I opened the PDF at 2 a.m. and the headline hit like a cold room: losses far larger than anyone had advertised. You scroll and the math keeps getting stranger. I felt the story pull tight—growth on one axis, a deeper hole on the other.
I’ve been tracking tech finances long enough to spot theater from trend. You want facts and a sense of whether this is a temporary wobble or structural trouble. I’ll walk you through the filings, the awkward accounting, and what it means if markets decide that promises beat profit.
I found a table that showed a $38.5 billion loss and had to read it twice
OpenAI’s leaked documents—pushed into public view by Ed Zitron and confirmed by the Financial Times—report a $38.5 billion loss in 2025 (≈ €35.4 billion). That dwarfs the $5.09 billion (≈ €4.7 billion) loss in 2024 and makes the scale of spending impossible to ignore.
The company did grow revenue from $3.7 billion (≈ €3.4 billion) in 2024 to $13.07 billion (≈ €12.0 billion) in 2025. Yet even that jump doesn’t cover the headline losses. The filings also show a net loss of $60.35 billion (≈ €55.5 billion) before accounting adjustments—numbers that read like a software-era horror story.
How much money is OpenAI losing?
The short answer: it depends on which line items you accept. The filings show a $38.5 billion operating loss, but accounting entries tied to the company’s conversion from non-profit to for-profit pushed the headline to roughly $60.35 billion (≈ €55.5 billion). Some of that $60 billion is the bookkeeping result of convertible interest rights being recorded as liabilities.
The ledger showed $17.2 billion paid to Microsoft in one year
I traced the vendor lines and kept stumbling on Microsoft. The leak lists $17.2 billion paid to Microsoft in 2025 (≈ €15.8 billion), including about $10.5 billion (≈ €9.7 billion) tagged as research and development—almost certainly model-training compute and cloud services.
That’s where the business story becomes tactical: OpenAI is pouring money into compute and data-center scale to keep model development moving. Think of the spend like a bonfire: it’s spectacular, it lights everything up, and it consumes wood fast.
Is OpenAI profitable?
Not by GAAP measures. If you strip out one-time conversion-related accounting hits, an FT source suggests a more modest $8 billion loss for 2025 (≈ €7.4 billion). That’s better, but still a meaningful burn against $13.07 billion (≈ €12.0 billion) of revenue—growth, yes; margin, not yet.
I read the notes about the non-profit-to-for-profit switch and saw a paperwork grenade
The conversion created a $41.55 billion accounting loss (≈ €38.3 billion) tied to convertible interests granted to early investors. The change also spawned a lawsuit from Elon Musk, which fizzled but left a paper trail that inflates headline losses under U.S. accounting rules.
The filings show $17.87 billion (≈ €16.5 billion) of the net loss attributed to “noncontrolling members capital,” which is another way the ledger makes the situation look worse than the ongoing burn alone.
Will OpenAI go public?
Yes—or at least it’s expected to later this year—so a full prospectus will give markets clearer numbers. But markets don’t always value profits. Investors have chased stories: SpaceX isn’t profitable yet still commands massive value based on future scenarios (and that’s a useful comparison when you’re pricing a company whose promise is enormous).
OpenAI’s public filing will force clarity. It will also force a market judgment: do you value scale and promise over present losses?
I compared the rhetoric to what the cash flow actually buys
Management has painted a future in which massive compute investments and data-center expansion justify today’s burn. On paper, the company has discussed pledging $1 trillion to build out infrastructure (≈ €920 billion), though that’s not the same as spent capital.
There’s a plausible path where rapid revenue growth—triple-digit percentage increases, strong enterprise deals, and Microsoft’s partnership—catches up to capex and operating cost. There’s also a path where spending outpaces returns for years. The current numbers let both narratives exist in the market like competing headlines.
So where does that leave you? If you’re watching for a technology winner, OpenAI’s traction is real: millions of users, enterprise contracts, and an enormous technical lead. If you care about cash flow and clean profitability, the filings are a warning light. The market will decide which lens it prefers—story or spreadsheet—but which would you trust when both are on fire?