Leaked documents published this week provide concrete terms behind the OpenAI and Microsoft arrangement, indicating OpenAI pays Microsoft roughly 20 percent of revenue on many products and that the deal spells out inference and cloud compute cost allocations. Those figures matter for AI pricing, total cost of ownership for hosted models, and the choices businesses make about cloud providers.

Background: Why the OpenAI and Microsoft deal matters

The partnership pairs OpenAI model technology with Microsoft Azure infrastructure and commercial channels. Two contractual levers are especially important for customers and competitors: revenue share and inference costs. In plain language, revenue share is the portion of product revenue paid to a platform partner, while inference costs are the compute expenses that accrue each time a model generates output.

Key findings from the leak

• Revenue share: The documents indicate a roughly 20 percent revenue share for many products, a meaningful figure for gross margins on AI services sold through Microsoft affiliated channels.
• Cost transparency: The papers disclose how inference and cloud compute costs are allocated between the parties, helping procurement teams model costs more accurately.
• Exclusive commercial commitments: Large Azure commitments and commercial exclusivity influence where and how OpenAI models are deployed in production.
• Timing and sources: The disclosures were published mid November 2025 and center on OpenAI and Microsoft as the major parties involved.

Implications for businesses and cloud competition

What follows from these terms for AI pricing and vendor strategy:

• Pricing and margins: A 20 percent revenue share raises the breakeven threshold for AI products sold through Microsoft affiliated channels. That pressure can push end prices up or compress vendor profitability when token based or API call rates are taken into account.
• Cloud choice and vendor lock in: Large Azure commitments and exclusive commercial terms increase the cost of designing for multiple cloud suppliers. Firms will weigh tighter product integration and go to market reach against reduced portability and potential lock in.
• Cost predictability: Disclosure of inference cost allocation is notable because inference often dominates variable expenses. Clearer rules on who pays for compute, embeddings storage and fine tuning help teams forecast total cost of ownership for Gen AI projects.
• Competitive dynamics: Public revenue share terms let rivals such as AWS and Google Cloud craft alternative offerings and price plans. They also make it easier for buyers to compare usage based pricing models and negotiate better terms.
• Regulatory and procurement risk: High visibility into commercial terms can attract scrutiny from customers, competitors and regulators concerned about platform gatekeeping and fair competition.

Practical takeaways

If your organization consumes hosted model access or builds solutions on top of provider APIs, consider these immediate actions:

• Revisit vendor economics and model unit economics to include a 20 percent revenue share and variable inference fees when calculating AI ROI.
• Demand billing transparency. Breakdowns that separate platform fees, per request compute charges, and token based billing matter for internal cost allocation.
• Design for portability when feasible. Where multiple cloud suppliers are an option, quantify the trade off between integration benefits and the cost of reduced mobility.
• Negotiate for predictable pricing structures such as tiered API rates, committed use discounts on Azure, or clarified responsibilities for fine tuning costs.
• Monitor policy and market developments. Highly publicized deals can reshape contract norms and influence future enterprise procurement practices.

Conclusion

These leaked documents give rare visibility into the financial plumbing of a major AI partnership, showing how platform economics translate into AI pricing, cloud selection and market strategy. For businesses moving AI into production, the key lesson is simple: understand both the revenue share terms and inference cost structure before committing to a vendor relationship. As AI spending grows, those contractual details will increasingly determine who captures value, how affordable AI services become, and how procurement teams forecast total cost of ownership.