OpenAI has posted a $10 billion quarter. It has also posted a quarter in which it lost more than that, on the measure it chose to show investors.

The figures were reported on Tuesday, November 10, by The Wall Street Journal and Reuters — the Journal from investor documents it reviewed — with The Information reporting the same day. Revenue for the three months to September 30 came to about $10 billion, up roughly 50% from about $6.7 billion in the second quarter, the Q2 figure implied by the reported growth rate. That is a sharp change of pace: The Decoder, reporting the Journal and The Information, put Q2 growth at 17 to 18% and year-ago revenue at $3.9 billion.

The loss grew. The margin improved. Both are true.

Adjusted operating loss widened to $10.9 billion, from $10.1 billion in Q2. The loss is bigger than the quarter's entire revenue. At the same time, the adjusted operating margin improved, from −150% in Q2 to −109%, because revenue grew much faster than the loss did. Both are true, and quoting only one gives a distorted quarter. The business became less unprofitable per dollar of sales and more unprofitable in total dollars, in the same quarter.

There is a second qualification, and it is not small. The $10.9 billion is an adjusted figure, and the adjustment excludes stock-based compensation expense of about $2.6 billion reported for the quarter. Stock pay is a real cost, paid in ownership instead of cash. A loss figure that leaves it out understates what the quarter cost under ordinary accounting, and any comparison built on the adjusted number inherits that gap.

The run-rate line needs its basis

Dataconomy reports that CFO Sarah Friar described a blockbuster quarter in interviews and put the annualized revenue run rate at about $50 billion. Treat that number carefully. Four quarters at the Q3 level would be $40 billion, not $50 billion, and the reporting reviewed does not specify which period or definition Friar's figure rests on. Until a basis is published, the figure on the record is the quarter itself: about $10 billion, as reported.

The mix shifted underneath the total. The Decoder reports that President Greg Brockman said enterprise revenue doubled in Q3 and is now the bulk of revenue, helped by demand for its coding tools. Indian Television's account of Friar's CNBC interview describes the split between enterprise and consumer as approximately equal. A source can be the largest while remaining approximately half, so the two descriptions are not necessarily inconsistent — but they do not provide an exact reconciled mix for anyone modeling the business. On the consumer side, The Decoder reported ChatGPT at 1.2 billion weekly users and 100 million paying subscribers, with 22 million new sign-ups reported.

Anthropic is the awkward comparison

The reports landed while Anthropic is on its IPO roadshow, and the side-by-side is unflattering to OpenAI. The Decoder reports Anthropic's third-quarter revenue at about $17 billion, up 47%, with an operating loss of $567 million. That Anthropic loss includes stock-based compensation; OpenAI's reported adjusted loss excludes $2.6 billion of it. The measures are not directly comparable without their full reconciliations, and the identified stock-pay difference is one reason why.

The sourcing gap is separate from the numbers. Anthropic's figures are available in a public IPO filing. OpenAI's reach readers through reporting on private documents: the company is private, it has published no filing for this quarter, and it did not immediately respond to Reuters' request for comment. That does not make its figures false. It means readers hold them at one more remove than the rival figures they are now compared against.

What to watch

OpenAI is raising capital at a reported $1.25 trillion valuation, with about $11 billion secured at first close, as ALFW reported on November 6. A private raise can proceed on investor documents alone. If OpenAI follows Anthropic toward a public listing, figures like these would eventually sit in a filing, with the adjustments itemized. The distance between the investor-document version of a quarter and a filed version is the thing to measure if that day comes.

Sources

  1. [1] Reuters (syndicated, November 10, 2026): OpenAI's quarterly revenue rose about 50% to $10 billion in Q3; operating losses widened to $10.9 billion from $10.1 billion; adjusted operating margin improved to −109% from −150%, per documents viewed by The Wall Street Journal; OpenAI did not immediately respond to a request for commentRead source
  2. [2] The Decoder (November 11, 2026, 12:12 UTC), reporting the WSJ and The Information: Q2 growth of 17–18%, year-ago revenue of $3.9 billion, $2.6 billion stock-based compensation excluded from the adjusted loss, year-ago margin of −97%, Brockman on enterprise doubling to become the bulk of revenue, 1.2 billion weekly users and 100 million paying subscribers, Anthropic's $17 billion Q3 revenue and $567 million loss including stock compensationRead source
  3. [3] Dataconomy (November 11, 2026): Friar's approximately $50 billion annualized run-rate statement and the clarification that her earlier 70% growth comment referred to run-rate growthRead source
  4. [4] Indian Television (November 11, 2026), summarizing the WSJ report and Friar's CNBC interview: revenue split described as approximately equally enterprise and consumer; Anthropic's Q3 revenue near $17 billion with a $567 million operating lossRead source
  5. [5] Invezz (November 11, 2026): Friar's "blockbuster quarter" description, Q2 revenue of approximately $6.7 billion, and the note that the adjusted figures are not net losses or cash consumedRead source
  6. [6] Unite.AI (November 10–11, 2026), summarizing Reuters: the $10 billion revenue and $10.9 billion loss figures, the margin improvement, and OpenAI's March 31, 2026 funding announcement ($122 billion committed at an $852 billion post-money valuation) as backgroundRead source