OpenAI recently told investors its annualized revenue was nearing $50 billion at the end of September, The Information reports. That’s about $20 billion below the roughly $68 to $70 billion figure that circulated in reports only weeks ago, including The Information’s own. The Financial Times reported the $50 billion number as well, and CNBC and Reuters later confirmed it.
The company is still growing fast. But this is the first time the headline number has moved down instead of up, and the market noticed.
⚡ Quick Summary
- The new number: OpenAI’s annualized revenue was close to $50 billion by the end of September.
- The old number: Recent reports put it around $68 to $70 billion.
- The gap: About $20 billion. Most of it seems to come from how revenue gets counted, not from a sudden drop in sales.
- Growth is still strong: OpenAI reported 77% run-rate growth in Q3 and 107% run-rate growth in its enterprise business, according to a person familiar with the investor communications.
- Market reaction: CNBC reports that Nvidia, Oracle, CoreWeave and other AI infrastructure stocks fell on the news.
🧮 Why the Number Shrank
The difference mostly comes down to accounting choices. The higher figure reportedly included gross revenue from sales through OpenAI’s partners. The point was to line OpenAI up against Anthropic, which counts revenue from sales through cloud partners like AWS and Google Cloud.
OpenAI’s own internal method leaves that partner revenue out. Use OpenAI’s method and you get about $50 billion. Use Anthropic’s and you get something closer to $70 billion.
That matters because the two methods aren’t the same. A Reuters analysis found that Anthropic pays cloud partners about 16% of every dollar it earns through them, and that channel made up about half of Anthropic’s revenue last year. So depending on the method, one company’s “revenue” can include money that never really stays with it.
According to The Information, OpenAI didn’t dispute the earlier, higher figures when they first came out.
📉 Why This Matters
What stands out here is how much weight one fuzzy number carries. Annualized run rate usually means taking one month’s revenue and multiplying it by 12. It’s a useful pulse check for fast-growing startups. It isn’t audited revenue, and companies have a lot of room to decide what goes into it.
That’s a real problem when the number is supporting:
- Huge compute commitments. OpenAI has signed multiyear infrastructure deals with Oracle, CoreWeave, Nvidia and others worth hundreds of billions of dollars.
- Private valuations. Investors price OpenAI and Anthropic largely off revenue multiples, so a $20 billion swing in the denominator changes the math.
- Public AI stocks. OpenAI’s revenue gets treated as a stand-in for overall AI demand. When it wobbles, chip and cloud stocks wobble too.
The selloff shows how closely Wall Street ties infrastructure spending to one company’s ability to make money. If OpenAI’s real revenue base is about 30% smaller than people thought, the gap between its spending plans and its income gets wider.
🔍 The Anthropic Angle
This also shows how heated the OpenAI and Anthropic rivalry has become. Both companies want the “fastest-growing” label, and both use revenue figures to win enterprise customers, hires and investors.
When the two can’t agree on what counts as revenue, side-by-side comparisons mean very little. Expect more scrutiny of how AI labs report their numbers, especially as IPO talk picks up for both.
🧭 What to Watch Next
- Standard disclosures. Investors will push for one consistent revenue definition across AI labs. Any IPO filing would require GAAP numbers, and those would end the debate fast.
- Infrastructure partners. Watch Oracle and CoreWeave earnings calls for questions about how exposed they are to OpenAI.
- Enterprise momentum. OpenAI’s 107% enterprise growth is the strongest part of this story. If it holds up, the headline correction matters less.
- Anthropic’s response. Don’t be surprised if Anthropic gets pressed to break out net revenue after partner payouts.
For practitioners and buyers, demand for AI tools hasn’t changed. The open question is whether the money behind the AI buildout is as solid as the headlines made it look. You can find more details in The Information’s original report.