Back in 2019, Microsoft and OpenAI announced their initial partnership, which included a $1 billion investment from Microsoft and the designation of Azure as OpenAI’s exclusive cloud partner. Since then, Microsoft has invested an additional $12 billion in OpenAI and introduced key changes to the partnership earlier this year.
Arguably, the Microsoft-OpenAI collaboration has grown into one of the most successful partnerships in the technology industry. Today, as part of OpenAI"s corporate restructuring, Microsoft and OpenAI have signed a new definitive agreement that"s said to ensure long-term success for both organizations. Microsoft’s stake in OpenAI Group PBC will be about 27%, valued at approximately $135 billion. Before this restructuring, Microsoft held a 32.5% stake on an as-converted basis in OpenAI’s for-profit entity.
While there are several changes to the provisions of the partnership, the key elements remain the same. OpenAI will continue to be Microsoft’s frontier model partner, and Microsoft will retain exclusive IP rights and Azure API exclusivity until Artificial General Intelligence (AGI) is achieved. Furthermore, even if OpenAI declares AGI, the claim will now be verified by an independent expert panel.
Microsoft’s IP rights for both models and products have been extended through 2032 and now cover models developed post-AGI, subject to safety guardrails. This is a key change from the previous agreement, which revoked Microsoft"s IP rights completely after AGI. Additionally, Microsoft’s IP rights to research will remain until either the expert panel verifies AGI or through 2030, whichever is first. Beyond that, Microsoft retains the non-Research IP rights. Finally, Microsoft’s IP rights now exclude OpenAI’s consumer hardware.
The changes to the agreement also introduce greater flexibility for both companies. OpenAI can now jointly develop products with third parties, with API-based products remaining exclusive to Azure, while non-API products may be hosted on any cloud provider. For example, OpenAI can host Sora on Google Cloud if required.
Microsoft, meanwhile, is free to independently pursue AGI development, either on its own or with external partners. If Microsoft leverages OpenAI’s IP to develop AGI before it is formally declared, the resulting models will be subject to compute thresholds that far exceed the scale of today’s leading training systems.
On the commercial side, a revenue-sharing agreement between the two companies will continue until AGI is verified, though payments will now be spread over a longer period. OpenAI has also committed to purchasing an additional $250 billion in Azure services, while Microsoft will be losing its right of first refusal as OpenAI’s compute provider. This will be key to OpenAI as it is partnering with a host of companies to build out hundreds of billions in AI infrastructure.
Finally, OpenAI can now provide API access to U.S. government national security customers regardless of cloud provider and is permitted to release open-weight models that meet defined capability criteria.