OpenAI-Microsoft, ending exclusive agreement to accelerate AI model dissemination

OpenAI and Microsoft are converting their existing exclusive contract into a non-exclusive arrangement, allowing OpenAI’s AI models to be offered not only on Microsoft Azure but also potentially deployed through other cloud platforms such as Amazon Web Services and Google Cloud. Previously, the partnership between the two companies was seen as a core pillar of the generative AI market, and this adjustment is understood as an effort to broaden OpenAI’s distribution channels while easing Microsoft’s infrastructure burden.

OpenAI and Microsoft announced on the 27th local time that they have revised their contract, transforming Microsoft’s previous exclusive rights to use OpenAI models into a non-exclusive license. As a result, GPT models can now be sold and supplied in cloud environments outside of Azure. However, OpenAI still regards Microsoft as a core cloud partner and commits to prioritizing deployment of its products on Azure without special technical restrictions. This means the change is not a complete breakup but an expansion of cooperation within the existing partnership framework.

The main framework of the contract has also been partially preserved. Microsoft will continue to hold the intellectual property rights to OpenAI models and products until 2032, but these rights are no longer exclusive. On the other hand, revenue generated from selling OpenAI models through Microsoft’s service network will no longer be paid as a separate share to OpenAI, which is seen as a less favorable change for OpenAI. Additionally, the condition that OpenAI will allocate a fixed proportion of its model and product sales revenue to Microsoft before 2030 remains in place, with industry estimates suggesting this proportion is around 20%. Notably, the previously controversial clause—that profit sharing would cease once artificial general intelligence (AGI), i.e., AI with human-level or higher general reasoning capabilities, is achieved—has been removed. Since there is no clear industry consensus on the definition of AGI, some believe it is difficult to use as a practical contractual standard.

This revision also signals a significant easing of recent tensions between the two companies. OpenAI has decided to launch enterprise-level AI services called Frontier based on Amazon Web Services infrastructure, while Microsoft last month indicated that this move might violate existing contracts and even mentioned the possibility of legal action. Ultimately, the contract adjustment is seen as enabling OpenAI to reach a broader customer base during its preparations for an initial public offering, while Microsoft can step back from the burden of continuously expanding data centers to meet OpenAI’s demands. Analysts from Evercore ISI and Barclays believe that Microsoft can now more efficiently allocate resources toward its own AI services like Copilot.

Market consensus generally interprets this move as a response to strategic needs and changing regulatory environments. OpenAI needs to promote its models more broadly to enhance enterprise value, while Microsoft can break free from a structure limited to specific partners and strengthen a multi-model strategy. At the same time, the weakening of exclusivity clauses also helps reduce pressure from antitrust investigations by competition regulators in various countries. Microsoft has been investing in OpenAI since 2019, and when OpenAI restructured into a profit-seeking nonprofit last October, Microsoft acquired about 27% of the company, making it the second-largest shareholder after the OpenAI Foundation. This trend indicates that the future AI industry may shift from a closed alliance of single companies toward a model that promotes AI through multiple cloud platforms and diverse distribution channels.