OpenAI's sales jumped 50% to $10 billion last quarter. It still lost more than it took in

OpenAI took in about $10 billion in the third quarter, up nearly 50% from the second quarter. It also reported an adjusted operating loss of about $10.9 billion. By our arithmetic, for every dollar of sales it spent roughly two.

The figures come from investor documents first reported by The Wall Street Journal, and from an interview with chief financial officer Sarah Friar on Tuesday. They are adjusted internal numbers, not audited results. Audited figures would first appear in an IPO prospectus. Friar has told employees OpenAI will be a public company in 2027.

The timing matters. Anthropic, OpenAI's closest rival, is expected to price its IPO around November 18 and start trading as early as November 19. Public investors are about to set a market price for one leading AI lab, and OpenAI's numbers are the closest comparison they have.

Growth came back, and the losses barely moved

The headline for OpenAI's backers is acceleration. Revenue grew only about 18% in the second quarter, to roughly $6.7 billion. Company president Greg Brockman called the first half of the year "pretty rough" at a Journal conference last week.

The rebound came from business customers. Friar said enterprise revenue doubled from the prior quarter. Corporate sales now make up about half of OpenAI's revenue, up from a little over a third three months earlier. She credited coding tools, security products and Dots, its AI assistant for companies. Consumer sales grew a slower 20%. Like the totals, these segment figures are company-reported and unaudited.

The loss side improved in relative terms only. The adjusted operating loss rose to $10.9 billion from $10.1 billion. But because sales grew faster, the adjusted operating margin improved to minus 109% from minus 150%, according to the figures OpenAI gave investors.

Forbes contributor Jon Markman, who is bullish, puts it this way: each extra dollar of sales came with about 24 cents of extra loss. That is the investor case in one ratio. Revenue is outrunning costs, but not yet covering them.

Some caution is warranted on the growth story too. Brockman had said annualized revenue rose 70% in the quarter. Friar later clarified that figure measured the monthly sales pace, not reported quarterly revenue, which grew closer to 50%. OpenAI told investors its pace was about $50 billion a year at the end of September, CNBC reported. It expects to reach $70 billion or more by year end, according to October reports, which would take roughly 40% more growth in a single quarter. Neither run-rate figure is audited.

Why the Anthropic comparison is not like-for-like

On paper, Anthropic looks far ahead. Its IPO filing shows third-quarter revenue of $16.96 billion, up 47% from the second quarter. Its operating loss was $567 million, which it tied to $5 billion of extra computing power bought at short notice. By Forbes' math, OpenAI's reported loss was roughly 19 times larger on about 59% of the sales.

Two caveats cut into that gap. First, OpenAI's figures are adjusted internal numbers, while Anthropic's come from a filing with the Securities and Exchange Commission. They are not measured on the same basis.

Second, the two companies count sales through cloud partners differently. When a customer buys Anthropic's models through a cloud partner, Anthropic books the full sale as revenue and records the partner's cut as a cost. OpenAI books only its own share of certain partner sales. Both approaches comply with US accounting rules, Implicator.ai reported, and accounting professor Francine McKenna notes the choice depends on which company controls the customer relationship. The practical effect is that Anthropic's top line includes money it passes on to partners, while some of OpenAI's does not.

That matters most for the loss comparison. OpenAI's gap with Anthropic on sales is partly an accounting choice. Its gap on losses is too large to explain that way, even allowing for the different bases. Anthropic is close to breaking even at the operating level, and OpenAI is not.

What investors should watch next

The first test is price. OpenAI reportedly raised $11 billion at a $1.25 trillion valuation, below the $1.5 trillion it first sought. Against its unaudited $50 billion annual sales pace, that is about 25 times revenue, by our calculation. Anthropic's IPO could value it at up to about $1.4 trillion, according to reports. Against its third-quarter sales multiplied by four, that is roughly 20 times revenue, again by our rough math. The two multiples use different revenue measures and accounting methods, so treat them as approximate.

On those rough numbers, OpenAI's private backers are paying a higher multiple for a company with far larger losses. If Anthropic's shares trade well, that gap may look like a bargain for OpenAI holders. If they trade poorly, it may look like a problem.

The second test is distribution. Friar said OpenAI's enterprise sales were held back by its Microsoft Azure contract, which limited selling through other clouds. She called broader cloud access a potential "very big tailwind in 2027." That is a forecast, not a result, and it depends on how fast OpenAI signs new partners.

The third test is cost. OpenAI has not published a breakdown of its costs. Its computing commitments include a reported $300 billion, five-year deal with Oracle and a $38 billion deal with Amazon Web Services. Those bills arrive whether or not the sales forecasts do.

For buyers of AI services, the takeaway is simpler. OpenAI and Anthropic are now fighting for the same corporate accounts, and OpenAI's enterprise sales doubled in a quarter. As long as that contest continues, large customers have leverage in negotiations with both labs.