OpenAI rewrites Microsoft deal
Azure exclusivity ends and AGI clause disappears, Big Tech partnership turns into a licensing arrangement with exit doors
Images
Image description
the-decoder.com
OpenAI and Microsoft have rewritten the contract that bound the most visible AI lab to a single cloud. According to The Decoder, the revised terms end Azure exclusivity and remove an “AGI clause” that previously tied intellectual-property rights to a future threshold of capability, while extending Microsoft’s non-exclusive license through 2032.
The immediate business change is distribution: OpenAI can now sell products through other cloud providers, a shift reportedly triggered by plans to offer services on Amazon Web Services. Microsoft will still get first launch on Azure, but “first” is no longer “only,” which matters in a market where the biggest cost is not model research but the electricity, chips, and data-centre contracts needed to run models at scale. Under the new structure, Microsoft no longer pays OpenAI a revenue share; instead, OpenAI continues paying royalties to Microsoft through 2030, with an overall cap, while Microsoft’s upside is increasingly routed through its equity stake rather than preferential access. The removal of the AGI clause also turns a once-ambiguous tripwire into a fixed term: Microsoft’s rights no longer depend on whether OpenAI declares it has crossed a line.
That kind of clarity is useful when partnerships start to look like rivalries. Microsoft has built Copilot into Windows and Office and is investing heavily in its own model work, while OpenAI has been assembling a broader supplier base for compute and infrastructure. A non-exclusive license reduces the risk that one party can freeze the other’s roadmap, but it also reduces the strategic value of being “the” cloud for OpenAI—especially as enterprise buyers increasingly want redundancy across vendors and jurisdictions. For Europe, the deal is another reminder that the AI stack is being shaped by private contracts between US firms, with capacity decisions—where data centres get built, which chips get booked first—made long before regulators see a product launch.
Altman and Nadella reportedly negotiated the changes personally over recent weeks. The new agreement keeps OpenAI’s products launching on Azure first, even as it opens the door for OpenAI to sell the same capability elsewhere.