Microsoft and OpenAI End Exclusive Partnership: What Changed and What It Means

On April 27, 2026, Microsoft and OpenAI announced they were rewriting the terms of the most consequential partnership in the AI industry. The exclusive cloud arrangement that gave Microsoft sole rights to host OpenAI's models is over. OpenAI can now run its products on any cloud provider it wants.

This is not a breakup. Microsoft keeps its 27% equity stake, its IP license through 2032, and its status as OpenAI's "primary cloud partner." But the walls around the garden are coming down, and the implications stretch from enterprise procurement to the cloud computing price wars.

Here is what changed, what did not, and why it matters if you are buying or building AI tools.

What the deal looked like before

When Microsoft invested $1 billion in OpenAI in 2019, the two companies signed an agreement that made Microsoft the exclusive cloud provider for OpenAI's models. OpenAI got compute. Microsoft got first-mover access to the most important AI models in the world, and the ability to integrate them into Azure, Copilot, and every product line it had.

The revenue sharing went both ways. Microsoft paid OpenAI for the IP license. OpenAI paid Microsoft a percentage of its revenue, reportedly around 20%, for the Azure infrastructure and partnership. There was also the much-discussed "AGI clause" that would have voided the exclusivity if OpenAI achieved artificial general intelligence.

For years, this arrangement worked. Azure became the default cloud for companies wanting to use GPT-4, GPT-5, and later models. Microsoft integrated OpenAI's technology into Copilot across Office, Windows, and developer tools. OpenAI got the compute it needed to train frontier models.

But the cracks started showing in early 2026.

What triggered the change

In February 2026, OpenAI signed a $50 billion deal with Amazon Web Services that included plans to run certain OpenAI models on AWS infrastructure. The Financial Times reported that Microsoft threatened legal action over that deal, arguing it violated the exclusivity terms.

Then, in an internal memo obtained by CNBC, OpenAI Chief Revenue Officer Denise Dresser told staff that the Microsoft partnership had "limited our ability to meet enterprises where they are -- for many that's [Amazon] Bedrock." She described the demand from enterprises wanting OpenAI models on AWS as "frankly staggering."

The pressure was coming from customers. Large enterprises that had standardized on AWS or Google Cloud did not want to add Azure just to run GPT models. They wanted OpenAI where their data already lived. OpenAI was leaving money on the table by being Azure-only.

The new terms, point by point

Here is what actually changed in the amended agreement:

Non-exclusive licensing. Microsoft's license to OpenAI's IP is now non-exclusive. OpenAI can offer its models through Amazon Bedrock, Google Cloud, Oracle, or any other cloud provider. Microsoft retains its license through 2032 and remains the "primary cloud partner," but that label now carries less weight.

Revenue cap. OpenAI continues to pay Microsoft the same percentage of its revenue (reported to be 20%) through 2030, but that total payment is now subject to a cap. The exact cap number was not disclosed. Once OpenAI hits it, Microsoft stops receiving additional revenue share from OpenAI's growth.

No more AGI clause. The revenue sharing is now described as "independent of OpenAI's technology progress." That is a direct reference to the AGI clause. Under the old deal, hitting some definition of AGI could have voided the arrangement. That contingency is gone.

Microsoft stops paying OpenAI. Under the previous deal, Microsoft paid OpenAI a revenue share as part of the licensing. That payment from Microsoft to OpenAI has ended.

Azure stays primary. Microsoft Azure remains OpenAI's primary cloud infrastructure provider. OpenAI's own workloads, training runs, and core infrastructure still run on Azure. What changed is that OpenAI can now also serve customers through other clouds.

What this means for enterprise AI buyers

If you are running a large organization and evaluating AI tooling, this deal changes your calculus in a few specific ways.

You no longer need Azure to get OpenAI models. If your company is an AWS shop, you will soon be able to access GPT models through Amazon Bedrock without setting up an Azure account. Amazon CEO Andy Jassy confirmed that OpenAI models will be available on Bedrock "in the coming weeks." The same will likely apply to Google Cloud and other providers over time.

Multi-cloud AI strategies get easier. Previously, organizations using multiple cloud providers had to route OpenAI API calls through Azure specifically. That meant separate contracts, separate billing, and separate compliance reviews just for one vendor's models. Now you can consolidate your AI vendor management under fewer procurement umbrellas.

Pricing competition may increase. With OpenAI models available on multiple clouds, providers will compete on the total package: pricing, throughput, latency, compliance certifications, and bundled services. Azure loses its monopoly premium on OpenAI access. Expect pricing to adjust accordingly.

The OpenAI vs. Anthropic vs. Google decision gets cleaner. Before this change, choosing OpenAI often meant choosing Azure. That bundled decision made comparisons harder. Now you can evaluate OpenAI's models against Claude and Gemini on the same cloud, with the same data governance framework, and make a more apples-to-apples comparison.

What this means for the cloud market

Microsoft gave up a significant competitive moat. Azure's exclusive access to OpenAI was a major selling point for enterprises considering cloud migration or multi-cloud strategies. Without exclusivity, Azure has to compete on its own merits: infrastructure quality, existing enterprise relationships, and the depth of its AI platform beyond just hosting OpenAI.

Amazon is the immediate winner. AWS customers who wanted GPT models had to maintain Azure subscriptions. That friction is gone. Bedrock already offers models from Anthropic, Meta, Cohere, and others. Adding OpenAI makes it the broadest model marketplace in cloud computing.

Google Cloud gains too, though its position is more complex since Google competes with OpenAI directly through Gemini. Offering OpenAI models alongside Gemini would be an acknowledgment that customers want choice, even when Google has its own frontier models.

What this means for Microsoft

Microsoft is not losing money here. The company keeps its 27% stake in OpenAI, which is worth many billions more than what it invested. It keeps revenue share payments through 2030, capped but still substantial. It keeps the IP license through 2032.

What Microsoft loses is strategic leverage. The exclusivity gave Microsoft a unique position: if you wanted GPT-4 or GPT-5 in production, you had to talk to Microsoft. That drove Azure adoption, enterprise agreements, and Copilot licensing. Without it, Microsoft has to earn that business on product quality rather than lock-in.

Microsoft's stock dipped on the news, which tells you how the market reads this. Investors valued the exclusivity premium. Removing it means Microsoft has to work harder to maintain its AI cloud revenue growth.

What this means for OpenAI

OpenAI gains flexibility. The company can now pursue enterprise deals on any cloud, removing a major objection from procurement teams that did not want to add Azure to their vendor list. This opens up the AWS-heavy enterprise market, which is the largest segment of cloud computing.

The cap on revenue share payments to Microsoft is also significant. Under the old deal, as OpenAI's revenue grew, so did the payments to Microsoft. With a cap, OpenAI keeps more of its revenue as it scales. That matters for a company reportedly targeting an IPO and needing to show improving unit economics.

The Amazon deal alone, worth $50 billion, validates the multi-cloud strategy. OpenAI can now pursue similar deals with Google, Oracle, and other providers without legal friction.

What stays the same

Despite the headlines, several important things did not change:

• •Microsoft remains OpenAI's largest shareholder at roughly 27% on an as-converted diluted basis
• •Azure continues to host OpenAI's training infrastructure and core workloads
• •The IP license runs through 2032
• •OpenAI still pays Microsoft 20% of revenue through 2030 (subject to the new cap)
• •The fundamental technology partnership, including Copilot integration across Microsoft products, continues

This is an evolution, not a divorce. Microsoft and OpenAI are still deeply intertwined. The difference is that OpenAI is no longer monogamous when it comes to cloud hosting.

Timeline of the partnership

Bottom line

The Microsoft-OpenAI partnership is still one of the most important relationships in tech. But it is no longer the walled garden it used to be. For enterprises, this is good news: more choice, more competition, and less lock-in. For Microsoft, it means competing on merit rather than exclusivity. For OpenAI, it removes a ceiling on enterprise revenue.

If you are evaluating AI tools or planning your cloud AI strategy, the calculus just shifted. The right move depends on your existing cloud footprint, your compliance requirements, and which models you actually need in production. Our AI tool comparison pages can help you evaluate the options side by side.