Leaked OpenAI Financials Show $21 Billion Losses on $13 Billion Revenue as IPO Nears

Independent journalist Ed Zitron obtained and published OpenAI's audited financial statements, which were subsequently reviewed by the Financial Times, Fortune, Ars Technica, and other outlets. The documents surfaced as OpenAI filed confidential S-1 paperwork with the SEC ahead of an expected initial public offering later in 2026. The leak gives investors and analysts an early look at numbers that would normally remain private until the formal IPO prospectus.

The documents cover 2024 and 2025, showing revenue growth from $3.7 billion to $13.07 billion year over year. This rapid expansion reflects the massive adoption of ChatGPT and OpenAI's enterprise API business. The same documents, however, reveal cost structures that are expanding even faster than top-line growth, raising fundamental questions about the economics of large-scale AI deployment.

The timing matters. OpenAI is simultaneously pursuing what could be the largest private capital raise in history, with reports of a $100 billion funding round that would value the company at up to $830 billion. The leaked financials provide the first verified window into whether that valuation bears any relation to underlying profitability.

OpenAI's 2025 revenue of $13.07 billion represents nearly fourfold growth from 2024's $3.7 billion, according to the audited statements reviewed by Fortune and the Financial Times. Cost of revenue alone hit $7.5 billion in 2025, up from $2.65 billion the prior year. When all expenses are tallied, the company posted approximately $21 billion in losses for the full year.

The gap between revenue and spending is widening, not closing. Research and development costs, compute infrastructure, and talent acquisition are consuming capital faster than subscription growth can replenish it. This pattern challenges the common Silicon Valley narrative that scale alone will solve AI economics. The documents suggest that even at $13 billion in revenue, OpenAI remains deeply unprofitable at the operational level.

The自发 The documents suggest that even at $13 billion in revenue, OpenAI remains deeply unprofitable at the operational level. The $21 billion loss figure implies the company is spending roughly $2.60 for every $1.00 it brings in, a ratio that would trigger immediate intervention at a conventional software company.

A substantial portion of OpenAI's spending flows to Microsoft, its primary cloud infrastructure partner. According to the leaked documents, Microsoft received $493.8 million from OpenAI in 2024. In the first three quarters of 2025 alone, that figure jumped to $865.8 million, as reported by ForkLog and StrictlyVC's newsletter.

This revenue-sharing arrangement sits at the heart of OpenAI's cost structure. The company relies on Microsoft's Azure data centers to train and run its models, making it simultaneously dependent on and financially beholden to a major competitor in the AI space. Microsoft itself is building competing products with Copilot and its own AI initiatives, creating a complex frenemy dynamic.

The payments to Microsoft represent more than a simple vendor relationship. They illustrate how the economics of generative AI currently funnel money from application layer companies back to the cloud infrastructure providers that own the underlying compute. For OpenAI, this means every new user and every additional query deepens a cost structure it cannot easily escape without massive capital investment in its own data centers.

The Information reported that OpenAI's internal projections imply losses tripling to $14 billion in 2026. This forecast, combined with the already-leaked 2025 actuals, paints a picture of a company that expects its cash burn to accelerate even as it chieves the largest possible scale.

The disconnect between growth and profitability is not unique to OpenAI, but the magnitude is striking. Few companies in history have sustained $14 billion annual losses while preparing for a public market debut. The bet that OpenAI and its investors are making is that artificial general intelligence, or something close to it, will emerge before the money runs out, transforming the economics of the business entirely.

This timeline pressure explains the urgency behind the reported $100 billion funding round. Nvidia, Microsoft, and Amazon are reportedly in discussions to contribute as much as $60 billion collectively, according to coverage by RDWorldOnline citing The Information and Wall Street Journal reporting. The company needs capital not merely to grow but to survive the gap between current spending and any hypothetical future revenue from transformative AI products.

The leaked financials arrive at a delicate moment for both OpenAI and the AI sector at large. Public market investors have shown growing skepticism toward companies with unsupervised losses, and OpenAI's numbers test the limits of what even growth-oriented funds will tolerate. The confidential S-1 filing on June 8, 2026, noted by Futuresearch, suggests the company is proceeding with IPO plans despite the disclosure risk.

For competitors, the documents validate the enormous capital requirements of training and operating frontier AI models. Companies like Anthropic, xAI, and Google DeepMind face similar cost structures, though none have had their internal financials exposed with this level of detail. The leak may also affect how cloud providers price their services to AI startups, now that the economics are public.

The ultimate question is whether OpenAI can transition from a research organization burning investor capital to a sustainable business before market conditions or investor patience shifts. With $25 billion in annualized recurring revenue now reported but losses still expanding, that transition remains theoretical. The IPO, if it proceeds, will test whether public markets agree that the future value of artificial general intelligence justifies present losses on a scale without precedent in technology history.