OpenAI's audited 2025 financials, obtained by journalist Ed Zitron and verified by the Financial Times, show a company with a serious gap between what it earns and what it spends. Revenue reached $13 billion last year, more than tripling from $3.7 billion in 2024. Costs reached $34 billion. The net loss was $38.5 billion.
The headline loss number needs a small asterisk. About $41.5 billion of the total was a one-time accounting charge tied to OpenAI converting from a non-profit into a for-profit company. Stripping that out, the operating loss was roughly $21 billion. That is still more than $1.50 spent for every $1.00 earned.
The Microsoft relationship is central to understanding these numbers. OpenAI paid Microsoft $17.2 billion in 2025 across server costs, research and development charges, and other fees. Microsoft paid OpenAI $303 million in return. OpenAI runs almost entirely on Microsoft's cloud, and that dependence now shows up clearly in the accounts. OpenAI has itself flagged this reliance as a significant risk in documents shared with investors.
Research and development costs alone, at $19.2 billion, exceeded OpenAI's total revenue for the year. Sales and marketing spending jumped from $1.1 billion in 2024 to $5.7 billion in 2025, a 418% increase in one year. The company is not just building; it is also spending heavily to pull in customers.
At the same time, OpenAI has confidentially filed for a stock market listing, with Goldman Sachs, Morgan Stanley, and JPMorgan advising on the process. The company is targeting a listing as early as the fourth quarter of 2026, at a valuation that analysts expect to exceed $1 trillion. Anthropic filed for its own listing a week earlier at a $965 billion valuation, so the two companies are racing each other to public markets.
OpenAI's own internal projections show it will not turn a profit until 2029, and that is the optimistic scenario. Some independent analysts now place the cash-flow-positive date at 2030, with cumulative losses potentially reaching $143 billion before the company gets there. The argument for the stock is that revenue is growing fast: monthly revenue hit $2 billion by the end of 2025, and some projections put annual recurring revenue above $20 billion entering 2026.
The core question for anyone watching this is whether AI costs fall fast enough. Every time someone uses ChatGPT, OpenAI pays for the computing power to generate that answer. Those costs have roughly doubled year on year. If costs keep rising in step with users, revenue growth does not help much. OpenAI's bet is that computing costs will eventually fall and that its 2029 revenue target, around $100 billion, will be large enough to cover everything. That is a big bet, and it requires years of continued investment before anyone finds out if it is right.
For business operators who already use ChatGPT or similar tools, the practical read is this: the AI tools you pay for today are being sold below cost. That means pricing will likely rise, competition will intensify as rivals try to grab market share, and the market will consolidate around whoever can build a sustainable cost structure first. The companies that survive will be those that reach massive scale. Everyone else will be acquired or shut down.