OpenAI Group PBC and Microsoft Corp. have revised several sections of the contract that governs their technology partnership.

The update, which was announced today, is at least the third that the companies have made since the start of 2025.

The partnership began in June 2019 with a $1 billion investment from Microsoft. Four years later, it became OpenAI’s exclusive cloud provider. The companies dropped the exclusivity clause last January, but kept many parts of the original agreement.

One of the sections that remained unchanged relates to OpenAI’s application programming interface. Developers use the API to integrate the company’s AI models into their applications. Until today, Microsoft was the only cloud provider that could offer the API through its cloud platform.

The revised agreement will enable OpenAI to “serve all its products to customers across any cloud provider.” In a post on X published today, Amazon.com Inc. Chief Executive Andy Jassy stated that the ChatGPT developer’s models will become available to Amazon Bedrock customers within a few weeks. 

The cloud giant’s customers will also receive access to an offering called the Stateful Runtime Environment. According to OpenAI, it can help developers manage the data that AI agents use to automate works. The offering also promises to ease certain related tasks, such as ensuring that AI agents adhere to a company’s cybersecurity policies.

Microsoft will retain enhanced access to OpenAI products. In particular, the company stated that “OpenAI products will ship first on Azure,” but only if certain conditions are fulfilled. The requirement won’t apply if “Microsoft cannot and chooses not to support the necessary capabilities.”

The original partnership agreement gave Microsoft an exclusive license to intellectual property related to OpenAI’s models and certain other offerings. Under the revised agreement, the tech giant has a nonexclusive license that will run through 2032. That creates an opportunity for OpenAI to ink technology sharing partnerships with other parties, which could translate into new revenue streams. 

The change marks the second time that Microsoft’s access to OpenAI intellectual property has been narrowed. The last such change was made last October, when the AI provider restructured as a public benefit corporation. That revision excluded intellectual property related to OpenAI’s upcoming consumer hardware lineup. 

Under the previous agreement, the companies paid a portion of their AI revenue revenue to one another. According to CNBC, Microsoft received a 20% cut from the sale of OpenAI products such as ChatGPT subscriptions. Those payments will continue “at the same percentage but subject to a total cap” through 2030. Microsoft, for its part, will stop giving OpenAI a share of its AI revenue.

The revisions come amid an effort by the tech giant to reduce its reliance on the ChatGPT developer’s technology. Earlier this month, Microsoft released a set of internally developed AI models optimized for use cases such as generating images and transcribing speech. At the same time, the companies continue to work closely together in other areas. Last year, OpenAI committed to purchasing $250 billion worth of Azure services.

Photo: Pixabay