OpenAI has told investors that its annualized revenue reached close to $50 billion at the end of September — roughly $20 billion below the figure that had been circulating from investor documents. The Financial Times reported the number, and markets moved immediately.

The gap is large. In recent months the working figure had been an annualized run rate of around $70 billion, said to come from materials shared during fundraising discussions. The update, reported by the FT, put the number at close to $50 billion.

The same afternoon, the shares of companies treated as proxies for AI demand fell: Oracle dropped more than 5%, AMD about 4.6%, Nvidia nearly 3%, and the Nasdaq had its worst day since July. One revenue figure, reported by one newspaper, was enough to revive doubts about the returns on the entire AI investment cycle — a reminder of how thin the evidence base under these valuations is.

Why the two numbers differ is not settled, and this is where caution is warranted. Some accounts suggest the lower figure reflects a narrower definition — one that excludes certain revenue booked through cloud partners — rather than softer demand. Others read it as a straightforward downgrade. OpenAI does not publish audited quarterly results, and it has not publicly confirmed the FT's number. The gap between $50 billion and $70 billion could therefore be partly an accounting question rather than a demand question. Both matter to a valuation, but they are different problems.

Separately, several outlets report that the company is pushing back its plans to go public. That, too, has not been officially confirmed.

Why it matters: the data center and chip industries have built their capital plans on the assumption that AI revenue keeps climbing steeply. If the reference number from the most important vendor is revised down by a third, that changes the arithmetic behind depreciation schedules, leases and credit lines — not just the mood on one trading day.

What to watch: whether OpenAI publishes audited figures or a clear definition of its revenue run rate; how its next funding round is priced; and whether other AI infrastructure names follow the market's lead.

Sources