For most of the past six years, the simplest way to use OpenAI's models in production was also the only way: Microsoft Azure. The 2019 investment, originally pitched as a $1 billion bet, locked OpenAI's products to a single cloud and gave Microsoft a license to anything OpenAI built, all the way up to and not including artificial general intelligence. The investment was later expanded to a reported thirteen billion dollars. That clause set the boundary for what Sam Altman could spin off into independent products. It also gave Microsoft an effective veto over what counted as AGI in the first place.
On Monday, April 27, 2026, the two companies blew up most of that arrangement.
In a joint blog post, Microsoft and OpenAI announced that Microsoft's license to OpenAI intellectual property is now non-exclusive, that OpenAI can sell its products on any cloud provider, and that Microsoft will stop paying any revenue share on OpenAI products it resells through Azure. OpenAI still owes Microsoft a revenue cut at the same 20% rate through 2030, but that obligation is now subject to a cap whose dollar amount the companies declined to disclose. The AGI trigger that previously ended Microsoft's license is gone. The license now runs to a fixed calendar date in 2032 regardless of what OpenAI builds before then.
Altman summarized the change on X with characteristic terseness: "We have updated our partnership with Microsoft. Microsoft will remain our primary cloud partner, but we are now able to make our products and services available across all clouds. Will continue to provide them with models and products until 2032, and a revenue share through 2030."
Investors took a moment to figure out who had won. Microsoft slipped about 0.20% intraday on the news. Alphabet, the most obvious beneficiary outside Amazon, jumped 2.10%. Amazon edged down 0.40%, which only made sense once analysts read the fine print: AWS already had its win locked in.
The Amazon Deal That Forced This Conversation
The new terms did not arrive in a vacuum. In February 2026, OpenAI announced a strategic partnership with Amazon worth up to $50 billion, with AWS designated as the exclusive third-party cloud distribution provider for OpenAI's enterprise platform Frontier. That deal was structurally incompatible with Microsoft's old exclusivity rights. For two months, Microsoft, OpenAI, and AWS lawyers negotiated what could and could not be sold where, with the threat of litigation hanging over every product launch.
The Monday announcement settled it. As TechCrunch put it, the rewrite ends Microsoft's "legal peril" over the Amazon arrangement. OpenAI is now free to push Frontier onto AWS, ChatGPT Enterprise onto Google Cloud, and the API onto whichever provider gives a customer the best discount. Azure remains the default, but only the default.
Microsoft framed the changes in characteristically corporate language. The company called itself OpenAI's "primary cloud partner" and noted that OpenAI products will continue to ship first on Azure unless Microsoft decides otherwise. That last clause matters. It preserves Microsoft's ability to cede a launch to a rival cloud when its own infrastructure cannot meet OpenAI's deployment needs, which the company has openly admitted has been a problem for the past two years.
What Each Side Actually Gained
Strip away the diplomatic language and the swap looks like this:
| Item | Microsoft Got | OpenAI Got |
|---|---|---|
| Revenue share to OpenAI | Stops paying | Loses that income |
| Revenue share from OpenAI (20%) | Continues through 2030, now capped | Cap protects long-term margin |
| Cloud exclusivity | Lost | Free to sell on AWS, Google Cloud |
| AGI trigger on license | Removed; runs to 2032 | Loses ability to end license by declaring AGI |
| Equity stake | Retained 27%, valued at $135 billion as of October 2025 recapitalization | Microsoft remains anchor investor |
| "Primary cloud partner" status | Retained | Azure stays the default |
The cap on OpenAI's revenue share is the line item analysts will watch most closely. OpenAI's annualized revenue is on track for $30 billion by mid-2026 according to numbers the company has shared with investors, and 20% of that pace would push the share well above what Microsoft historically modeled. A cap protects OpenAI's long-term margin once the platform crosses the threshold. It also protects Microsoft from an embarrassing public number in a future regulatory filing.
The AGI clause removal is the line item that mattered to OpenAI's leadership. Under the old terms, Microsoft's license expired the moment OpenAI's board formally declared the company had achieved AGI. That gave Microsoft a structural incentive to dispute any such declaration, and gave OpenAI a structural incentive to delay it. With the trigger replaced by a calendar date, both sides are freed from a fight neither wanted.
How the Six-Year Marriage Came Apart
The 2019 deal was a product of a different OpenAI: a small research lab with a non-profit governance structure, no consumer products, and Greg Brockman doing most of the engineering management himself. Microsoft's $1 billion bought exclusive cloud rights, exclusive licensing of OpenAI's IP up to AGI, and a financial relationship that gave Microsoft 75% of OpenAI's profits until its investment was repaid.
What no one in 2019 anticipated was ChatGPT. The November 2022 launch turned OpenAI from a research lab into the fastest-growing consumer software product in history. Suddenly the same exclusivity terms that had protected Microsoft's investment were also choking OpenAI's ability to serve customers who refused to put their workloads on Azure. Salesforce, Oracle, and most of the Fortune 100 either had multi-cloud commitments or actively preferred AWS. OpenAI was leaving real money on the table every quarter.
The relationship went public over the course of 2025 and early 2026:
What Practitioners Should Expect
For ML engineers and platform teams, three things change immediately and one thing changes slowly.
What changes immediately: choice of cloud for OpenAI inference. Until Monday, running GPT-5.5 in production effectively meant Azure. Going forward, the same model can be served from AWS Bedrock, Google Cloud Vertex AI, and Azure OpenAI Service. Pricing across providers will not equalize on day one, but the negotiating power shifts toward the customer.
What changes immediately: Frontier's reach. OpenAI's enterprise platform was always going to live primarily on AWS under the February deal. It can now land legally without anyone in Redmond filing a lawsuit. Banks, healthcare systems, and regulated workloads that already had AWS commitments now have a legitimate path to OpenAI's most secure offering.
What changes immediately: data residency conversations. EU and APAC customers who could not use Azure for sovereignty reasons can now negotiate with whichever hyperscaler has the right region.
What changes slowly: integration depth. Azure's deep stack of OpenAI integrations, including the Cognitive Services tooling, Microsoft Fabric connectors, and Sentinel security workflows, are not getting replicated on AWS or Google Cloud overnight. Engineering teams that built around Azure-specific OpenAI APIs will see those products supported, but the tooling parity is years away.
For context: LDS previously covered OpenAI's internal disclosure that Microsoft could "sink the business" and the friction over the $50 billion Amazon arrangement. The Monday announcement is the resolution of those tensions, not the start of them.
The Counterargument: Microsoft Won the Long Game
Not everyone reads the changes as a Microsoft loss. Bret Kinsella, a longtime AI infrastructure analyst, pointed out on X that Microsoft is keeping the equity stake, the technology license, and primary cloud status, while shedding a revenue obligation it never wanted to pay in the first place. The 27% stake, valued at $135 billion in the October 2025 recapitalization, makes Microsoft by far the largest single beneficiary of OpenAI's growth trajectory regardless of which cloud the products run on.
There is also a regulatory angle. Antitrust scrutiny of the original arrangement had been building on both sides of the Atlantic. The UK Competition and Markets Authority and the European Commission both opened reviews of the partnership during 2024, and the FTC under Lina Khan launched a study of generative AI investments in early 2025. Removing the exclusivity language defuses the most aggressive antitrust theory: that Microsoft was using its OpenAI stake to foreclose cloud competition. As one analyst quoted by VentureBeat put it: "They're trading a contractual lock that regulators were going to break anyway for a revenue cap and a cleaner equity story."
The skeptical read is that Microsoft is making a virtue of necessity. The optimistic read is that Microsoft anticipated where this was headed and got ahead of it. Neither read changes the practical outcome: starting now, the AI cloud market has three serious vendors of OpenAI's frontier models instead of one.
The Bottom Line
For six years, the Microsoft-OpenAI deal was the most consequential commercial arrangement in software, and it was held together by a piece of contract language drafted before ChatGPT existed. That language assumed an OpenAI that was small, science-driven, and unlikely to need scale beyond what Azure could provide. None of those assumptions survived 2024.
The Monday rewrite acknowledges what every customer already knew: AI workloads will run wherever the GPUs are cheapest and the latency is lowest, and no exclusivity contract can change that. The new terms protect Microsoft's equity, protect OpenAI's margin, defuse regulatory pressure, and give Amazon and Google the legal cover to build real OpenAI businesses. The losers, to the extent there are any, are the lawyers who spent the past three months arguing about whose cloud Frontier could land on.
As Sam Altman wrote at the end of his X post: "a revenue share through 2030." It was an oddly precise way to end a statement about a six-year breakup. The two companies are still tied together by the calendar. They are no longer tied together by anything that prevents either of them from working with everyone else.
Sources
- The next phase of the Microsoft-OpenAI partnership (Microsoft Official Blog, April 27, 2026)
- The next phase of the Microsoft OpenAI partnership (OpenAI, April 27, 2026)
- Microsoft to Stop Sharing Revenue With OpenAI (Bloomberg, April 27, 2026)
- OpenAI shakes up partnership with Microsoft, capping revenue share payments (CNBC, April 27, 2026)
- OpenAI ends Microsoft legal peril over its $50B Amazon deal (TechCrunch, April 27, 2026)
- Microsoft and OpenAI gut their exclusive deal, freeing OpenAI to sell on AWS and Google Cloud (VentureBeat, April 27, 2026)
- OpenAI breaks free of Microsoft's cloud (Axios, April 28, 2026)
- OpenAI Breaks Free From Exclusive AI Pact With Partner Microsoft (Seattle Times, April 27, 2026)
- Microsoft to Get 27% of OpenAI, Access to AI Models Until 2032 (Bloomberg, October 28, 2025)