§ News
By AI Blog Editor
Nov 12, 2026 · 25 min read
OpenAI's quarterly revenue crossed $10 billion for the first time — and lost $10.9 billion in the same quarter. Anthropic booked $17 billion.
On Nov 11, 2026, three outlets confirmed OpenAI's Q3 2026 revenue crossed $10 billion for the first time, up about 50% from Q2. The same disclosure put the quarter's operating loss at $10.9 billion. In the same quarter, Anthropic booked $17 billion on a $567 million loss.

On Wednesday November 11, 2026, three outlets confirmed the same number: OpenAI's Q3 2026 revenue crossed $10 billion for the first time, up about 50% from Q2. TechCrunch sourced the figure to Bloomberg, The Decoder to the Wall Street Journal and The Information, PYMNTS to the same pair. In the same disclosure, OpenAI's adjusted operating loss grew to $10.9 billion, up from $10.1 billion in Q2. And in the same quarter, Anthropic — the lab OpenAI was supposed to be two years ahead of — booked $17 billion in revenue on a $567 million operating loss, per the IPO prospectus it filed the day before.
Read the three numbers in a row. OpenAI Q3: $10.0B revenue, $10.9B loss. Anthropic Q3: $17.0B revenue, $0.57B loss. OpenAI is the lab with more than a billion weekly users and a $1.25 trillion private valuation, founded six years ahead of its biggest competitor. It is also the lab that, in Q3 2026, booked 59% of its rival's revenue on 19x its operating loss. The story of the AI frontier for most of 2024 was "OpenAI dominates, Anthropic survives." The story for Q3 2026 is a different one.
The three outlets agree on the spine; the scoop is split two ways
TechCrunch attributes the $10 billion figure to Bloomberg, "citing anonymous sources." The Decoder and PYMNTS attribute it to the Wall Street Journal and The Information. PYMNTS adds that the figures come from an OpenAI investor presentation on Tuesday November 10. None of the three lifts the scoop from the other two; the underlying read is one leak into Bloomberg and a separate, broader leak into the WSJ-plus-Information pair, same quarter, same numbers, same day.
The spine the three agree on:
- Q3 2026 revenue: ~$10 billion, up about 50% from Q2's $6.7 billion.
- Q3 adjusted operating loss: $10.9 billion, up from $10.1 billion in Q2.
- Q3 adjusted operating margin: -109%, up from -150% in Q2 (and -97% a year earlier).
- Annualised run rate: $50 billion.
- ChatGPT subscribers: 100 million paying, out of roughly 1.2 billion weekly active users.
- Year-ago Q3 revenue: $3.9 billion.
That is a cleanly-reported quarter: three independent tech-press reads, two independent original scoops, numbers that match at every seam. The headline growth rate is real. The headline loss figure is real. The Q2 base is real. No outlet is laundering a vendor framing — the WSJ's own URL slug carries the phrase "continues to suffer heavy losses as revenue rises to $10 billion," which is as close to a verdict as a trade-press URL writes.
The twist nobody at OpenAI is highlighting is the comparison.
Anthropic's Q3 was $17 billion on a $567 million loss
Anthropic's S-1 amendment, filed with the SEC on Monday November 10 ahead of the IPO roadshow The Loop covered yesterday, put Anthropic's Q3 2026 revenue at $17.0 billion, up 47% from Q2, on an operating loss of $567 million including stock-based compensation. The Decoder's Nov 11 piece is explicit: Anthropic "already overtook OpenAI on revenue in Q2," and OpenAI's new Q3 number "still trails Anthropic." PYMNTS carries the same comparison in the same frame.
The two labs now sit on opposite sides of every profitability ratio that matters.
Q3 2026 | OpenAI | Anthropic |
|---|---|---|
Revenue | $10.0B | $17.0B |
Adjusted operating loss | $10.9B | $0.57B |
Operating margin | -109% | -3% |
Year-ago revenue | $3.9B | ~$2.5B (implied from 47% YoY framing) |
Growth QoQ | +50% | +47% |
Weekly active users | ~1.2B | not disclosed |
Paying subscribers | 100M | not disclosed |
Current valuation | $1.25T (private round) | $1.4T (IPO top of range) |
Money raised toward latest round | $11B of $30B planned | $26B private placement + $57–$64B public offering |
The parallel-growth number (50% vs 47% quarter-over-quarter) is what the comparison looks like at the top of the funnel. The parallel-loss number — a 19x operating-loss gap on 59% of the revenue — is what it looks like at the bottom. OpenAI is burning twenty dollars for every dollar of loss Anthropic burns, while booking fewer dollars of revenue. That is not a growth story and not a scaling story. It is a unit-economics story, and the unit economics are pointing in opposite directions.

The framing OpenAI chose, in Brockman's words and in the WSJ's
Greg Brockman, OpenAI's president, framed the quarter as a turning point for enterprise revenue. Per The Decoder, Brockman named JPMorgan Chase as a major new enterprise customer, said the enterprise business doubled in Q3, and said enterprise now makes up most of the company's revenue. OpenAI's new head of sales, Dali Rajic, was named in the same briefing without a direct quote.
The frame is coherent on its own terms. Enterprise-SaaS economics have been the lab's weakest category through 2024-25, and if Brockman is right that the enterprise book doubled in a single quarter, that is the most structural piece of the story. ChatGPT ads ran at an annualised $1 billion run-rate in August per The Decoder, so the revenue mix is also tilting away from consumer subscriptions — 100 million paying subscribers out of 1.2 billion weekly users is a conversion rate of about 8%, which matches the 8.5% figure the WSJ reported from the same presentation. A conversion rate in the single digits at consumer scale is not unusual, but it is also not a growth lever that can double in a quarter the way enterprise contracts can.
The frame OpenAI isn't choosing is the one the WSJ's URL writes for them. "Continues to suffer heavy losses as revenue rises to $10 billion" is the WSJ reading of the same investor presentation Brockman is reading as a turning point. Both can be true: this was both the first quarter revenue crossed $10 billion and another quarter of operating losses in the ten-figure range, bigger in absolute terms than the one before it. The question is which half of that sentence matters more to the people writing cheques in the current funding round.
The $1.25 trillion round, the $11 billion raised, and the Anthropic gravity well
OpenAI is currently raising fresh capital at a $1.25 trillion valuation, per The Decoder, and has raised less than half of the planned $30 billion so far. The Loop covered the $11 billion raised and the family-office shift on November 1. Eleven days later, the context around that $11 billion has only gotten less flattering: Anthropic opened its IPO roadshow on Nov 10 at a $1.4 trillion top-of-range on $57–$64 billion in public proceeds plus a $26 billion private placement from its chip and cloud suppliers, all priced into a single transaction.
The side-by-side is exact: both labs are telling institutional capital, this week, that they are worth about $1.3 trillion. One is doing it with a Q3 that broke even within 3% of operating margin and a run-rate heading toward $81 billion. The other is doing it with a Q3 that lost $10.9 billion on $10 billion of revenue, with its planned round two-thirds unsubscribed. These are not two prices for the same asset. They are two different offers, priced the same.
That pricing symmetry is the thing to watch into Q1 2027. If the OpenAI round closes anywhere near its $30 billion target at the current valuation, the market will have decided that the loss trajectory is a reinvestment story — the burn is cash going into compute that will throw off multiples of its cost once inference scales. If the round stalls below $15 billion and the valuation drifts to the low trillions, the market will have decided that the loss trajectory is a burn-rate story — the labs with the better unit economics are worth more.
OpenAI will tell investors the first version. The Q3 numbers tell the second.
Three other things sit on the same release
- The adjusted operating margin improved from -150% to -109% quarter-over-quarter. That is a real directional win, the first material margin improvement OpenAI has posted in six quarters. It matters less than it sounds because the absolute loss still grew ($10.1B → $10.9B). Growth outran efficiency gains, which is the pattern a reinvestment story needs — until the day it needs to stop being the pattern.
- Stock-based compensation climbed to $2.6 billion in the quarter. That is money the operating-loss figure excludes but shareholders dilute for. OpenAI's quarterly SBC alone is 4.6x Anthropic's entire Q3 operating loss. The adjusted margins are already ugly; the diluted ones are worse.
- The investor presentation was on Tuesday November 10 — the same day Anthropic opened its IPO roadshow. Both labs picked the same calendar window to talk to institutional capital. OpenAI did it in a closed room with an NDA. Anthropic did it in an S-1 amendment with an SEC stamp on the top page. One of those two documents is going to be bound by securities law when the quarter closes. The other is not.
What this means
The structural narrative has inverted on the two numbers that matter most. Revenue: Anthropic ahead by 70%. Operating loss: OpenAI ahead by 19x the wrong direction. The lab with the user base, the brand and the first-mover lead is also, in Q3 2026, the lab with worse unit economics and a slower-closing round at a lower valuation. For the investors who priced the OpenAI round at $500 billion in late 2024, this is the first quarter where the Anthropic comparison is a problem that doesn't go away by Q1.
The $10 billion milestone is less structural than the $10.9 billion loss. Every press release about Q3 will lead with the revenue figure because it is a round number and it is up and to the right. Every twenty-page memo from a sell-side analyst will lead with the margin figure, because -109% at $10 billion of quarterly revenue is a cash story, not a growth story. Watch which one the next OpenAI fundraising deck leads with.
Enterprise doubled, and that is the one half of the OpenAI story that is actually new. The consumer-subscription conversion rate is pinned at single digits by scale; the ads business is throwing off a billion-dollar annualised run rate but not materially more; the API business is bounded by model competition from Anthropic and Google. Enterprise is where the margin can move. If Brockman is right that the enterprise book doubled in Q3, the question for Q4 is whether it doubles again. If it does, the loss trajectory changes. If it doesn't, the comparison to Anthropic gets worse.
The $1.25 trillion round is the market's second opinion on the quarter. Anthropic's IPO roadshow is set to price around November 19, give or take a day. OpenAI's round has been open since the summer and has brought in $11 billion of the $30 billion target, per The Loop's November 1 story. The two labs are now, functionally, running a public price-discovery process in parallel — one of them in a registered offering and one of them in a private round that has leaked piece-by-piece into the trade press. If both close at anything resembling $1.3 trillion, the market will have accepted that unit economics do not price in at frontier scale. If Anthropic's prices at the top of range and OpenAI's drifts down, the market will have decided they do.
The quarter that OpenAI wanted to be remembered as "the first ten-billion-dollar quarter" is instead the first one where the obvious comparator booked seventy per cent more revenue on five per cent of the operating loss. One of the two numbers writes the headline for next week's earnings coverage. The other writes the headline for the fundraising memo in March.
The Loop will file on the OpenAI round close, Anthropic's first trade, and whichever of the two moves the valuation ratio first.
* * *
Thanks for reading. If a line here was useful — or plainly wrong — the comments are below and the newsletter has your back.
Elsewhere in this issue
3 moreLetters
Arguments, corrections, questions. Anonymous comments allowed; be kind, be specific.