A set of leaked internal documents gives one of the clearest looks yet at how much OpenAI is handing over to Microsoft under their multi-year partnership. These figures raise fresh questions about OpenAI’s revenue, cost structure and cloud dependency.
What the leaked documents show
- In 2024, OpenAI reportedly paid Microsoft US $493.8 million in revenue-share payments.
- In the first three quarters of 2025, the payment amount reportedly jumped to US $865.8 million.
- These payments are said to reflect about 20% revenue share from OpenAI’s revenue, under the partnership with Microsoft.
- Concurrently, documents show OpenAI’s spending on inference (i.e., running its AI models) via Microsoft’s Azure cloud may have reached US $8.65 billion in the first nine months of 2025, and about US $3.8 billion in 2024.
Why this matters
Revenue transparency
The numbers provide one of the rare public glimpses into the financial terms of the OpenAI-Microsoft deal, which has until now remained largely opaque.
Cost and margin pressure
High payments plus huge inference spending suggest that while OpenAI’s top-line may be scaling, costs are also very large — raising questions about profitability and cost discipline.
Cloud dependence & strategic implications
OpenAI is heavily reliant on Microsoft’s Azure infrastructure. The payments underscore how Microsoft benefits financially (via revenue share) and strategically (via compute and access).
Competitive & investor implications
Such disclosures influence how investors view OpenAI’s business model and how competitors or regulators see the cloud/AI landscape — especially with respect to vendor concentration, margin pressure and revenue share deals.
Background & context
- Microsoft invested billions into OpenAI, securing exclusive cloud-compute rights and model licensing arrangements.
- It has been widely reported that OpenAI’s revenue-sharing deal gives Microsoft roughly 20% of OpenAI’s revenue. TechCrunch
- The AI industry is under pressure, as running large models is extremely compute-intensive; cost curves, cloud partnerships and licensing economics matter deeply.
- Until now, OpenAI and Microsoft provided little public breakdown of how the monetisation and cost flows worked between them.
Implications for India & global tech ecosystem
- For Indian AI firms and cloud customers: this leak highlights how critical cloud-compute cost structures and vendor terms can be in AI-business models — something Indian startups should be keenly aware of.
- For Indian cloud and infrastructure policy: heavy dependence on foreign cloud vendors and revenue-share models might be a wake-up call for developing more local compute capacity and diversified cloud relationships.
- For global AI start-ups and investors: transparency about revenue share and inference spend is critical. This leak may trigger more scrutiny of other AI firms’ economics.
- For enterprise customers: understanding how underlying cloud partnerships drive cost and margin could affect how enterprises negotiate AI service deals and choose vendors.
Key things to watch & caveats
- These are leaked documents, not official disclosures. Both OpenAI and Microsoft have not publicly confirmed all the figures.
- The payments shown are likely “net revenue share” from OpenAI to Microsoft; the overall financial flows (including what Microsoft pays OpenAI, training cost subsidies, cloud credits) are more complex.
- The inference-cost numbers (US $8.65 billion) are based on extrapolations and may not represent full disclosure of all compute costs or cloud vendors.
- Revenue figures derived from the 20% share assumption are estimates; actual revenue may differ.
- The long-term profitability or sustainability of this model remains uncertain: huge cost implies that forecasts and business models should be scrutinised.
Conclusion
The leaked document revealing how much OpenAI pays Microsoft offers a rare and revealing peek into the economics of one of the most closely watched tech partnerships. That OpenAI paid Microsoft nearly half-a-billion in 2024 and over eight-hundred million by Q3 2025 underscores both the size of the business and the complexity of the cost structure. For AI startups, cloud customers, investors and policymakers — particularly in regions like India — this episode reinforces the importance of understanding cloud economics, revenue share dynamics and infrastructure dependence.
