A new cache of leaked documents is offering a rare inside look at one of tech’s most closely watched partnerships: OpenAI and Microsoft. After months of financial speculation surrounding OpenAI’s explosive growth and talk of an eventual IPO, these documents provide new clues into how much the AI company is actually paying its biggest backer.
According to materials reviewed by tech blogger Edward Zitron, Microsoft received $493.8 million in revenue-share payments from OpenAI in 2024. That figure surged to $865.8 million in just the first three quarters of 2025, underscoring OpenAI’s rapid financial expansion, and its equally rapid cost escalation.
Revenue-Sharing Arrangement Under New Scrutiny
The disclosures align with a long-circulated figure: OpenAI reportedly shares 20% of its revenue with Microsoft. This stems from Microsoft’s investment exceeding $13 billion into the AI developer. However, neither company has ever publicly confirmed that percentage.
The situation becomes more complex because Microsoft also shares revenue back with OpenAI. Roughly 20% of the income from Bing and the Azure OpenAI Service is reportedly returned to OpenAI, given that both rely heavily on the company’s models. But, according to a source who spoke with TechCrunch, the leaked numbers reflect Microsoft’s net revenue share, after subtracting what Microsoft paid out to OpenAI.
Since Microsoft never breaks out Bing or Azure OpenAI revenue in earnings reports, it remains difficult to estimate how much it is refunding to OpenAI on its end.
Even so, the payment numbers provide valuable hints. Based on the 20% revenue-share figure, OpenAI’s revenue appears to have hit at least $2.5 billion in 2024, and more than $4.3 billion in the first three quarters of 2025. Independent reporting from The Information places 2024 revenue closer to $4 billion and early 2025 revenue at $4.3 billion.
CEO Sam Altman recently suggested that OpenAI’s real financial performance is far higher, saying revenue is “well more” than $13 billion annually and could surpass a $20 billion annualized run rate by year-end. He has even floated the possibility of hitting $100 billion in revenue by 2027.
Huge Compute Bills Are Eating Into Revenue
While revenue is climbing fast, costs appear to be rising even faster. Zitron’s analysis suggests OpenAI spent around $3.8 billion on inference compute in 2024. By the first nine months of 2025, that number jumped to $8.65 billion.
Inference, the compute used to run and generate responses from AI models, is largely a cash expense. A source told TechCrunch that, unlike training compute (mostly covered by Microsoft-issued credits), inference costs require real cash outflow.
OpenAI still leans heavily on Microsoft Azure for compute, though it has recently diversified with deals involving CoreWeave, Oracle, AWS, and Google Cloud. Prior reports estimated OpenAI’s total compute spend at $5.6 billion in 2024, with $2.5 billion in cost of revenue for just the first half of 2025.
These numbers indicate that OpenAI may be spending more on running models than it earns, a reality that is fueling fresh debate about the sustainability of the AI boom.
If the industry’s leading AI company is still in the red despite record demand, what does that mean for dozens of AI startups raising at ever-inflating valuations?
