OpenAI Q3 2026 revenue
$10.0B
Quarter ended Sept 30, 2026; first quarterly print disclosed Nov 10, 2026
OpenAI adjusted operating margin
-109%
Q3 2026, improved from -150% in Q2 2026 while the loss still widened to $10.9B
Anthropic Q3 2026 revenue
$16.96B
Quarter ended Sept 30, 2026, per amended Form S-1 filed Nov 2, 2026
Anthropic IPO multiple
20.7x
At the $78 top of the range, fully diluted, on $67.8B annualized Q3 revenue
OpenAI private mark
31.3x
$1.25T pre-money mark against $40.0B of annualized Q3 revenue
Anthropic committed compute spend
$895.4B
Per its S-1; $69.8B falls due within 12 months of Sept 30, 2026
Both labs reported inside the same eight-day window, and the gap between them is the entire argument. OpenAI handed investors a quarterly profit-and-loss statement for the first time on Nov 10: $10.0 billion of revenue in the three months to Sept 30, up about 50% from $6.7 billion, with an adjusted operating loss of $10.9 billion. Anthropic, which prices on Nov 18, reported $16.96 billion of revenue in the same quarter and reversed a $974 million operating profit into a $567 million loss — on 70% more revenue and one nineteenth of the burn.
The two prints
One Week, Two Quarterlies, and a $90 Billion Question Mark
OpenAI's Nov 10 disclosure was a company event, not a leak. Documents reviewed by The Wall Street Journal put quarterly revenue at about $10 billion and the adjusted operating loss at $10.9 billion, widening from $10.1 billion, with the adjusted operating margin improving to -109% from -150%. The next day, CFO Sarah Friar told Yahoo Finance that the September exit run rate was roughly $50 billion, that enterprise revenue doubled quarter on quarter, and that the mix had reached 50/50 between enterprise and consumer on more than 100 million subscribers and 1.2 billion weekly active users. OpenAI's 2025 accounts had leaked in June; this was the first quarter the company itself put in front of investors.
Anthropic had already answered the same quarter in public filings. Its amended Form S-1, filed Nov 2, shows third-quarter revenue of $16.96 billion and a $567 million GAAP operating loss on $17.52 billion of operating expenses. Nine months to Sept 30, 2026 brought $33.24 billion of revenue, a $1.26 billion operating loss, and a $104.0 billion net loss that the filing attributes overwhelmingly to non-cash remeasurement on convertible notes and similar instruments.
| Metric, quarter ended Sept 30, 2026 | OpenAI | Anthropic |
|---|---|---|
| Revenue | ~$10.0B | $16.96B |
| Sequential revenue growth | ~+50% (from $6.7B) | +47% (from $11.56B) |
| Operating result | -$10.9B (adjusted) | -$567M (GAAP) |
| Operating margin | -109% | -3.3% |
| Nine-month operating loss | Not disclosed | -$1.26B |
| Nine-month net loss | Not disclosed | -$104.0B |
| Capitalization | $1.25T pre-money mark, Oct 22, 2026 | $70-$78 IPO range; up to $1.405T fully diluted |
| Multiple on annualized Q3 revenue | 31.3x | 20.7x (fully diluted) |
The core arithmetic
The Multiple Gap Is the Whole Argument
Strip out the narratives and two numbers decide this. Anthropic's IPO range of $70 to $78 values it at about $1.02 trillion on shares outstanding and up to $1.405 trillion fully diluted. OpenAI's mark is $1.25 trillion pre-money, set in talks reported Oct 22, 2026. On the same measure — equity value against revenue — Anthropic prices at 20.7x its annualized third quarter and OpenAI at 31.3x.
OpenAI's Private Mark Carries the Higher Revenue Multiple
Equity value divided by revenue, on the same basis for both labs. Anthropic at the $78 top of the range: $1.405 trillion fully diluted and $1.02 trillion on shares outstanding, against $67.8 billion of annualized Q3 revenue. OpenAI: $1.25 trillion against $40.0 billion of annualized Q3 revenue, or the $50 billion September exit run rate.
Unit: x revenue
Anthropic IPO at $78, shares outstanding
$1.02T / $67.8B annualized Q3 revenue
15
Anthropic IPO at $78, fully diluted
$1.405T / $67.8B annualized Q3 revenue
20.7
OpenAI on $50B exit run rate
$1.25T / $50B September run rate
25
OpenAI on $40B annualized Q3
$1.25T / $40.0B annualized Q3 revenue
31.3
Run Anthropic's own multiple against OpenAI's numbers and the private mark stops looking like a mark. At 20.7x, OpenAI's $40.0 billion of annualized third-quarter revenue is worth about $828 billion, and its $50 billion exit run rate about $1.04 trillion. Against the $1.25 trillion being discussed, the private mark embeds a premium of roughly $215 billion to $423 billion — 17% to 34% — for the lab with the smaller revenue base and the vastly larger loss.
- In 2025, Anthropic's $4.585 billion of revenue was 35% of OpenAI's $13.07 billion. In Q3 2026, Anthropic's quarterly revenue was 170% of OpenAI's. The revenue ranking has inverted inside twelve months.
- OpenAI's third-quarter operating loss of $10.9 billion is 19 times Anthropic's $567 million. Anthropic's revenue is 1.7 times OpenAI's; its loss is a nineteenth as large.
- The $1.405 trillion fully diluted figure implies roughly 18 billion shares against the 13.09 billion outstanding after the offering. Paper to be priced is about 37% of the company, and the OpenAI mark carries the same hidden dilution.
- The gap is not about growth. Anthropic grew 47% sequentially and OpenAI about 50%. On forward revenue the two are within a few points of each other, which is why the print is a repricing event rather than a growth verdict.
- Listed comparables do not rescue the private mark either: Anthropic's IPO multiple sits above the enterprise value to sales multiple the market pays for NVIDIA, Microsoft, Alphabet, Amazon, CoreWeave and Nasdaq.
| Company | Enterprise value | Revenue basis | EV / sales |
|---|---|---|---|
| NVIDIA | $4.20T | TTM to Jul 2026, $303.0B | 13.9x |
| Microsoft | $3.98T | TTM, $346.4B | 11.5x |
| CoreWeave | $94.5B | TTM to Sep 2026, $9.89B | 9.6x |
| Alphabet | $3.66T | TTM to Sep 2026, $471.1B | 7.8x |
| Nasdaq | $70.4B | TTM, $9.1B | 7.7x |
| Amazon | $3.09T | TTM to Sep 2026, $797.5B | 3.9x |
Balance sheet
The Raise Does Not Fix the Commitment — It Fronts One Year of It
Anthropic's balance sheet is the reason this offering is larger than anything in market history, and it is also the reason the equity is hard to read. At Sept 30, 2026 the company held $131.8 billion of cash and short-term investments against $895.4 billion of committed compute and infrastructure spend, of which $69.8 billion comes due within twelve months. Total stockholders' equity was negative $149.9 billion, a deficit created largely by fair-value accounting on convertible notes and similar instruments, not by the operating business.
| Item | Amount | Basis |
|---|---|---|
| Cash, equivalents and short-term investments | $131.8B | Sept 30, 2026 |
| Pro forma cash after offering and placement | $212.7B | Capitalization table |
| Total stockholders' equity (deficit) | $(149.9)B | Sept 30, 2026 |
| Pro forma stockholders' equity | $23.4B | Capitalization table |
| Committed compute and infrastructure spend | $895.4B | Commitments note |
| Of which due before Dec 31, 2030 | $447.5B | Commitments note |
| Due within 12 months of Sept 30, 2026 | $69.8B | Commitments note |
| Net IPO proceeds at the $74 midpoint | ~$84.6B | Use of proceeds |
| Total raise including the $26.0B placement | Up to $90.4B | Reuters, Nov 10, 2026 |
The arithmetic is uncomfortable in a way the headline obscures. Net proceeds of about $84.6 billion cover roughly one year of the $69.8 billion of contracted spend that falls due by Sept 30, 2027, and pro forma cash of $212.7 billion funds about a quarter of the $895.4 billion total commitment. The IPO is not a capital-freedom event. It converts a private mark into a public price one year ahead of the obligations the money is actually for — and along the way it flips a $149.9 billion equity deficit into $23.4 billion of pro forma equity.
Upstream
Where the Cash Actually Goes: $895 Billion Into Silicon
The money does not stay in the model business. Anthropic's $895.4 billion of committed compute and infrastructure spend is contracted future demand for accelerators, and OpenAI's own capacity disclosure shows the same direction of travel: 0.2 gigawatts of compute in 2023, 0.6 in 2024 and 1.9 in 2025, on ARR that went from $2 billion to $6 billion to $20 billion. The chip layer is monetizing these commitments today, while the equity layer that funds them is repriced in public next week.
- NVIDIA grew revenue 106% year over year to $96.2 billion in the quarter ended July 26, 2026, on 75.0% gross margin and $63.7 billion of operating income. Anthropic's commitments are orders of magnitude for that order book, and they are signed, not forecast.
- NVIDIA also sits inside the customer base rather than outside it: OpenAI's own February 2026 announcement of $110 billion in new investment named $30 billion from the chipmaker. It is supplier, lender and holder of the labs' equity risk at the same time.
- The foundry and memory layer sits further upstream still — TSMC and SK Hynix convert accelerator orders into wafers and high-bandwidth memory. Neither shows a valuation that assumes permanent lab pricing power.
- CoreWeave is the clearest listed read on lab demand and the clearest listed warning: third-quarter revenue of $3.66 billion and $207 million of operating income, but a $709 million net loss after $846 million of interest expense, with debt at 1.39 times market capitalization.
- Put side by side, the model layer and the infrastructure layer have swapped places. Anthropic's $1.405 trillion fully diluted value is roughly 15 times CoreWeave's entire $94.5 billion enterprise value, on 4.6 times the revenue.
Downstream
Who Actually Holds the Marks
Three listed companies carry the exposure, and each carries it differently. Microsoft owns roughly a quarter of OpenAI on an as-converted basis and already books the consequences in its income statement: $272 million of net gains on its OpenAI investments in the September 2026 quarter, against $4.1 billion of net losses in the same quarter a year earlier. Alphabet is buying $10.0 billion of Anthropic stock in the concurrent placement — 135,135,135 shares at the assumed $74 price — while Claude runs on Google Cloud. Amazon is the channel: Anthropic's S-1 discloses Claude on AWS, Google Cloud and Microsoft Azure.
- Microsoft does not consolidate OpenAI's $10.9 billion quarterly loss; it marks the stake. That is why a $272 million gain can sit in the same quarter as a record operating loss at the investee, and why Anthropic's public price will eventually show up in Microsoft's non-operating line.
- Alphabet put $10.0 billion into Anthropic at a public price in a stock that trades at 7.8 times enterprise value to trailing sales, against roughly 21 times for Anthropic at the top of its IPO range. It is buying a faster-growing asset with cheaper currency.
- Amazon earns on the commitments themselves. Anthropic's $69.8 billion of spend due within twelve months is contracted cloud demand landing on a business that posted $201.9 billion of third-quarter revenue, up 12%, at a 13.9% operating margin.
- The concentration is the risk on the other side of the trade. Anthropic's filing shows no single customer above 10% of nine-month revenue with the top five near 25%, so no one hyperscaler can switch Anthropic's compute on and off — but no lab can renegotiate its channel economics either.
- Every hyperscaler is now a lender as well as a landlord. Alphabet at $10.0 billion in the placement makes it a holder of Anthropic equity; Amazon and Microsoft hold their own stakes. A bad tape for lab equity becomes a mark-to-market problem for the customers.
Capital markets
The Biggest Equity Raise in History, Underwritten by Two Cheap Stocks
The fee pool is as historical as the deal. Anthropic is offering 825 million shares at $70 to $78 — 801.4 million primary and 23.6 million from selling stockholders — for up to $64.35 billion of gross proceeds, plus a $26.0 billion concurrent placement, with Goldman Sachs and Morgan Stanley as lead underwriters on a 30-day option for a further 123.75 million shares. Pricing is set for Nov 18, with trading on Nasdaq under the symbol ANT on Nov 19.
Anthropic IPO terms
Price range
$70.00-$78.00 per share
Shares offered
825.0M (801.4M primary, 23.6M selling stockholders)
Gross IPO proceeds
Up to $64.35B
Concurrent private placement
$26.0B; Google LLC $10.0B
Pricing and debut
Priced Nov 18, 2026; trades Nov 19, 2026
Ticker and venue
ANT, Nasdaq Global Select Market
Lead underwriters
[Goldman Sachs](gs), [Morgan Stanley](ms)
Value at the top of the range
About $1.0T on shares outstanding; up to $1.405T fully diluted
Shares outstanding after the offering
13.09B across all classes
The banks are not pricing the asset the way the market is. Goldman Sachs trades at 13.1 times trailing earnings and 2.2 times book on trailing revenue of $71.0 billion; Morgan Stanley trades at 15.0 times trailing earnings on trailing revenue of $79.8 billion. The Anthropic deal alone, at $64.35 billion of gross proceeds, is worth more than the entire market capitalization of the exchange that hosts it — Nasdaq stands at $61.4 billion.
- For Goldman Sachs, an equity underwriting of this size is a step-function fee event against a $71.0 billion revenue base, and the firm trades at 13.1 times earnings with a $1,030.67 average analyst target against a $926.81 close.
- For Morgan Stanley, the institutional fee is joined by the retail channel: the deal targets roughly 20% retail allocation, and the firm trades at 15.0 times trailing earnings on $79.8 billion of trailing revenue.
- For Nasdaq, the listing is a rounding error against $9.1 billion of trailing revenue at 7.7 times enterprise value to sales, but the index and franchise consequences of admitting the largest frontier lab are not.
Horizons
What Moves First, and What Breaks the Thesis
Over days to quarters, this is a price-discovery event with six scheduled triggers. The print on Nov 18 sets the first public price for frontier model equity; the aftermarket on Nov 19 sets the level from which every private lab mark is judged; and the next hard data points arrive in Microsoft's December-quarter filing and NVIDIA's fiscal quarter ended Oct 25, 2026.
- Where the print lands inside $70 to $78: the top of the range holds roughly 21 times annualized Q3 revenue, the bottom cuts it to about 19 times. A price below issue on day one is the loudest possible verdict on private marks.
- Whether the $26.0 billion placement expands beyond Google LLC's $10.0 billion, and who else signs up at a public price.
- Whether Alphabet and Amazon disclose the Anthropic positions they now hold, turning private marks into reported assets in the December quarters.
- NVIDIA's fiscal quarter ended Oct 25, 2026, which is the cleanest read on whether the $895.4 billion of lab commitments are being honoured or deferred.
- Microsoft's December-quarter 10-Q, due in late January 2027, which shows whether the $272 million OpenAI mark-to-market gain repeats or reverses.
Over one to three years, the thesis rests on two things going wrong for the private mark and one thing going right for the supply chain. It fails if the public market pays more than 31 times sales for model equity — then OpenAI's $1.25 trillion is validated rather than arbitrated. It also fails if token prices deflate faster than volume grows, which is the classic failure mode of two labs both compounding revenue at roughly 47% to 50% per quarter. It succeeds for the supplier layer as long as the commitments hold: $447.5 billion of Anthropic's $895.4 billion falls due before the end of 2030, and contracts signed today convert into revenue for NVIDIA, TSMC and SK Hynix regardless of how the labs' equity reprints.
Synthesis
Two Prices for the Same Business, and Only One of Them Is Public
OpenAI and Anthropic are about to carry two prices for the same business. Anthropic's will be struck in public at $70 to $78 on Nov 18. OpenAI's was struck privately, at $1.25 trillion, two weeks after the company cut its own revenue guidance by $20 billion. Facts, inference and speculation separate cleanly here. The facts are the two quarterlies, the $895.4 billion of commitments, the negative $149.9 billion equity deficit and the price range. The inference is that Anthropic's print will pull OpenAI's private mark toward public comparables. The speculation is whether the market rewards a model business at 21 times sales or punishes it.
What is not speculative is the direction of travel in the revenue ranking, the size of the commitment book and the fact that a nineteen-fold larger quarterly loss sits on a smaller revenue base at the higher multiple. OpenAI's $1.25 trillion mark asks investors to pay more per dollar of sales for the lab that loses more of it, and the first public price struck for frontier AI equity will put a number on whether that trade was a good one.
For the listed companies, the transmission runs through three channels rather than one: mark-to-market on equity stakes (Microsoft, Alphabet), contracted infrastructure demand (NVIDIA, TSMC, SK Hynix, CoreWeave), and underwriting and listing fees (Goldman Sachs, Morgan Stanley, Nasdaq). The Nov 18 print touches all three. Only one of them is a valuation question; the other two are revenue questions, and they are not waiting on the print to resolve.
Where this event lands on listed stocks
- Books a $272M net gain on its OpenAI investments in the September 2026 quarter, against $4.1B of net losses a year earlier
- Holds roughly a quarter of OpenAI on an as-converted basis, so a lower public AI multiple reverses the mark-to-market gain
- Near term: the December-quarter filing in late January 2027 shows whether the OpenAI fair-value gain persists or unwinds
- Anthropic also sells through Azure, so a public Anthropic listing does not cost Microsoft cloud revenue
- Buys 135.1M Anthropic shares for $10.0B in the concurrent placement, locking a public-price mark on a stake
- Trades at 7.8x enterprise value to trailing sales against roughly 21x for Anthropic at the top of its IPO range
- Claude runs on Google Cloud, so Anthropic's 47% sequential revenue growth is partly Alphabet's cloud revenue
- One to three years: Anthropic's listing gives Alphabet's stake a daily mark its private rounds never had
- Captures contracted AWS demand as Anthropic's $895.4B of committed spend lands, with $69.8B due within 12 months
- Third-quarter revenue of $201.9B grew 12% at a 13.9% operating margin, and Anthropic's channel revenue scales on top of it
- Trades at 3.9x enterprise value to trailing sales, the cheapest sales multiple in the AI stack
- Risk: Anthropic's channel margin, not Amazon's, absorbs the compute cost, and Anthropic's own GAAP loss reappeared in Q3
- Grew revenue 106% year over year to $96.2B in the quarter ended July 26, 2026, on 75.0% gross margin
- Anthropic's $895.4B of committed compute spend is contracted demand for the same GPUs however lab equity reprints
- Near term: the fiscal quarter ended Oct 25, 2026 is the next read on lab compute orders
- Risk to the thesis: a violent repricing of lab equity eventually curtails new commitments rather than existing ones
- Generated $207M of operating income in Q3 2026 but still lost $709M after $846M of interest expense
- Trades at 9.6x enterprise value to trailing sales, roughly half the multiple Anthropic's IPO range implies
- Bull case: lab demand converts backlog into contracted revenue; bear case: repriced customers renegotiate multi-year terms
- Debt at 1.39x market cap leaves the balance sheet exposed to a financing window that can close abruptly
- Earns an underwriting fee on up to $64.35B of gross IPO proceeds, the largest equity offering in market history
- Trailing revenue of $71.0B means the Anthropic deal is roughly the size of a full year of the firm's top line
- Trades at 13.1x trailing earnings and 2.2x book, a fraction of the multiple it is underwriting
- Risk: an Anthropic break below issue would hit both the fee pool and the mark on the position
- Co-leads the largest IPO in market history and is targeting roughly 20% retail allocation through its wealth channel
- Trailing revenue of $79.8B and a 15.0x trailing multiple keep the fee pool material against the earnings base
- The wealth channel monetizes the debut when private AI exposure is scarce for ordinary portfolios
- Risk: a soft aftermarket slows the follow-on pipeline the franchise depends on
