The $25B that isn't $25B

At the end of February, OpenAI's revenue was running at a pace that, projected across a full year, comes to about $25 billion. That number is now the headline figure, and it is the one to read carefully. It is a run-rate: a recent month annualized, not money in the door. What OpenAI actually booked across all of 2025 was around $13.1 billion. Its own forecast for 2026 is a loss of roughly $14 billion — meaning the company expects to lose more this year than it earned in the whole of last year.

$600 billion in compute committed through 2030 — about 24 years of revenue at the current $25 billion run-rate.

All of this is happening as it prepares to go public. A private round that closed in March put about $122 billion into the company at a valuation north of $840 billion, the largest such round on record, and a confidential filing for an IPO targeting up to $1 trillion was reportedly lodged in May, with a listing possible as early as this autumn.

A trillion-dollar listing forces the question of whether public markets can price the AI build-out at all. OpenAI has committed to something like $600 billion in compute through 2030 — roughly 46 times what it booked last year ($13.1 billion) — and the bet is that revenue compounds fast enough to outrun the spend before the spend outruns patience. An IPO of this size is the first real test of that arithmetic in front of investors who can sell the next day.

One side-light: Nvidia's Jensen Huang has put $30 billion into OpenAI and $10 billion into Anthropic, and framed both as his last pre-IPO moves on the logic that going public closes the window to invest. TechCrunch's Connie Loizos doubts that explanation, noting it is hard to square with how late-stage private investing works — and reading the retreat (Nvidia's OpenAI stake was trimmed from a mooted $100 billion) as an exit from a situation that has grown complicated fast.