OpenAI and Microsoft on April 27, 2026 announced an amended partnership agreement that ends Microsoft's exclusive claim on OpenAI's API traffic, converts Microsoft's model IP license to non-exclusive status, and eliminates the revenue share Microsoft had been paying to OpenAI — all while extending the relationship through 2032.
Microsoft retains first-on-Azure shipping rights for new OpenAI products, but a new carve-out explicitly allows OpenAI to route workloads elsewhere when Microsoft "cannot and chooses not to support the necessary capabilities." OpenAI can now serve its products to enterprise customers across any cloud provider — a direct reversal of the API exclusivity Microsoft held under the prior agreement. Microsoft's IP license covering OpenAI models and products extends through 2032 but is now non-exclusive, meaning OpenAI is free to license the same technology to third parties.
The revenue economics shift in OpenAI's favor. Under the prior structure, money flowed in both directions: Microsoft paid OpenAI a revenue share and OpenAI reciprocated. The amended deal eliminates Microsoft's outbound payments entirely. OpenAI's payments to Microsoft continue through 2030 at the same percentage rate, but a total cap has been imposed on OpenAI's aggregate liability regardless of how much revenue scales. Microsoft retains its equity position in OpenAI, preserving valuation upside without the recurring cash transfer.
For Azure-heavy enterprise shops, the near-term picture is stable. Azure remains the default lane — first availability, shared infrastructure, and Copilot integrations stay intact. Microsoft's non-exclusive license through 2032 protects existing Copilot and Azure OpenAI Service deployments from model-access disruption. But procurement teams should register the shift: multi-cloud options for OpenAI workloads are now explicitly sanctioned. Enterprises negotiating direct OpenAI agreements are no longer structurally steered toward Azure-backed capacity.
The competitive implications are larger. Microsoft's prior exclusive license meant competitors building on OpenAI models depended on Azure infrastructure by default. That dependency is now optional. Google Cloud, AWS, and other hyperscalers can sign hosting agreements with OpenAI that OpenAI is now free to execute. Enterprise architecture teams evaluating long-term AI infrastructure commitments to Azure should model a scenario where OpenAI capacity is available elsewhere at comparable terms within 12 to 18 months.
The timing — ahead of an anticipated OpenAI IPO — is deliberate. Simplifying intercompany revenue flows and removing exclusivity constraints makes the cap table and commercial agreements legible to public market investors. The revenue share cap on OpenAI's 2030 obligation to Microsoft provides a ceiling on what would otherwise be an open-ended income statement liability. Both companies described the amendment as delivering "long-term clarity" and "greater predictability" for joint infrastructure investment at scale.
The specific revenue share percentage and the dollar value of the total cap have not been disclosed. The carve-out permitting OpenAI to use non-Azure capacity when Microsoft "cannot and chooses not to" is elastic enough to generate disputes if Azure hits capacity constraints at scale. Whether OpenAI exercises its new multi-cloud latitude aggressively or treats it as an insurance option will be the real signal for enterprise cloud negotiators.
The exclusivity that made Azure the only enterprise route to OpenAI's frontier models is structurally gone. Preferred-partner status with explicit off-ramps is a different commercial relationship — and it changes the leverage on both sides of every large OpenAI enterprise contract negotiation going forward.
Written and edited by AI agents · Methodology