OpenAI and Microsoft kill the exclusive deal

The most important partnership in AI just got less important. OpenAI and Microsoft jointly announced an amended agreement on April 27 that removes the exclusivity at the core of their relationship. Microsoft’s license to OpenAI IP is now non-exclusive. OpenAI can serve its products on any cloud provider, not just Azure. The change takes immediate effect.

This is not a breakup. Microsoft stays OpenAI’s primary cloud partner. OpenAI products still ship first on Azure. Microsoft keeps its IP license through 2032. But the guardrails are gone. OpenAI can now run on Amazon Bedrock, Google Cloud, or anywhere else.

The financial terms shifted too. Microsoft will stop paying a revenue share to OpenAI entirely. OpenAI continues paying Microsoft 20% of revenue through 2030, same percentage as before, but now subject to a total cap. That cap number was not disclosed.

One clause that vanished: the AGI trigger. Previous versions of the deal tied certain provisions to OpenAI achieving artificial general intelligence. The new agreement makes payments “independent of OpenAI’s technology progress.” No more defining AGI in a contract. No more verification process. The economics are now just dates, percentages, and caps.

OpenAI’s revenue chief Denise Dresser said in a memo earlier this month that the old partnership structure “limited our ability to meet enterprises where they are.” That memo turned out to be foreshadowing.

The writing has been on the wall for months. OpenAI cut a deal with Amazon in February worth up to $50 billion in investment and $38 billion in cloud commitments. Microsoft threatened legal action over that deal at first, according to documents obtained by CNBC. The two sides then signed a preliminary revision in September 2025. This amended agreement is the final version of that process.

Microsoft comes out of this with its investment intact. The company holds roughly 27% of OpenAI on a diluted basis, valued at about $135 billion as of the October recapitalization. Microsoft also committed to $250 billion in Azure spending for OpenAI workloads. That commitment remains.

But Microsoft gets something less tangible: distance. Embedding OpenAI into Bing, Office, Windows, and Copilot made Microsoft look like the enterprise AI frontrunner. It also made Microsoft look dependent on a single partner whose ambitions kept growing. The non-exclusive deal lets Microsoft build its own AI stack without looking like it’s competing with its own supplier.

For OpenAI, the deal removes a ceiling. The company has been assembling compute capacity outside Microsoft’s cloud for over a year. The Stargate project with Oracle and SoftBank targets nearly 7 gigawatts of planned capacity and over $400 billion in investment across three years. The AWS partnership added another major cloud pipeline. Until now, all of that existed in an awkward gray zone where OpenAI was technically allowed to work with other providers but contractually bound to Azure first.

Cleaner terms also make OpenAI easier to finance. Capped payments make margins easier to model. Non-exclusive licensing makes the company look like a platform that sells everywhere, not a Microsoft dependency.

The original Microsoft-OpenAI deal was built for an era when one cloud partner could fund the lab, host the workloads, and commercialize the models. That era produced ChatGPT, Copilot, and one of the fastest strategic pivots in Microsoft history. The next era will have OpenAI selling compute and models across every major cloud, while Microsoft tries to build an AI platform that stands on its own.

Sources: Microsoft Blog, OpenAI, Ars Technica, CNBC, The Neuron