Noted AI critic and bubble watch enthusiast Ed Zitron has been teasing a major scoop for days, and we finally got it: He seems to have gotten his hands on OpenAI’s financial documents and found the company has been burning through cash at an astronomical rate—significantly more than had been reported previously. The figures were verified by the Financial Times, which seemed less alarmed than Zitron.
Zitron found that OpenAI lost around $38.5 billion in 2025, a massive uptick in burn rate from the company’s $5.09 billion in 2024. He also found that the company, while growing its revenue, is still struggling to make good on the promise of being a trillion-dollar company. In 2024, the company generated $3.7 billion in revenue, which jumped to $13.07 billion in 2025. For what it’s worth, OpenAI previously claimed to achieve $20 billion in annualized revenue in 2025.
That difference could probably be chalked up to funky business math, and there’s a lot of that going on in this story. As Zitron notes, OpenAI actually had a net loss of $60.35 billion in 2025, but marked down about $17.87 billion of that to “net loss attributable to noncontrolling members capital.” The company also experienced a $41.55 billion loss related to its conversion from a non-profit to a for-profit entity, a decision that landed it in a very dumb lawsuit with co-founder Elon Musk that amounted to nothing. The Financial Times explained that, under US accounting rules, investors in OpenAI received convertible interest rights when the company made the switch, and those interests went on the company’s ledger as liabilities, hence the loss.
FT also cited a person apparently familiar with OpenAI’s financial situation and concluded that OpenAI’s losses for 2025 were closer to $8 billion once you strip out all of the stuff that should be one-time costs. That’s certainly a friendlier interpretation of the situation, though it doesn’t totally explain away what is unquestionably an astronomical burn rate for OpenAI, which Zitron found forked over $17.2 billion to Microsoft for expenses in 2025, including nearly $10.5 billion specifically set aside for “research and development” expenses, likely related to training new OpenAI models.
One way or another, OpenAI’s losses are growing. It’s revenue is also up, but it’s not clear whether it’s growing at a rate that can support the sheer amount of cash that the company is shoveling into furnaces right now to expand its operation and train its models. Remember, this is a company that, on paper, has pledged (though not actually spent) $1 trillion on data center buildouts. It’s possible for the company to both be growing its revenue extremely quickly and also not generating nearly enough to keep up with its expenditures.
With OpenAI set to go public later this year, we’ll likely get a much clearer look at its financial situation before long. Of course, it might not matter much when it comes to how the market values it. SpaceX isn’t profitable and is banking on putting one million people on Mars as a potential revenue generator, and yet the market is pumping that stock up. We’re currently in a market where theoretical profit is worth more than actual profit. In that world, real losses probably don’t matter.
