OpenAI and Microsoft Strike a Truce: What the New Deal Means for the Future of AI

The deal between OpenAI and Microsoft, renegotiated in early April, ends the cloud-hosting exclusivity clause that required OpenAI to run its commercial workloads exclusively on Microsoft Azure. Under the new terms, OpenAI can sell its models to customers using any cloud provider, including Amazon Web Services and Google Cloud, while Microsoft retains a revenue-sharing arrangement from those sales.

The exclusivity arrangement had been in place since Microsoft first invested $1 billion in OpenAI in 2019, later increasing its total commitment to over $13 billion. As OpenAI’s compute demands grew exponentially, the restriction became a source of friction. Sam Altman sought the flexibility to offer deployment options across clouds, while Microsoft wanted to protect its investment and keep Azure as the default platform for OpenAI’s models.

The structural tension became visible in November 2023, when OpenAI’s board briefly fired Sam Altman. The episode showed how deeply the two companies were intertwined and prompted a board restructuring that gave Microsoft a non-voting observer seat.

Regulatory scrutiny also pushed the companies toward change. The Federal Trade Commission opened an investigation into the Microsoft-OpenAI relationship, examining whether the exclusive hosting deal gave Microsoft an unfair advantage in the cloud market, where it trails Amazon Web Services and competes with Google Cloud. Unwinding the exclusivity proactively reduces that antitrust risk.

For enterprise customers, the change removes a practical barrier. Many companies hesitated to adopt OpenAI’s models because doing so required a parallel commitment to Azure. Now a bank running on AWS can buy GPT access and deploy it through its existing infrastructure, eliminating the friction of managing a second cloud relationship for AI.

OpenAI is also building toward infrastructure independence. The company has been developing its own data center partnerships, including a project internally called “Stargate” that could cost over $100 billion, and has been constructing its own hardware capabilities. Reducing reliance on Azure is part of a longer-term strategy to operate as a platform company rather than a tenant.

The cloud providers have different incentives. Google has its own Gemini models and bundles them aggressively with its cloud offering. Amazon has invested over $4 billion in Anthropic, a direct OpenAI competitor, and offers models through its Bedrock platform. Neither is eager to host a rival’s product. The deal matters most for enterprise buyers who want OpenAI’s models without being locked into any single cloud.

Microsoft is hedging its bets. The company has accelerated development of its own foundation models through an internal team. By loosening the dependency on OpenAI, Microsoft protects itself if the relationship sours again or if OpenAI’s model quality declines. The move trades guaranteed Azure revenue for strategic flexibility.

The restructuring also affects the legal fight involving Elon Musk, a co-founder who left OpenAI and sued the company, alleging it abandoned its nonprofit mission and became too controlled by Microsoft. By giving OpenAI independence in cloud choices, the company can argue it is not a subsidiary of Microsoft, removing one piece of evidence Musk’s legal team had used.

The new deal does not resolve all complications. OpenAI still needs enormous computing power, and Microsoft’s data centers are built to handle that load. The transition away from Azure will take years. The revenue-sharing agreement and board observer seat remain in place, preserving Microsoft’s financial interest in OpenAI’s success.

What to watch next: how quickly OpenAI diversifies its cloud infrastructure, whether enterprise adoption accelerates now that the lock-in is gone, and whether the FTC closes its investigation or shifts to new questions about the remaining revenue-sharing and board structures.