Leaked financial docs show OpenAI is losing billions of dollars a year
Ars Technica
Leaked SEC docs show OpenAI's revenue hit $13 billion in 2025, but it still lost nearly $21 billion. That's a lot of red ink for a company promising Wall Street it'll turn a profit by 2030.
Based on reporting by Ars Technica — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
Ed Zitron got his hands on audited financials that OpenAI would probably rather keep quiet while it preps for an eventual IPO, and the numbers tell a story of a company scaling revenue impressively while burning cash even faster. Revenue jumped from $3.7 billion in 2024 to $13.07 billion in 2025, and the Financial Times says monthly run rates hit nearly $2 billion by December, meaning growth didn't just happen — it accelerated through the year.
But spend a minute on the expense side and the picture gets uglier. R&D costs alone ballooned from $7.81 billion to $19.18 billion, more than the company's entire 2025 revenue. A chunk of that, $10.59 billion, went straight to Microsoft. Inference costs — what OpenAI calls cost of revenue — nearly tripled, from $2.65 billion to $7.5 billion, as more people than ever started poking at ChatGPT and its API. Sales and marketing spending grew fivefold too, from $1.11 billion to $5.73 billion, which is a strange thing to see from a product that supposedly sells itself on hype alone.
Add it all up and OpenAI's operating loss went from $8.78 billion in 2024 to $20.92 billion in 2025. That's not a rounding error. And it's happening at a company that has told investors, repeatedly, that profitability is coming by 2030. There is a silver lining buried in the math: losses as a share of revenue actually shrank, from 237 percent to 160 percent, which suggests the unit economics are improving even if the absolute dollar figures look terrifying.
Still, 160 percent of revenue lost is not a business anyone would call healthy by traditional standards. OpenAI is betting that scale eventually flips the equation, the same bet every capital-intensive AI lab is making right now. Whether investors buying into an IPO will have the patience for five more years of numbers like this is the real question these documents raise.
My take — AI-written commentary, not fact-checked reporting
None of this surprises me — building and running frontier models costs an obscene amount of money, and OpenAI's Microsoft dependency alone should worry anyone thinking about buying into an eventual IPO. The percentage-of-revenue improvement is the only number that matters here, and it's still nowhere close to sustainable; I'd rather see a profitable open-weight lab with modest revenue than another Silicon Valley cash furnace dressed up as inevitability.
Read more about this at: Ars Technica