For the past several years, if your company wanted to use OpenAI models through a cloud provider, that cloud provider was Microsoft Azure. That was by design, a cornerstone of the $13 billion investment Microsoft made in OpenAI starting in 2019. Now that cornerstone has been removed.
OpenAI and Microsoft announced a restructured partnership in late April 2026. Both companies called it a win. By most accounts, they are both right.
What the Original Deal Was
To understand what changed, it helps to know what existed before.
When Microsoft invested in OpenAI, it got something unusually valuable: exclusive rights to be the cloud provider for OpenAI's products. This meant that if you wanted to access GPT-4 or any other OpenAI model through cloud infrastructure, only Azure was an option. Customers who had existing relationships with AWS or Google had to route their OpenAI usage through Microsoft anyway, which was exactly the kind of leverage Microsoft was paying for.
That exclusivity is now gone.
What Actually Changed
The new deal replaces the exclusive arrangement with a nonexclusive license. Microsoft retains the right to use OpenAI's models and technology through 2032, but it no longer holds the exclusive. OpenAI can now distribute its models through any cloud provider.
The financial terms shifted too. Microsoft previously received a payment from OpenAI when Azure customers paid for OpenAI access. That flow stops. OpenAI will continue paying Microsoft a revenue share at the same 20% rate, but the total obligation is now capped, which limits Microsoft's upside as OpenAI's revenue grows.
Within 24 hours of the announcement, Amazon confirmed that AWS Bedrock would offer OpenAI's latest models, Codex, and OpenAI's agent-building product. OpenAI also expanded its existing AWS commitment by $100 billion over eight years.
Why Both Sides Said Yes
Microsoft got something real out of this too. It shed the legal exposure that came with being OpenAI's exclusive partner. As OpenAI's business grows and the regulatory environment around AI tightens, being the exclusive distributor of the world's most widely used AI system carries its own risks. Microsoft CEO Satya Nadella was direct: the new structure gives Microsoft room to compete aggressively on its own AI products rather than simply being the infrastructure layer for OpenAI's.
For OpenAI, the motivation is more straightforward. Exclusivity was useful when Microsoft's investment was keeping the company alive. Now that OpenAI has its own revenue, products, and capital, exclusivity is a constraint. It limits where OpenAI can sell and creates dependency on a single relationship.
What This Means in Practice
For most people using AI tools, nothing changes immediately. If you are already using ChatGPT or the OpenAI API through Azure, that keeps working exactly as before.
The change matters most for businesses making cloud decisions. OpenAI models are now available to AWS customers without routing through Azure, which is significant for the enormous share of enterprise technology that runs on Amazon's cloud. Companies that had been choosing between their existing AWS infrastructure and Azure-only OpenAI access no longer face that trade-off.
It also accelerates competition between the major cloud providers on AI. AWS, Google Cloud, and Azure are now all competing to be the preferred way to access the same OpenAI models, which means pricing pressure, better tooling, and more integration options for enterprise customers. That is good news for anyone buying AI services rather than selling them.
The Bigger Picture
The OpenAI-Microsoft restructuring is part of a broader trend. As AI models become more widely available, the competition shifts from "who has the best model" to "who has the best platform to deploy, manage, and integrate AI at scale."
Microsoft is betting that its deep integration into enterprise software, including Office, Teams, Azure, and GitHub, is a more durable advantage than exclusivity ever was. OpenAI is betting that model quality and brand recognition will keep it relevant regardless of which cloud infrastructure carries its products. Both bets can be right simultaneously, which is probably why both sides agreed to the new terms.
This is not a breakup. It is two companies recognizing that their interests have evolved, and adjusting accordingly.