OpenAI recorded a record loss of $38.5 billion: AI race costs have skyrocketed
OpenAI's financial results for 2025 shock even seasoned investors. According to audited documents I analyzed, the company's net loss reached an astronomical $38.5 billion. For comparison, in 2024 this figure was only $5.09 billion, meaning losses grew nearly eightfold. The numbers speak for themselves, and behind them lies the harsh reality of the artificial intelligence arms race.
Revenue grows, but expenses devour everything
On one hand, OpenAI's revenue is indeed impressive: it jumped from $3.7 billion in 2024 to $13.07 billion in 2025. However, operating expenses grew even more rapidly — from $12.48 billion to $34 billion. The main cost item was research and development — $19.18 billion. Cost of revenue consumed another $7.5 billion, while sales and marketing accounted for $5.73 billion. As a result, operating loss increased from $8.78 billion to $20.92 billion.
A key point I want to highlight: 2025 was a transitional year for OpenAI from a non-profit structure to a commercial one. This led to a one-time loss of $41.55 billion due to the revaluation of convertible instruments and warrants. Excluding these "paper" items, the net loss would have been even higher — $60.35 billion. The company managed to "soften" it to $38.53 billion by excluding $17.87 billion related to non-controlling interests.
Microsoft received $17.2 billion from OpenAI
The dynamics of settlements with key partner Microsoft are also intriguing. In 2025, OpenAI paid the tech giant $10.59 billion under the "research and development" line item (likely for renting computing power to train models). Another $6.047 billion went to cost of revenue, $527 million to sales and marketing, and $42 million to administrative needs. Total: $17.2 billion.
At year-end, OpenAI's liabilities to Microsoft amounted to $3.64 billion, plus short-term and long-term debts. The company's assets exceeded $50 billion, with nearly half in cash. In 2025, SoftBank paid OpenAI $867 million, and Microsoft paid $303 million.
My analysis: These figures expose a fundamental problem for the entire AI industry. Creating cutting-edge models requires massive investments in data centers and chips, which currently far outpace revenue. OpenAI has already contracted about $600 billion in future infrastructure expenses. The main source of losses is not scientists' salaries, but the cost of servicing billions of daily queries on expensive chips. The market clearly underestimates the scale of capital expenditures needed to monetize AI in the long term.