The $20 Billion Disagreement
Ten days after I wrote to watch the filing, the math leaked early: Axios reported OpenAI at roughly $70 billion in annualized revenue; the Financial Times reports the company told investors it is approaching $50 billion. The gap is the story.
The Financial Times reports that OpenAI told investors its annualized revenue is approaching $50 billion. Nine days earlier, Axios put the number at roughly $70 billion. One of those numbers is wrong, and the $20 billion between them is the entire story.
Ten days ago I wrote that the way to read the rumored OpenAI IPO is to watch the filing, not the valuation pop. The filing has not arrived yet. The math leaked early anyway, and it arrived the way leaked math always arrives: as a disagreement.
what the two numbers are
Axios, about nine days ago: OpenAI's annualized revenue is roughly $70 billion. The Financial Times, this week: the company told investors the number is approaching $50 billion. Both outlets are reportedly describing the same metric, annualized revenue, which is the current run-rate projected across a full year. It is the single most important number in the AI business right now. It is the number the IPO will be priced against. And two credible outlets disagree about it by $20 billion.
There are innocent explanations. Different definitions, different dates, different sources inside the company with different incentives. Annualized revenue is a constructed number, not a bank balance. It moves with assumptions. But $20 billion is not a rounding difference. It is nearly 30 percent of the larger figure.
why the gap matters more than the number
Here is the thing I keep coming back to: it does not matter much which number is right. What matters is that the most important number in the industry can be contested by $20 billion in public, nine days apart, and nobody can settle it.
This is what private markets do. Without a filing, every number is somebody's number. The company tells investors one thing, the press reports another, and the truth sits somewhere in the fog between them. Everyone involved has an incentive to shape the figure. Investors want the growth story intact. The company wants the IPO priced well. Journalists want the scoop. The number becomes a negotiation conducted in public.
An S-1 ends the negotiation. One audited number, under penalty of law, in a document everyone reads. That is why I wrote ten days ago that the filing is the business and the pop is theater. The pop is priced on vibes. The filing is priced on arithmetic.
the honest read
The daylight I predicted is arriving early, and it is arriving messily. Before the prospectus, before the roadshow, before a single audited figure, the revenue number is already a public fight. That tells you something about how badly the market wants this number to be big, and how uncertain the reality underneath it still is.
SoftBank completed its $30 billion investment on October 1, which means the biggest checks are still being written against the private-market numbers. The IPO, reportedly targeted for mid-November, will be the moment those numbers meet daylight. Until then, treat every revenue figure as a bid, not a fact.
Watch the filing. It is still the only number that counts. Everything before it is negotiation.