Microsoft and OpenAI have restructured their long-term partnership to allow OpenAI access to computing resources from competitors. The updated agreement moves away from strict exclusivity, granting OpenAI more freedom while maintaining key elements of their collaboration.
Microsoft retains a “right of first refusal” for new capacity and continues its exclusivity over OpenAI’s API and technology for products like Copilot. OpenAI can now build additional capacity, primarily for research and model training, with Microsoft’s support.
The partnership, which extends through 2030, is critical for both companies. OpenAI depends on vast computing resources for its ambitious AI models, requiring billions of dollars in infrastructure.
Previously, OpenAI was restricted to Microsoft’s Azure cloud platform for its compute needs. However, reports suggest that some OpenAI shareholders were dissatisfied with Azure’s pace in meeting these requirements, prompting OpenAI to partner with Oracle for additional resources earlier this year, with Microsoft’s approval.
This development follows broader industry moves, including a joint venture involving Microsoft, OpenAI, Nvidia, Oracle, and Arm to build U.S.-based data centers called Starbase.
However, the partnership has faced tension in recent months, particularly after OpenAI CEO Sam Altman’s brief ousting and Microsoft’s acquisition of Mustafa Suleyman and his team from Inflection in a $650 million deal. The latter move reportedly strained relations between the two organizations.
Adding complexity to their partnership is a unique clause tied to artificial general intelligence (AGI). If OpenAI develops AGI capable of generating over $100 billion in profits, it could limit Microsoft’s access to subsequent models. Initially intended to prevent the commercialization of highly advanced AI, OpenAI is now reportedly reconsidering this clause to secure further funding from Microsoft.
