OpenAI lost $38.5 billion in 2025, almost eight times more than in 2024. It will still go public

The well-known analyst Ed Zitron has leaked the audited financial statement of OpenAI for 2024 and 2025. The data is overwhelming and shows how the company, which lost $5 billion in 2024, lost almost eight times more in 2025: $38.5 billion. These colossal amounts do not seem to be an obstacle to the company’s new ambition: going public.

It looks like a big hole… When analyzing the 2025 numbers, it is clear that OpenAI’s operating business is not the real cause of this hole in its accounts. Much of the net loss is due to the transition that the company made from an entity non-profit (non-profit) to a traditional business corporation (for-profit). By doing soUS tax regulations caused a loss of $41.55 billion due to changes in the value of convertible interests and stock options (warrants) that had been agreed with partners and investors.

…but maybe it’s not. The fascinating thing about this situation is that although the data is worrying, it also contains a probable contradiction. OpenAI records this colossal loss of $41.55 billion not because business is bad, but precisely because it is worth much more than before. How do they explain in Financial Timesupon becoming a for-profithad to update all those “accounting promises” at a fair and reasonable value, which generated that notable negative impact on the balance sheet.

The “real” loss. The leaked balance sheet explains that if this “technical” loss from revaluation and some other tax credits is discounted, the pure operating loss from its traditional commercial activity is around a much more acceptable figure and less than $8 billion. It is still a huge amount (in 2024, we repeat they lost 5.08 billion), but it changes the perspective.

Investors still believe in OpenAI. These data may seem terrifying and should make investors flee. They are doing just the opposite because they firmly believe in the future of the company. A few months ago the company raised an absolutely astronomical investment round of 122 billion dollars to reach a valuation of $852 billion.

At the moment the one that wins is Microsoft. The leak also shows who is currently the big financial winner of this AI fever: Microsoft. In 2025, OpenAI paid the Redmond giant a total of $17.2 billion to be able to use the computing capacity of its cloud infrastructure, Azure. The amount Microsoft paid OpenAI for licenses or services was ridiculous by comparison: $303 million.

Source: Sherwood News.

Revenue is growing. For investors, the metric that is sustaining optimism is the speed at which OpenAI has managed to grow its revenue. The company closed 2025 with consolidated revenues of 13.07 billion dollars, almost tripling the 3.7 billion in 2024. But what is really notable is the evolution of its annualized income (Annual Run Rate, ARR), which allows projecting what is expected to be earned at the end of the year. OpenAI started at a pace of $1 billion per quarter, and then accelerate and end up closing with a turnover of more than 2,000 million per month (per month!).

The condemnation of everything free. OpenAI’s commercial strategy, however, may have been its great Achilles heel. The company has paid a high price for wanting to be the free AI of the end user. Keeping hundreds of millions of people querying for free on ChatGPT certainly has a huge impact on operating costs. That contrasts with Anthropic’s approach, which from the beginning focused on business users who pay in much greater proportion. This tactic has allowed the rival company to achieve something unusual: make money with AI. With small print, but they win it.

And the IPO, what? The truth is that OpenAI has already sent the confidential documentation that the Securities and Exchange Commission (SEC) needs to start the IPO process. That doesn’t necessarily mean such an IPO is near, but there’s a problem: Anthropic has taken the exact same step. If the company led by Dario Amodei comes forward in that appearance on Wall Street, it will be another reputational battle won just at the moment when OpenAI is generating the most doubts.

In Xataka | Anthropic is at the most important moment in its history and has a warning: we must lift the AI ​​accelerator