OpenAI’s eye-catching $70 billion annualized-revenue figure began circulating in September, presenting the company as roughly comparable with rival Anthropic’s reported run rate. But the number was not a straightforward measure of OpenAI’s own reported sales: it was an effort to “gross up” revenue for an apples-to-apples comparison using Anthropic’s methodology, Axios reported.¹

The distinction lies in cloud-partner transactions. Under the accounting approach described by accounting professor Francine McKenna, Anthropic can record the full value of a sale made through a cloud provider — $100 on a $100 transaction — then list the provider’s share as an expense. OpenAI, by contrast, records only its own share of certain partner sales. Both approaches are GAAP-compliant, but they reflect different judgments about who controls the customer relationship and delivers the product.¹

More recently, OpenAI has reportedly told investors that annualized revenue is “approaching $50 billion,” about $20 billion below the earlier headline figure.² The Financial Times reporting, relayed by TechCrunch, framed the $70 billion number as a comparison exercise initiated by investors rather than a like-for-like disclosure of OpenAI’s usual revenue tally.²

That matters because annualized revenue is a Silicon Valley shorthand, projecting a recent sales period across a full year — useful for fast-growing private companies, but less persuasive than audited actual revenue in public markets. The question is especially sharp for OpenAI as it seeks to justify enormous spending and financing: TechCrunch noted the company raised $122 billion in a March round, while leaked 2025 financials showed roughly $13 billion in revenue alongside much heavier costs.²

The revised figure does not necessarily signal vanishing demand. It does, however, show how dramatically the story changes when identical AI sales are counted through different accounting frames.