Microsoft‑OpenAI deal loses exclusivity

Cloud infrastructure is the core of this story. Not chatbot hype — plumbing, control, and who gets to sell what. For years, Microsoft had the inside track on OpenAI because Azure was effectively the commercial home for OpenAI’s models. On April 27, 2026, that changed. Microsoft and OpenAI amended their deal so OpenAI can serve products across any cloud provider, while Microsoft’s license to OpenAI IP became non-exclusive. ### What actually got loosened? The big shift is exclusivity. Microsoft is still OpenAI’s primary cloud partner, and OpenAI products are still supposed to ship first on Azure unless Microsoft can’t or won’t support what’s needed. But OpenAI is no longer boxed into Azure for customer delivery. It can now offer its products through other clouds too, and Microsoft’s rights to OpenAI models and products now run through 2032 on a non-exclusive basis. ### Did the companies break up? No — basically the opposite. They’re still tightly linked, just under looser rules. Microsoft remains a major shareholder in OpenAI, and both companies framed the rewrite as a way to make the partnership more predictable while continuing joint work on datacenters, chips, and security. So this is not a divorce. It’s a contract rewrite after both sides outgrew the original arrangement. ### What changed on money? The economics got simpler, but not small. Microsoft will no longer pay a revenue share to OpenAI. OpenAI will keep paying Microsoft through 2030, at the same percentage as before, but now with a total cap. CNBC said that percentage is 20%, based on a person familiar with the agreement. That matters because it replaces a fuzzier long-term dependency with a dated, capped stream of payments. ### Why did OpenAI want this? Because enterprise customers don’t all live on Azure. OpenAI had started to run into a pretty basic sales problem — big companies often want AI tools where their data and existing systems already sit. If those customers are heavily on AWS or Google Cloud, forcing them through Azure adds friction. OpenAI’s own revenue leadership. ### Why did Microsoft agree? Because Microsoft kept a lot of upside even while giving up exclusivity. It still gets OpenAI products first on Azure in many cases, still collects revenue share through 2030, still holds a major equity stake, and still has IP access through 2032. The catch is that Microsoft gave up a moat — exclusive control over distribution — in exchange for a more durable, clearer arrangement. ### How fast did rivals move? Immediately. On April 28 — one day after the amended deal — OpenAI began offering its latest models and Codex on Amazon Bedrock, and the companies launched tooling for developers to build AI agents on AWS. Reuters also said OpenAI had committed to major AWS spending over eight years and had lined up 2 gigawatts of compute using Amazon’s Trainium chips. That’s not a symbolic side deal. That’s scale. ### Why does AWS matter so much here? Because AWS is the biggest proof that exclusivity really ended. For years, Microsoft’s OpenAI tie-up gave Azure a rare “you can only get this here” advantage. Now Amazon can tell customers to keep their data, apps, and AI stack in one place. Matt Garman’s line at the launch event was blunt: customers wanted OpenAI models inside AWS because their production systems and data were already there. ### Bottom line? The old Microsoft-OpenAI deal was built like a gated community. The new one looks more like an anchor tenancy. Microsoft is still the lead partner, but not the only landlord. That weakens one of Azure’s cleanest AI advantages — and gives OpenAI a much bigger commercial map.