Sam Altman and Satya Nadella
For years, OpenAI‘s biggest financial backer was also its biggest constraint. That changed Monday. OpenAI and Microsoft announced a revised partnership that ends Microsoft’s exclusive license to OpenAI’s technology—the same technology Microsoft had been backing since 2019, to the tune of $13 billion.
The new deal allows OpenAI to sell its products freely across rival cloud platforms, including Amazon Web Services and Google Cloud. It’s as much an admission as it is an agreement: the old arrangement wasn’t working for OpenAI anymore.The tension had surfaced publicly earlier this month, when OpenAI revenue chief Denise Dresser sent an internal memo telling staff that the Microsoft partnership “has limited our ability to meet enterprises where they are—for many, that’s Bedrock.”
AWS’s Bedrock platform gives companies access to all major AI models, and OpenAI had been effectively locked out of competing there. That memo landed less than two months after Amazon announced plans to invest up to $50 billion in OpenAI as part of a sweeping strategic partnership—a deal Microsoft had reportedly considered legal action to stop.
Microsoft keeps cloud primacy, just not the monopoly
Under the reworked terms, Microsoft stays OpenAI’s primary cloud partner.
OpenAI products still ship on Azure first, unless Microsoft decides otherwise. But OpenAI can now serve its full product lineup to customers on any cloud—a meaningful shift for enterprise deals that increasingly run through AWS or Google Cloud.Microsoft’s license to OpenAI’s IP runs through 2032, now non-exclusive. The revenue structure has also been cleaned up: OpenAI continues paying Microsoft a 20% revenue share through 2030, now capped and no longer tied to OpenAI’s technology milestones.
Microsoft stops paying a revenue share to OpenAI entirely. Shares of Microsoft dipped roughly 1% in premarket trading Monday, as some investors read the shift as the company losing a hard-won competitive edge.The two companies say the work ahead remains ambitious—scaling gigawatts of new datacenter capacity, collaborating on next-generation silicon, and applying AI to cybersecurity. The chip collaboration is notable: both companies have been working to reduce their dependence on Nvidia by developing custom AI silicon, and the revised partnership keeps that work intact.
The AGI clause that spooked Microsoft’s lawyers has been removed
Buried in the old agreement was a provision few outside the companies fully understood—but that Microsoft’s legal team almost certainly did. If an expert panel agreed that OpenAI had reached artificial general intelligence, the revenue-sharing arrangement could have ended abruptly. The uncertainty that clause created for Microsoft was real. It’s now gone. Revenue payments from OpenAI to Microsoft run through 2030 regardless of any AGI declaration, giving both sides a cleaner, more predictable runway.Microsoft remains a major OpenAI shareholder, its stake valued at roughly $135 billion—about 27% of the company on a diluted basis. Wedbush analyst Dan Ives called the restructuring a net positive for Microsoft: six years of locked-in IP control, a significant equity position, and an end to the “ongoing partnership limbo.” For OpenAI, it removes one of the more visible structural barriers to an IPO, which the company has signaled could arrive as soon as this year.Jury selection in Elon Musk’s federal lawsuit against OpenAI and Microsoft began the same day—a reminder that not all of this partnership’s complications resolve so cleanly on paper.
