Analysis
The Financial Times reported this week that OpenAI's annualized revenue, when compared apples-to-apples with Anthropic's, runs roughly $20 billion below the $70 billion figure that's been circulating -- not because OpenAI lied about its growth, but because the two companies have been measuring revenue differently, as Newcomer laid out this week. OpenAI counts net revenue; Anthropic counts gross revenue, including the cut that cloud resellers like AWS, Azure and Google Cloud take when they sell Claude access. Bloomberg later clarified OpenAI still expects $70 billion in annualized revenue by year-end, on a net basis -- which Pulse covered as OpenAI's own year-end target days before the accounting gap surfaced.
Our Read
Until either company files an S-1, every headline comparison between OpenAI's and Anthropic's revenue is a communications choice, not a financial statement. Gross-versus-net isn't a minor footnote in businesses this size -- it's the difference between a reseller's take rate mattering or not, and reseller take rates on cloud-distributed AI products are not small. Dan Primack's framing is the right one: "This OpenAI/FT stuff is the sort of thing that would be eliminated if both it and Anthropic go public. Apples/apples GAAP accounting for all to see." Every later-stage investor marking up secondary shares off a blended ARR number right now is underwriting a metric that neither lab defines the same way, and that neither has to defend under audit.
“## Our Read Until either company files an S-1, every headline comparison between OpenAI's and Anthropic's revenue is a communications choice, not a financial statement.”
That doesn't mean the underlying growth is fake. Both companies are plausibly growing revenue at a pace that would have seemed absurd two years ago, and the market reaction -- AI stocks wobbling on the FT story -- treated an accounting-convention gap as if it were a growth miss, which it wasn't.
Room for disagreement: Sophisticated growth-stage investors already apply their own haircuts to self-reported ARR and know gross and net figures diverge; the FT story is only shocking to anyone treating a single blended number as audited fact in the first place. The underlying growth trend at both labs is directionally real, and the gap here is about disclosure discipline rather than fraud -- it resolves the moment either company files to go public, which both are reportedly moving toward.
The number to track going forward isn't $70 billion. It's whether either company starts publishing a gross/net split voluntarily, before a regulator or an S-1 forces the issue.
