OpenAI is burning through cash at an unprecedented scale, with newly leaked financial documents revealing a staggering $38.5 billion loss in 2025. The figures, initially uncovered by tech critic Ed Zitron and subsequently verified by the Financial Times, highlight a massive acceleration in the company's burn rate compared to its $5.09 billion loss in 2024. While the AI giant is undeniably growing its top line - jumping from $3.7 billion in revenue in 2024 to $13.07 billion in 2025 - it is struggling to outpace the astronomical costs of training next-generation models. This actual revenue also falls short of the company's previous claims of achieving a $20 billion annualized run rate.
The raw financial data paints an even more complex picture of the company's internal accounting. Zitron's analysis indicates that OpenAI actually recorded a net loss of $60.35 billion in 2025. However, the company marked down approximately $17.87 billion of that figure as "net loss attributable to noncontrolling members capital." Furthermore, OpenAI absorbed a massive $41.55 billion loss directly tied to its controversial restructuring from a non-profit to a for-profit entity.
According to the Financial Times, US accounting rules required that investors receive convertible interest rights during this corporate switch. These interests were recorded on the company’s ledger as liabilities, artificially inflating the deficit. When stripping away these one-time restructuring costs, sources familiar with the financials suggest the actual operational loss for 2025 sits closer to $8 billion. While this is a friendlier interpretation, it still represents a massive capital drain driven heavily by infrastructure demands.
A significant portion of this capital is flowing directly to Microsoft. The leaked documents show OpenAI handed over $17.2 billion to its primary backer in 2025 for various expenses. Of that total, nearly $10.5 billion was specifically earmarked for research and development, which primarily covers the immense compute power required to train new AI models. This aggressive spending aligns with the company's broader, albeit theoretical, pledge to invest up to $1 trillion in future data center buildouts.
The IPO Reality Check: Selling Theoretical Profit
With OpenAI preparing to go public later this year, these leaked financials set the stage for a highly unconventional initial public offering. The core tension is no longer about whether OpenAI can generate revenue - $13.07 billion is a formidable baseline - but whether its growth rate can ever mathematically catch up to the sheer volume of cash required to sustain its compute infrastructure. The $10.5 billion R&D bill to Microsoft proves that the cost of intelligence is scaling just as fast as the revenue it generates.
However, the public market's current appetite for tech giants suggests these losses might not deter investors. Much like SpaceX, which commands a massive valuation based on the future promise of Mars colonization rather than immediate profitability, OpenAI is selling a paradigm shift. In a market environment that currently values theoretical dominance over actual profit margins, OpenAI's astronomical burn rate will likely be pitched as a necessary moat against competitors, rather than a fundamental business flaw.
