The $70 billion revenue figure that circulated around OpenAI last week was assembled by investors, not published by the company. The Financial Times reports that they built it from information OpenAI had shared with them, in attempts to set the company beside Anthropic, and that OpenAI has now told those investors its run rate is “approaching $50 billion.”
TechCrunch’s Lucas Ropek relayed the FT’s reporting on Thursday. Everything here is therefore secondhand twice over. The FT obtained the lower figure, TechCrunch describes what the FT found, and OpenAI has published no audited accounts to check either against. TechCrunch said it had asked OpenAI for comment.
Annualized revenue, also called a run rate, is a recent period’s takings multiplied out to cover a year. It is a projection from a slice of time, not money the company has banked over twelve months. Companies pick the slice themselves, which is one reason two run rates can look alike and mean different things.
The two labs also do not count the same way. Anthropic folds in the sales its cloud partners make when it reports annualized revenue. OpenAI does not. Put the two numbers side by side and one of them contains revenue that the other has deliberately left out, so the gap between them says little about which business is larger or growing faster. A fair comparison needs one side restated to match the other, and nothing in the reporting says anyone did that.
The brief does not state Anthropic’s own figure, only that $70 billion would have made OpenAI competitive with it.
A figure built by investors, passed to reporters, and published as a company’s revenue is a sell-side estimate wearing a news headline. It moved by $20 billion in about a week, and nothing about OpenAI’s actual business changed in that time. Only the person doing the arithmetic did.
TechCrunch notes that OpenAI raised $122 billion in a single round in March. Leaked 2025 financials showed roughly $13 billion in revenue against considerably higher spending, and the IPO once rumored for this year has slipped to early 2027. Each of those facts is easier to defend when the revenue story looks strong, which gives investors a reason to prefer the larger number.
The same logic applies in both directions. Headline revenue for the two biggest AI labs reaches the public through leaks and investor decks, not through filings, and nobody outside the companies can check how the numbers were put together. Until one of them lists its accounting rules or files public accounts, any run rate for either company is a claim to be sourced before it is compared, and a ranking built on two of them is worth very little.
Based on reporting by Lucas Ropek for TechCrunch, published 8 October 2026, which in turn reports on the Financial Times.
