3. Operating losses have nearly closed. Adjusted EBITDA loss narrowed to roughly $400,000 from $1.6 million in Q1. On a $16.7 million revenue base with 34.7% gross margins, the company is within one good quarter of consistent operating profitability — a milestone almost no drone small-cap has reached.

4. The balance sheet removes the usual small-cap failure mode. UMAC holds roughly $229 million in cash plus over $86 million in short-term investments and carries no debt. That is more than $315 million of net liquidity against a $2.0 billion market cap. It can fund capacity expansion for years without returning to the market — which matters enormously given its dilution history.

BUSINESS OVERVIEW

What Unusual Machines actually does

Unusual Machines is an Orlando-based manufacturer of drone components, founded in 2019 and led by CEO Allan Evans. The company's origin is unusual in the literal sense: it was assembled in 2024 out of consumer drone businesses divested by Red Cat Holdings, which wanted to focus purely on military platforms. UMAC inherited Fat Shark, the leading brand in ultra-low-latency FPV video goggles, and Rotor Riot, a direct-to-consumer store for acrobatic FPV drones.

That consumer heritage is now almost irrelevant to the investment case. The pivot came in September 2025 with the acquisition of Rotor Lab, an Australian designer and manufacturer of electric motors and propulsion systems for unmanned aircraft, in a $7.0 million all-stock deal including a $3.0 million earnout. Rotor Lab brought precision-wound motors spanning sub-400W units for small drones up to high-power motors for larger rotary and fixed-wing platforms — and, critically, brought the manufacturing know-how to build them outside China.

Today UMAC sells NDAA-compliant motors, flight controllers, and components to American drone builders. Red Cat — its own corporate parent, in a neat reversal — became the first U.S. drone manufacturer to source motors from UMAC. Dynamic Aerospace Systems and Performance Drone Works followed. The company reports headcount roughly doubling to about 240 employees as it scales production, and management has been explicit that demand exceeds supply into 2027.

INDUSTRY OPPORTUNITY

Washington just made domestic drone parts a policy priority

For a decade, the economics of small drones were settled: China built them, and everyone else assembled them. DJI and its supply chain drove component costs to levels no American manufacturer could match, and the Pentagon's NDAA restrictions created a compliance requirement without creating a domestic industry capable of satisfying it.

The August 13 proclamation attacks that directly. Drones above 25 kilograms maximum takeoff weight, drones with thermal imaging, their docking stations, and their sensitive components now face a 100% ad valorem duty. Smaller drones and remaining components face 25%. Allied nations get preferential treatment — 15% for the EU, Japan, South Korea, Switzerland, Liechtenstein and Taiwan; 10% for the UK — which tells you the policy is aimed squarely at China rather than at imports generally.

The number that matters for UMAC is the phase-in. Sensitive-category tariffs bite in 21 days. Component tariffs on non-sensitive parts arrive in 180 days. That is a compressed window in which every American drone builder must re-source its supply chain, and there are very few domestic motor manufacturers with production already running. UMAC's problem for the next eighteen months is not finding customers. It is building fast enough to serve them.

COMPETITIVE ADVANTAGE

Why Unusual Machines might win

The moat is timing, and timing in manufacturing is harder to replicate than it sounds. UMAC bought Rotor Lab in September 2025 and spent the following year installing capacity, hiring, and qualifying parts with defense customers. A competitor deciding today to enter domestic drone-motor manufacturing faces eighteen to twenty-four months of tooling, qualification, and NDAA compliance before shipping a single unit — by which point UMAC will have locked in the design wins that matter.

Its position in the value chain is also better than the obvious comparison. Red Cat builds finished drones and must win individual platform competitions against Skydio, Anduril, and a dozen well-funded startups; lose the competition, lose the revenue. UMAC sells motors and flight controllers into whoever wins. It is the picks-and-shovels position in a gold rush that Washington just declared strategically essential — and it is why 95% enterprise revenue mix is a more durable number than it looks.Where UMAC is weak is scale and manufacturing maturity. This is a company with roughly 240 employees and a 34.7% gross margin, learning to run high-volume precision manufacturing in real time. Margin at that stage tends to wobble as new staff are trained and new lines are qualified, and management has been candid that supply is the binding constraint. The competitive advantage is real; the operational execution required to convert it is the whole ballgame.

FINANCIAL DEEP DIVE

The numbers: what they mean and what to watch

Metric               Value
-------------------  ---------------------
Last Close (Aug 13)  $27.24
Current Price        $34.06 (+25.0%)
52-Week Range        $7.25 – $34.93
Market Cap           $1.99B
P/E (TTM)            Negative (-156.6)
Price / Book         3.48
Dividend Yield       None
Q2 Revenue           $16.72M (+687% YoY)
Q2 Gross Margin      34.7%
Q2 Net Loss          -$7.78M (-$0.16 EPS)
Q2 Adj. EBITDA       -$0.4M (vs -$1.6M Q1)
Cash + Investments   ~$315M, no debt
TTM Revenue          $31.9M
EV / TTM Revenue     ~53x
Q3 Internal Target   $12M – $14M
Q4 Internal Target   $25M
Shares Outstanding   58.5M (float 45.3M)
Avg. Volume (30d)    3.7M shares

Start with the line most people skipped. Management's internal target for Q3 is $12 million to $14 million — below the $16.7 million just delivered. That is a guided sequential decline, and it is not a warning sign so much as a description of how lumpy enterprise defense orders actually are. But it does mean the next print will look like deceleration, and a stock trading at these multiples has no tolerance for a headline that reads "revenue fell."

The valuation is where honesty is required. At $34.06, UMAC carries a $1.99 billion market cap. Strip out roughly $315 million of cash and investments and you get an enterprise value near $1.68 billion against $31.9 million of trailing revenue — about 53 times sales. Even granting management its $25 million Q4 target and annualizing it to a $100 million run rate, you are paying roughly 17 times forward sales for a company that has never posted a full year of operating profit.

One more thing deserves flagging. The $10.3 million net profit reported in Q1 2026 was almost entirely investment gains on the cash pile, not operating performance. Anyone building a model off that quarter is building it off the treasury department, not the factory floor.

MANAGEMENT TEAM

Who is running this?

Allan Evans has been CEO since the company's formation and has executed the pivot from consumer FPV retailer to defense component supplier with unusual discipline for a micro-cap. The Rotor Lab acquisition was small, all-stock, and strategically precise. Just as instructive, Evans cancelled the planned Aloft acquisition in June 2025, citing strategic misalignment and incurring no fees — small-cap CEOs almost never walk away from announced deals, and the willingness to do so is a genuine signal.

His communication style is also worth noting. On the Q2 call he declined to issue formal guidance while volunteering specific internal targets, and framed the quarter as "continuing to dramatically increase revenue while getting closer to consistent profitability." That is a more measured posture than the promotional register this sector usually operates in.

Then there is the part nobody can ignore. Donald Trump Jr. joined UMAC's advisory board in November 2024 as an investor, and the stock briefly doubled on the announcement. Defense contracts and favorable policy have followed. Investors should be clear-eyed about what this does and does not mean: it does not guarantee a single contract, and it is not a substitute for execution. What it does mean is that UMAC's share price responds to political headlines far more violently than an industrial component maker's fundamentals would ever justify — in both directions.

BULL CASE

5 reasons this stock could dramatically outperform

1. The tariffs are a multi-year demand transfer, not a one-day pop. Every American drone manufacturer must re-source components within 21 to 180 days, and there is no meaningful domestic alternative with production already running. This is not a quarter of upside; it is a structural shift in who gets the orders through 2028.

2. The Q4 target of $25 million implies a $100 million run rate exiting the year. If management hits it, UMAC finishes 2026 having grown revenue more than tenfold in twenty-four months. Growth rates like that in a policy-protected market are what re-rate small caps into mid caps.

3. Operating breakeven is one quarter away. An adjusted EBITDA loss of roughly $400,000 on $16.7 million of revenue means the very next revenue step-up flips this business into profitability. The moment UMAC prints positive operating income, an entirely new class of institutional buyer is permitted to own it.

4. The balance sheet lets them build without asking permission. With over $315 million in cash and investments and no debt, UMAC can expand capacity aggressively into a supply-constrained market — the single highest-return use of capital available to it — without the dilutive raise that would normally cap the story.

5. Institutions are accumulating at a pace that is difficult to reverse. Holdings went from 17,000 shares to 37.1 million in two years, across 240 institutional owners now including BlackRock, State Street, Vanguard, Citadel, and Tudor. That base bought at far lower prices and is not the marginal seller.

BEAR CASE

5 reasons this could disappoint

1. The valuation has already priced the tariff, and then some. Fifty-three times enterprise value to trailing sales is a number reserved for companies with proven operating leverage. Wall Street's average 12-month target of $37.75 implies roughly 11% upside from here, and Maxim's Matthew Galinko — the most engaged analyst on the name — raised his target to $30, which sits below today's price. The sell side is already behind the stock, not ahead of it.

2. Q3 is guided to decline sequentially. Management's internal target of $12–$14 million against Q2's $16.7 million means the November print will show revenue going backwards. On a stock that has run 370% off its low, a headline like that is exactly the kind of thing that starts a 30% drawdown regardless of the long-term story.

3. Dilution is this company's established pattern. The share count has more than tripled over the trailing twelve months. The cash pile exists precisely because management has been willing to sell stock into strength — and with the stock at an all-time high, the temptation to do it again is at a maximum. Every share issued at $34 is cheap capital for the company and expensive dilution for anyone who bought at $34.

4. The policy tailwind is reversible by the same mechanism that created it. Tariffs imposed by proclamation can be modified, exempted, or litigated. Allied-nation carve-outs at 10–15% already leave a competitive lane open, and a future administration — or an adverse trade ruling — could unwind the entire premium embedded in this price.5. Supply constraints cap the upside they cannot cap the expectations. Management says demand exceeds supply into 2027. That is a wonderful problem framed as a bottleneck, but it means UMAC cannot convert the tariff windfall into revenue as fast as the market is now assuming. Meanwhile margins wobble as they train staff and qualify lines — the exact stage where small industrial companies stumble on quality control and supplier delays.

VALUATION

Building the model: Bear, Base, Bull

Scenario  Key Assumption                         Price Target  Probability
--------  -------------------------------------  ------------  -----------
Bear      Q3 misses; dilutive raise at highs     $14.00        35%
Base      Q4 hits $25M; ~17x fwd sales holds     $32.00        45%
Bull      Capacity unlocks; 2027 revenue >$200M  $60.00        20%

Run the probability-weighted math and you get an expected value of roughly $31.30 per share — about 8% below today's $34.06. That is the uncomfortable conclusion this analysis arrives at, and I am not going to dress it up: after a 25% single-session move, the risk-adjusted expected return on UMAC is slightly negative. The business deserves a premium. It does not deserve this premium on the day the news broke. The bull case is genuinely large and genuinely achievable, but a 35% probability of a fall to $14 — a 59% drawdown — is what the combination of stretched multiples, guided sequential decline, and an established dilution habit actually implies.

KEY METRICS TO WATCH

The five signals that tell you if the thesis is working

01 — Q3 revenue against the $12–$14 million internal target. The November 5 print is the whole ballgame. Green flag is anything above $14 million, which would mean the tariff pull-forward is already landing. Yellow flag is a print below $12 million, which would signal the supply constraint is worse than management has framed it.

02 — Adjusted EBITDA crossing into positive territory. The loss narrowed to roughly $400,000 in Q2. Green flag is the first positive quarter, which unlocks institutional mandates that screen on profitability. Yellow flag is the loss widening back above $1 million, which would mean scaling costs are outrunning revenue.

03 — Share count. Shares outstanding sit at 58.5 million. Green flag is a flat count through year-end, proving the $315 million cash position is genuinely sufficient. Yellow flag is any registered offering or ATM activity at these levels — the clearest signal that management views its own stock as expensive.

04 — Gross margin against the 34.7% baseline. Green flag is margin holding or expanding as volume scales, which would confirm the manufacturing process is under control. Yellow flag is a drop below 30%, the classic tell that a small industrial company is buying revenue with unprofitable throughput.

05 — Institutional share count in the Q3 13F cycle. Holdings went from 17,000 shares to 37.1 million in two years. Green flag is continued accumulation into the tariff news. Yellow flag is the first sequential decline — evidence that the smart money that bought at $9 is using $34 as its exit.

MY TAKE

The bottom line

I want to be precise about what I think here, because it would be easy to read this as bearish and it is not.

Unusual Machines is a legitimately good business that got dramatically better this week. Revenue up 687%. Enterprise mix at 95%. Operating losses nearly closed. $315 million of net cash and no debt. A tariff regime that just handed it pricing power against the Chinese supply chain that spent a decade making its business impossible. And an institutional base that went from 17,000 shares to 37.1 million — the exact inverse of the pattern you see in a dying company. Almost everything I want to see in a small-cap industrial story is present.

What is also present is a $2.0 billion market cap on $31.9 million of trailing revenue, a guided sequential revenue decline for the quarter that reports in November, a share count that tripled in a year, and an average analyst target that sits 11% above the current price with the most engaged analyst on the name carrying a target below it. Those two paragraphs describe the same company on the same day.

The resolution is not "good company, bad stock." It is "good company, wrong entry." Buying an industrial manufacturer on the session it gaps 25% on a policy headline is how retail investors reliably transfer money to people who bought it in November at $7.25. The tariffs do not expire. The supply constraint runs into 2027. There will be a Q3 print in November that gives the market something to be disappointed about, and that is when this becomes interesting.

I would rather own UMAC at $22 in December than at $34 today, and I think the odds strongly favor getting that chance.

The business inflection is real, the policy moat is real, and the price already assumes both — you are being asked to pay for the next two years of execution on the day the good news printed.

ANALYST RATING

🟡 HOLD — BUY THE PULLBACK

No new capital at $34. For existing holders, trim into strength and hold the core. For new positions, wait for the November Q3 print and size at 1% to 2% with a target entry in the low $20s — the story is a 2027 story, and you will be paid better for your patience than your urgency.

Sources & Further Reading

This is not financial advice. Always do your own research.

SNAP: The Turnaround Nobody Owns Anymore

Colby Hill — Fri, 14 Aug 2026 14:17:11 GMT

August 14, 2026 | Deep Dive by Colby

OPENING HOOK

Why this story matters right now

Here is a number worth sitting with. Adjusted EBITDA of roughly $250 million — a six-fold increase from the same quarter a year ago.

That is not a rounding error or a one-quarter accounting quirk. That is a company that cut 16% of its workforce in April, stripped more than $500 million out of its annualized cost base, and then watched the savings land exactly where management said they would.

And the stock still trades at $5.30, down 25% over the past twelve months and more than 40% below its 52-week high of $9.28.

Meanwhile, the institutions that are supposed to price this correctly have been walking away. Fintel data shows institutional holders owned just over 1.0 billion shares in July 2025. As of August 2026, that figure is 706 million — a 30% exodus in thirteen months, into the teeth of the best margin quarter in company history. Snap Inc. is the most-mentioned ticker on FinTwit right now, and the reason is simple: the fundamentals and the ownership base are moving in opposite directions, and one of them is wrong.

INVESTMENT THESIS

The core reasons to own this stock

1. The cost structure is finally right-sized, and it is showing up in cash. Snap generated $176 million in operating cash flow and $121 million in free cash flow in Q2 while still posting a GAAP net loss. Free cash flow at a company the market treats as a perpetual cash incinerator is the single most underappreciated fact in this story.

2. The revenue mix is quietly improving. Advertising revenue grew a modest 9% to $1.28 billion. But other revenue — Snapchat+ subscriptions and adjacent products — grew 85% to $316 million. That is the high-margin, recurring, algorithm-proof piece of the business, and it is now roughly one-fifth of total revenue and compounding at a rate the ad business has not seen in five years.

3. Guidance implies acceleration, not stabilization. Management guided Q3 revenue to $1.70–$1.74 billion and adjusted EBITDA to $300–$350 million. The midpoint of that EBITDA range annualizes to roughly $1.3 billion against a $9.5 billion market cap — and the company holds $2.7 billion in cash and marketable securities against that. This is no longer a story stock. It is starting to look like a cash-flow stock priced like a story stock.

4. The setup is asymmetric because expectations are on the floor. Truist and BofA both cut targets to $7 and stayed on the sidelines. Institutions have been net sellers for a year. When the marginal seller is exhausted and the numbers keep improving, the re-rating tends to happen faster than anyone models.

BUSINESS OVERVIEW

What Snap actually does

Snap Inc. is a Santa Monica–based technology company built around Snapchat, the ephemeral visual messaging app co-founded by Evan Spiegel and Bobby Murphy in 2010. The product's original insight — that most communication should disappear — turned into a camera-first social network that now reaches 493 million daily active users, up 5% year over year, and is approaching one billion monthly users.

The business monetizes in three ways. The largest by far is advertising: brands and small businesses buy placements across Stories, Spotlight, and the ad platform, generating $1.28 billion in Q2. The second is Snapchat+, a consumer subscription tier that, alongside other products, drove the $316 million "other revenue" line — the fastest-growing part of the company. The third, and the one that consumes the most capital and the most investor patience, is augmented reality hardware.

That third leg is about to face its verdict. On September 16, Snap holds a dedicated launch event for Specs, its $2,195 consumer AR glasses. This is the culmination of roughly a decade of AR spend, and it is the single largest swing factor in the stock over the next two quarters.

INDUSTRY OPPORTUNITY

Digital advertising consolidated — and then the survivors got cheap

The digital ad market did not shrink. It concentrated. Meta, Google, Amazon, and TikTok absorbed a widening share of incremental budget over the past three years, and every subscale platform — Snap, Pinterest, Reddit, X — was forced to justify its existence to media buyers on performance rather than reach.

Snap lost that argument for most of 2022 through 2024. It is winning a version of it now. World Cup–related brand spending contributed meaningfully to the June quarter, and management specifically called out better momentum with large North American advertisers alongside continued strength among small and medium-sized businesses. Nineteen percent revenue growth is not a market-share loser's number.

The more interesting structural point is what happens to advertising when a platform stops needing advertising to be profitable. At a 58% gross margin with an annualized adjusted EBITDA run rate approaching $1.3 billion, Snap no longer has to win the ad-share war outright. It only has to stop losing it while the subscription and hardware businesses compound underneath.

COMPETITIVE ADVANTAGE

Why Snap might win

The moat is demographic, and it is more durable than the market credits. Snapchat retains near-saturation penetration among users aged 13 to 24 in the United States, Western Europe, and increasingly India — the cohort every advertiser wants and no advertiser can reliably reach elsewhere at scale. Meta owns breadth; Snap owns a specific, high-value slice of attention that Instagram has spent nine years and enormous engineering resources trying to copy without fully dislodging.

The second advantage is AR depth that competitors cannot buy quickly. Meta is further along commercially with its Ray-Ban partnership, and Apple has vastly more capital, but Snap has the largest deployed base of AR lenses and creator tooling in the world — hundreds of thousands of developers who have already built for its stack. Specs at $2,195 is not a mass-market product and will not be judged on unit volume; it is judged on whether Snap can convert a decade of Lens Studio infrastructure into a platform other people want to build on.

Where Snap loses is scale economics. It runs the business with roughly 5,261 employees after the April restructuring against Meta's tens of thousands, and it rents most of its infrastructure — the reason $1.65–$1.70 billion in annual infrastructure cost sits permanently on the income statement. That is the structural tax on being the fourth-largest player in a business with three giants.

FINANCIAL DEEP DIVE

The numbers: what they mean and what to watch

Metric                Value
--------------------  ------------------
Last Close (Aug 13)   $5.30
Current Price         $5.47
52-Week Range         $3.81 – $9.28
Market Cap            $9.54B
P/E (TTM)             Negative (-28.5)
Price / Book          4.54
Dividend Yield        None
Q2 Revenue            $1.599B (+19% YoY)
Q2 Adj. EBITDA        ~$250M (~6x YoY)
Q2 Net Loss           -$164M (-38% YoY)
Q2 Free Cash Flow     $121M
Daily Active Users    493M (+5% YoY)
Q3 Revenue Guide      $1.70B – $1.74B
Q3 Adj. EBITDA Guide  $300M – $350M
Cash & Securities     $2.7B
Avg. Volume (30d)     39.7M shares

The headline tension is between the GAAP loss and the cash generation, and the gap between them is almost entirely stock-based compensation of roughly $1.05 billion for the full year. That is real dilution and it deserves to be penalized — shares outstanding sit at 1.745 billion against a 1.093 billion float. But it is not a cash cost, and the market is currently pricing Snap as though the GAAP loss and the cash reality are the same number. They are not.

The balance sheet has quietly become a strength rather than an overhang. Snap holds $2.7 billion in cash and marketable securities after repaying $47 million of convertible notes due this month, and after retiring more than $2 billion of converts originally due in 2027 and 2028. The refinancing wall that hung over this equity for three years has largely been dismantled.

What to watch next is the trailing revenue base of roughly $6.35 billion against a $9.54 billion market cap — about 1.5 times sales, and materially less on an enterprise basis given the cash. For comparison, that is a fraction of what the market pays for any other platform growing revenue at 19% with expanding margins. The question is not whether Snap is cheap. It is whether the discount is a mispricing or an accurate reflection of terminal risk in North America.

MANAGEMENT TEAM

Who is running this?

Evan Spiegel has been CEO since founding the company in 2010, which makes him one of the longest-tenured founder-CEOs in large-cap technology — and, depending on who you ask, either the reason to own this or the reason to avoid it. He and co-founder Bobby Murphy retain super-voting control, meaning public shareholders have essentially no governance leverage. That structure has drawn open criticism, including calls in the business press this year for Spiegel to step aside after a decade of shareholder underperformance.

The counterargument is that Spiegel just did the thing founders famously refuse to do. In April he cut 1,000 employees — 16% of headcount — closed 300 open roles, took $95–$130 million in charges, and explicitly framed the move around AI-driven efficiency and a "crucible moment" for the company. Whatever one thinks of the framing, the $500 million of annualized savings is showing up on schedule, and Q2's EBITDA beat is the direct evidence.

Notably, the pressure was external. Irenic Capital Management has pushed publicly for cost reductions and now sits among Snap's ten largest institutional holders. When an activist is inside the register and the founder controls the vote, the useful question for investors is whether the cost discipline survives the next product cycle — because nothing about the share structure requires it to.

BULL CASE

5 reasons this stock could dramatically outperform

1. The Q3 EBITDA guide is beatable, and a beat forces a re-rate. Management guided $300–$350 million while noting that the restructuring savings will be "more fully reflected in Q3 and beyond." Q2 already beat consensus by roughly 35%. A print above $350 million in early November would put the annualized run rate near $1.4 billion against a $9.5 billion market cap and make the current multiple indefensible.

2. September 16 is a free option the market has priced at zero. Specs launches at $2,195 with a dedicated event. Nobody models meaningful hardware revenue, no analyst target embeds it, and the sell side has effectively written the AR investment off as sunk cost. If Specs lands as a credible developer platform rather than a curiosity, the entire narrative around Snap's decade of R&D spending inverts in a single session.

3. North American users have stopped falling. NA daily actives held at 92 million sequentially for the first time in several quarters, even though the year-over-year comparison is still down 7%. Sequential stabilization is how every user-metric turnaround begins, and it is the specific data point that unlocks the analysts who have been sidelined on secular-decline grounds.

4. Other revenue is compounding at 85% and nobody is valuing it. The $316 million subscription and other-revenue line, growing 85% year over year, would be worth several billion dollars as a standalone recurring-revenue business. Inside Snap it is buried under an advertising multiple applied to the whole company. That sum-of-the-parts gap closes as the mix keeps shifting.

5. Positioning is a coiled spring. Institutions dumped roughly 300 million shares over thirteen months, short interest and skepticism are elevated, and the consensus stance is Hold with $7 targets. When a stock at $5.47 with $2.7 billion in cash and improving cash flow has no marginal seller left, modest good news produces outsized moves.

BEAR CASE

5 reasons this could disappoint

1. The 19% revenue growth is flattered and does not repeat. World Cup brand spending was explicitly called out as a Q2 contributor, and it does not recur. Strip it out and core advertising grew 9% — below the digital ad market's overall growth rate, which means Snap is still losing share to Meta and TikTok on the primary business line. The Q3 guide implies growth decelerating from here.

2. North America is still structurally shrinking. Ninety-two million NA daily actives is down 7% year over year, and North America is where the revenue per user actually is. One quarter of sequential stability is noise until it is three. If the decline resumes in Q3, the entire "stabilization" thesis collapses and the stock revisits the $3.81 low.

3. Specs could be an expensive embarrassment. A $2,195 consumer AR device in 2026 is a hard sell against Meta's dramatically cheaper Ray-Ban line and whatever Apple ships next. If September 16 produces underwhelming demand, Snap will have spent a decade and billions of dollars to arrive at a niche product — and the market will demand the AR budget be cut, removing the only real growth optionality in the story.4. Dilution eats the equity story from the inside. Roughly $1.05 billion of annual stock-based compensation against a $9.5 billion market cap is more than 10% of the company issued to employees every year. Even if EBITDA doubles, per-share value creation is heavily taxed, and the GAAP net loss of $164 million shows the company is not yet earning its way past that drag.

5. The legal and regulatory overhang is real and getting worse. Snap faces ongoing social media addiction litigation, and proposed smartphone bans in schools directly target the demographic that constitutes its core user base and its entire advertising pitch. An adverse ruling or a wave of state-level restrictions would hit engagement precisely where Snap is least able to absorb it.

VALUATION

Building the model: Bear, Base, Bull

Scenario  Key Assumption                          Price Target  Probability
--------  --------------------------------------  ------------  -----------
Bear      NA users fall again; Specs flops        $3.50         30%
Base      Cost cuts hold; ~$1.3B EBITDA run rate  $7.00         50%
Bull      Specs scales; NA DAU inflects positive  $12.00        20%

Run the probability-weighted math and you get an expected value of roughly $6.95 per share — about 27% above the current $5.47. That is a positive expected return, but it is not the kind of asymmetry that justifies a large position, because the distribution is wide and the bear case is a 36% drawdown to the March low. The honest read is that this is a coin-flip business wrapped in an improving cost structure, and you are being paid a fair but not extraordinary premium to take that bet. The single variable that collapses the range is North American users. If NA daily actives print flat or positive in November, the bear case largely disappears and the base case becomes the floor.

KEY METRICS TO WATCH

The five signals that tell you if the thesis is working

01 — North America daily active users. The number to beat is 92 million. Flat or higher sequentially in the Q3 print is a green flag and the most important single data point in this story. Another sequential decline of 2% or more is a yellow flag that invalidates the stabilization thesis outright.

02 — Adjusted EBITDA against the $300–$350 million guide. A print at or above $350 million confirms the restructuring savings are permanent rather than timing-driven — green flag. Anything below $300 million means the cost cuts are being reinvested faster than management implied, and the cash-flow story loses its anchor.

03 — Core advertising revenue growth ex-World Cup. Q2 advertising grew 9%. Green flag is reacceleration toward low double digits without a sporting-event tailwind. Yellow flag is any print below 7%, which would confirm ongoing share loss to Meta and TikTok in the primary business.

04 — Other revenue and Snapchat+ momentum. The line printed $316 million, up 85%. Green flag is Q3 above $350 million with growth holding above 60%. Yellow flag is growth decelerating below 50%, which would mean the subscription tier has hit its addressable ceiling far earlier than bulls assume.

05 — Institutional share count in the Q3 13F cycle. Holdings sit at 706 million shares, down from 1.01 billion in July 2025. Green flag is the first sequential increase — evidence the exodus has ended and professional money is re-entering. Yellow flag is another leg lower, which would mean the people with the best access to management still want out.

MY TAKE

The bottom line

I have watched a lot of turnarounds get priced wrong in both directions, and Snap is the rare one where the operating evidence is genuinely ahead of the stock. A company that generates $121 million of free cash flow, holds $2.7 billion in cash, grows revenue 19%, and guides EBITDA up 40% sequentially does not normally trade at 1.5 times sales. Something is being mispriced here.

But I want to be honest about what that something is. The market is not confused about Snap's cost cuts — it can read the same release I can. What it is pricing is the possibility that North America is in permanent structural decline, that $1.05 billion in annual stock comp permanently caps per-share value, and that a $2,195 pair of glasses is not a business. Those are not unreasonable fears. They are the fears that produced a 30% institutional exodus over thirteen months, and institutions are not always wrong.

What tips me positive is the sequencing. The cost structure is fixed before the September 16 catalyst, not after. That means Snap gets to find out whether Specs works from a position of cash-flow positivity rather than desperation — and if the answer is no, the company is still generating cash and can cut the AR budget. The downside is bounded by a real balance sheet in a way it has not been since 2021.

This is a position I would take small, sized for a 36% drawdown, and hold through November's print. Not because I am confident. Because the payoff structure is favorable and the crowd has already left.

You are buying the best margin quarter in Snap's history at a price the market last paid when the company was burning cash — and getting a September 16 hardware catalyst thrown in for free.

ANALYST RATING

⚡ SPECULATIVE BUY

Small starter position — 1% to 2% of portfolio maximum, sized so a retest of the $3.81 low is survivable. Time horizon is two quarters, with the November Q3 print and the September 16 Specs launch as the two decision points that determine whether you add or exit.

Sources & Further Reading

This is not financial advice. Always do your own research.