Microsoft and OpenAI End Their Exclusive Deal

Published: April 28, 2026

After seven years of exclusivity, Microsoft and OpenAI have rewritten their partnership. OpenAI can now sell its models on any cloud provider. Microsoft keeps a non-exclusive license to OpenAI's IP through 2032, but will no longer pay a revenue share. The AGI clause — the provision that would have fundamentally altered the relationship if OpenAI achieved artificial general intelligence — has been removed entirely.

What changed:

  • OpenAI is no longer Azure-only. The ChatGPT maker can now serve customers across AWS, Google Cloud, Oracle, or any other provider. Azure remains the "primary cloud partner" and gets first dibs on shipping new products — but only if Microsoft can support the "necessary capabilities."

  • Revenue share is dead. Microsoft stops paying OpenAI a cut of what it resells. OpenAI continues paying Microsoft through 2030, capped at an undisclosed total, but the flow is now one-directional.

  • The AGI clause is gone. Previously, if OpenAI built artificial general intelligence, the Microsoft deal would change materially. That language has been stripped out. With no concrete definition of AGI and frontier models improving every quarter, the clause had become a legal ambiguity neither side wanted.

  • Microsoft remains a major investor. The $13.75 billion already poured in doesn't change. Microsoft still owns a significant stake. But the strategic rationale has shifted from "we own the distribution channel" to "we own the enterprise layer above it."

Why now:

The original deal made sense in 2019 when OpenAI was a research lab and Azure was the only hyperscaler willing to bet big. In 2026, frontier models are a commodity. GPT-5.5, Claude, Gemini, DeepSeek — the gap between them narrows monthly. Exclusivity stopped being a competitive advantage and started being a constraint.

For OpenAI, the problem was distribution. Being locked to Azure meant losing enterprises that standardized on AWS or Google Cloud. CEO Sam Altman has been publicly courting multiple cloud partners for months; this just formalizes what was already happening.

For Microsoft, the problem was strategic. Every dollar it sent to OpenAI as revenue share was a dollar it couldn't invest in its own models, its own agents, or its own Copilot ecosystem. The breakup frees capital and removes the appearance of conflict when Microsoft promotes its own AI tools alongside OpenAI's.

What it means:

This doesn't hurt Microsoft in the short term. Enterprises that spent years building on Azure won't migrate overnight. Azure still gets OpenAI products first. Microsoft still has the deepest integration between Copilot and OpenAI's stack.

But the long-term shift is real. OpenAI is now a model vendor selling to everyone, not a Microsoft subsidiary in all but name. Microsoft is now an AI platform competing with everyone, including OpenAI. The partnership that defined the first phase of the generative AI era is formally over.

The second phase is multipolar. No single cloud owns the best models. No single model owns the best distribution. The winners will be whoever builds the layer above both — the orchestration, the agents, the workflow integration that makes models actually useful.

Microsoft thinks that's Copilot. OpenAI thinks that's ChatGPT and Codex. Everyone else is betting on open models, custom inference, or vertical-specific agents.

The monopoly on the future just ended. The fight for what comes next starts now.