Is OpenAI profitable? Not close: Q3 brought $10B and a bigger loss

OpenAI is not profitable yet, though its numbers are moving in a better direction. Here’s the third quarter, the one that ended in September, as reported:

• Revenue: about $10 billion, up nearly 50% from the second quarter and nearly threefold from a year earlier.
• Operating loss: $10.9 billion, up from $10.1 billion in the second quarter.
• Adjusted operating margin: negative 109%, an improvement from negative 150% the quarter before.
• Annual revenue run rate: $50 billion, per Bloomberg.

The details were shared with investors on a call on November 10. None of this is audited public reporting, since OpenAI is still a private company, so treat it as a strong report rather than an official filing.

Both things can be true at once, and the reason is scale. A margin tells you how much a company loses or keeps for every dollar of sales. At negative 150%, OpenAI was losing a dollar and a half for every dollar it earned. At negative 109%, it is losing a little more than a dollar per dollar. That’s real progress.

But because sales grew so fast, the total loss in dollars still crept up. When you sell a lot more of something that costs you more than you charge, the hole gets deeper even as the price gap narrows. Some market watchers read the better margin as lowering the odds of a worst-case cash crunch for OpenAI. The bigger dollar loss shows how far it still has to go.

The single biggest reason OpenAI loses money is the cost of computing power. Training new models and serving a huge user base requires huge fleets of chips in data centers, and OpenAI mostly rents that capacity through partners rather than owning it. Its suppliers include Microsoft, Oracle, AWS, CoreWeave and Google Cloud, with Nvidia chips at the heart of both training and day-to-day answering.

Chief financial officer Sarah Friar has argued that revenue tracks the amount of compute the company can get. In a January article, she said annualized revenue topped $20 billion in 2025 with about 1.9 gigawatts of available capacity, and that more capacity earlier would have meant faster growth. The catch is timing. Those compute deals are signed years ahead, so OpenAI pays for capacity before it knows exactly how much demand will show up. Sometimes capacity runs ahead of usage, and sometimes it’s the other way round, as Friar herself has acknowledged.

The money keeping the lights on

A company losing over $10 billion a quarter needs deep pockets behind it, and OpenAI has them for now. Back in March, investors committed $122 billion, valuing the company at $852 billion after that money came in. It is now finishing a new round at a $1.25 trillion valuation, seeking up to $20 billion, with an initial $11 billion part already done.

It has also pushed back plans to go public until 2027. Until it lists, the public mostly learns its numbers through reports like this one, based on what it tells investors.

The comparison everyone is making is with Anthropic, the maker of Claude. Anthropic reported third-quarter revenue of about $17 billion, and it said its annual run rate climbed to $81 billion by the end of October. Its growth has started to slow, though: about 47% from the second quarter, down from more than 140% the quarter before. One report says Anthropic has reached a positive adjusted operating profit, which would put it on the other side of the line OpenAI hasn’t crossed.

Two caveats keep this from being a clean scoreboard. The firms count revenue differently: Anthropic includes sales made by its cloud partners, while OpenAI reports only its share of some partner sales. And Anthropic’s edge is mostly in selling to businesses, while OpenAI says it has more than a billion users, most of them likely on free or cheap plans. Anthropic is aiming for a $1.4 trillion valuation in a stock market debut expected later in November.

For anyone who uses ChatGPT, the important part is how OpenAI is trying to grow out of the loss. The recent rebound came from a mix of tighter pricing, discounts, better models and a heavier focus on business tools and AI agents, software that carries out tasks for you rather than just chatting. Friar said business growth jumped 70% from the second quarter to the third, and enterprise demand doubled.

That points to a few practical things to watch:

• The free tier is now a business too. OpenAI describes its free plan as supported by advertising and commerce, alongside paid subscriptions for people and workplaces.
• Shopping is moving into the chat. On November 10, OpenAI let businesses claim profiles inside ChatGPT and launched agents for US online stores.
• Prices have gone down, not up, lately. OpenAI has leaned on price cuts to win customers back from rivals. Tighter pricing and discounts were both part of how it grew sales in the third quarter.

If you rely on ChatGPT for work, none of this changes your account today. It does explain why new paid add-ons, business plans and ads keep arriving.

The path is fairly clear on paper: revenue has to keep growing faster than the cost of compute. The third quarter showed that can happen, since sales rose about 50% while the loss grew only slightly. If that pattern holds for several more quarters, the margin keeps closing toward zero.

The risks are just as clear. Cheaper open models, many of them from Chinese labs, put pressure on prices, as do Google, Meta, Amazon and Microsoft. And because compute is booked years ahead, a slowdown in demand would leave OpenAI paying for capacity it can’t fill. The next real test comes with the fourth-quarter numbers, which will show whether the margin keeps improving.