Key Takeaways:
OpenAI's audited 2025 financials show a company scaling revenue faster than almost any enterprise software business in history — while burning cash at a pace that raises fundamental questions about the economics of frontier AI.
OpenAI generated $13.07 billion in revenue in 2025, more than tripling the $3.7 billion it reported a year earlier, according to audited financial statements first published by blogger Ed Zitron and independently verified by the Financial Times. But total costs and expenses reached roughly $34 billion, producing an operating loss of about $21 billion — up from $8.78 billion in 2024.
"The revenue trajectory is extraordinary by any measure, but the cost structure reveals the brutal math of frontier AI," said Gil Luria, managing director at D.A. Davidson. "OpenAI is spending $1.60 for every dollar it brings in, and the biggest line item — $19.18 billion in research and development — shows no sign of peaking."
The headline GAAP net loss of $38.53 billion includes roughly $41.55 billion in non-cash charges tied to OpenAI's conversion from nonprofit to for-profit status, including changes in the fair value of convertible interests and warrant liability. Strip those out, and the adjusted loss comes to about $8 billion, according to a Financial Times source. The distinction matters: the $38.5 billion figure is an accounting artifact of the restructuring, not a measure of cash burn.
The cost structure behind the losses
R&D consumed $19.18 billion in 2025, up from $7.81 billion the prior year, making it by far the largest expense. Within that figure, $10.59 billion went to Microsoft in intercompany payments for cloud computing and model training infrastructure, according to Ars Technica's line-item reporting. Cost of revenue — the expense of running inference for ChatGPT and API customers — rose to $7.5 billion from $2.65 billion, reflecting the massive compute requirements of serving hundreds of millions of users.
Sales and marketing spending jumped to $5.73 billion from $1.11 billion, as OpenAI invested heavily in enterprise sales teams and consumer brand marketing. General and administrative costs reached $1.57 billion, up from $907 million.
The company's expense-to-revenue ratio improved year over year — it spent $2.37 to generate every dollar of revenue in 2024 versus $1.60 in 2025 — but remains far from the rule-of-40 benchmark that software investors typically demand from high-growth companies.
What the numbers mean for the AI sector
OpenAI's financials offer the clearest window yet into the economics of frontier AI development. The company's $13 billion in revenue already exceeds the annual sales of established enterprise software giants like Workday and ServiceNow, yet its R&D spending alone is larger than the entire market capitalization of most AI startups.
The three distinct loss measures — $21 billion operating, $38.5 billion GAAP, and roughly $8 billion adjusted — require careful interpretation. The adjusted figure is the closest proxy for cash losses from operations, and it suggests OpenAI burned roughly $8 billion in 2025 before accounting for the one-time restructuring charges. Internal management documents previously reported by The Information projected a $14 billion loss for 2026, though those estimates predated the audited 2025 results.
OpenAI has filed with the SEC for an initial public offering expected later this year, making these numbers a preview of what the S-1 will contain. The company's ability to reach profitability will depend on whether it can slow R&D spending growth without losing the performance race against rivals like Anthropic, Google's DeepMind, and Meta. Cutting the $5.73 billion marketing budget would be easier but could slow enterprise adoption at a critical moment.
For investors evaluating the AI infrastructure trade, the data points are sobering. Even the world's most prominent AI company, with a first-mover advantage and a consumer brand that rivals Google and Apple, is spending far more than it earns. If OpenAI cannot reach operating breakeven within two to three years, the implications extend beyond its own valuation — they call into question the $500 billion in projected AI infrastructure spending that cloud providers and chipmakers are betting on.
This article is for informational purposes only and does not constitute investment advice.
