The partnership that gave Microsoft a commanding position in enterprise AI — and gave OpenAI the compute and capital to reach escape velocity — has been significantly renegotiated. Both companies announced on April 27 that they have amended the terms of their deal, and the changes move in one clear direction: OpenAI gets more freedom, and the exclusivity that once defined the arrangement is effectively over.
Under the new terms, OpenAI can now offer its products through any cloud provider. AWS, Google Cloud, or anyone else can carry OpenAI's APIs and enterprise offerings. Microsoft remains OpenAI's primary cloud partner and retains first-access rights to new products — unless Microsoft cannot support the required capabilities, in which case OpenAI is free to go elsewhere. That carve-out matters. As OpenAI pushes deeper into specialized inference, custom silicon, and agentic workloads, "can Microsoft support this?" will increasingly be a live question.
On the licensing side, Microsoft's rights to OpenAI's models and products extend through 2032, but are now non-exclusive. Microsoft also stops paying revenue share to OpenAI as part of the amendment. OpenAI, for its part, continues making revenue-share payments to Microsoft through 2030 — at the same percentage as before, but subject to a total cap. Microsoft retains its equity stake in OpenAI.
Why now
The original deal was structured when OpenAI was a research organization dependent on a single infrastructure partner. That relationship made sense in 2019 and still largely made sense through 2023. It makes less sense now. OpenAI is the world's most-used AI product company, is in the middle of a large fundraise, and is actively trying to court enterprise customers who are already committed to AWS or Google Cloud. Exclusive cloud terms create friction in every one of those conversations.
From Microsoft's perspective, the concessions here are real but manageable. They lose exclusivity and stop paying revenue share — but they keep their license, their equity position, and preferential access as primary cloud partner. Microsoft is also deep enough into its own AI product buildout (Copilot, Azure AI Foundry, the GitHub Copilot ecosystem) that it no longer needs OpenAI's exclusivity as a competitive moat the way it once did.
What changes for builders
- Multi-cloud OpenAI is real now. If you've been waiting to build on OpenAI APIs but your infrastructure is locked to AWS or GCP, that friction is largely gone. Expect OpenAI to start competing more directly for workloads outside Azure.
- The platform stack is getting more modular. The tight Microsoft-OpenAI bundle was a forcing function for many enterprise AI deployments. With looser coupling, cloud selection and model selection start to decouple — which is better for buyers and more complex for sales teams.
- The license through 2032 still matters. Microsoft retains the right to use OpenAI's models in its own products for years. Copilot and Azure AI services built on top of OpenAI aren't going anywhere. The non-exclusivity applies to who else can use OpenAI — not to Microsoft's ability to keep using it.