The OpenAI Microsoft partnership has entered a new and contradictory phase: OpenAI publicly states that Microsoft has limited its ability to build a customer base, even as it commits to spending $250 billion on Azure services over multiple years. This tension reveals how dependent Microsoft has become on OpenAI’s success, and why OpenAI is now diversifying its cloud infrastructure relationships.
Key Takeaways
- OpenAI claims Microsoft restricted its customer growth despite calling the partnership foundational
- Microsoft holds 27% of OpenAI Group PBC, valued at roughly $135 billion as of 2025
- OpenAI committed to $250 billion in Azure services post-October 2025, representing 45% of Microsoft’s $281.3 billion AI backlog
- New deal loosens exclusivity: OpenAI can develop non-API products with third parties; API products remain Azure-exclusive until 2032
- OpenAI partnered with Amazon and acquired $350 million in CoreWeave shares for infrastructure diversification
The OpenAI Microsoft partnership paradox
Microsoft has invested multibillion dollars in OpenAI since 2016, including a $10 billion commitment announced in January 2023, making Azure OpenAI’s exclusive cloud provider from the outset. Yet OpenAI now says this foundational relationship has constrained its growth. The complaint is not rhetorical—it is backed by structural changes in their October 2025 deal that explicitly allow OpenAI to develop non-API products with any cloud provider, not just Microsoft. This loosening of restrictions signals OpenAI’s frustration with Microsoft’s gatekeeping and its determination to build independence.
The numbers tell the story. Microsoft’s Intelligent Cloud division, which includes Azure, generated $32.91 billion in sales in Q2 FY2026 (ended December 2025), up 28.8% year-over-year. OpenAI accounts for roughly 45% of Microsoft’s $281.3 billion AI services backlog—a staggering concentration that makes Microsoft more dependent on OpenAI than OpenAI is on Microsoft. This asymmetry explains why OpenAI feels emboldened to negotiate harder terms and pursue alternative partnerships.
How the new deal reshapes the OpenAI Microsoft partnership
The restructured agreement, announced alongside OpenAI’s October 2025 conversion to a for-profit public benefit corporation, fundamentally changes the relationship’s constraints. Under the new terms, OpenAI can develop consumer products and non-API offerings with any cloud provider—a critical escape hatch from Azure exclusivity. API products, however, remain locked to Azure until 2032. Microsoft also retains broader rights to OpenAI’s models and products (excluding consumer hardware) through 2032, and exclusive licensing of GPT-class models ends that same year.
The deal also includes a revenue-sharing mechanism that ends upon AGI confirmation by an independent panel, freeing both parties from ongoing revenue splits once artificial general intelligence is deemed achieved. This structure allows OpenAI to pursue aggressive diversification while Microsoft maintains a long-term financial stake and API exclusivity—a compromise that satisfies neither party fully but prevents a clean break.
Why OpenAI is pursuing Amazon and CoreWeave
OpenAI’s new partnership with Amazon and its March 2025 deal with CoreWeave—acquiring $350 million in shares for $11.9 billion over five years—are direct responses to Microsoft’s perceived limitations. CoreWeave, an AI infrastructure provider where Microsoft was the largest customer in 2024, gives OpenAI independent compute capacity outside Azure’s control. The Amazon relationship, praised explicitly by OpenAI as strategic for expanding market presence, signals a deliberate shift toward multi-cloud operations.
These moves are not about abandoning Microsoft—OpenAI’s $250 billion Azure commitment ensures continued deep integration for API products. Rather, they are about preventing any single cloud provider from constraining product innovation or customer acquisition. OpenAI has grown to five million business users, and that growth increasingly depends on freedom to choose infrastructure partners based on performance, cost, and availability rather than contractual obligation.
Microsoft’s precarious position in AI infrastructure
Microsoft’s heavy reliance on OpenAI revenue creates strategic vulnerability. With OpenAI representing 45% of its $281.3 billion AI backlog, Microsoft cannot afford to lose OpenAI’s business—yet it lacks leverage to prevent OpenAI from diversifying. The non-OpenAI portion of Microsoft’s backlog grew 28%, suggesting the company is building alternative revenue streams, but this growth cannot offset the risk of losing OpenAI’s dominance in the AI market.
Meanwhile, Microsoft’s 27% stake in OpenAI Group PBC, valued at roughly $135 billion, is a significant asset, but it does not translate to operational control. OpenAI’s nonprofit arm holds a $130 billion stake, maintaining structural independence that allows leadership to pursue partnerships Microsoft opposes. This structural imbalance—where Microsoft has capital but limited governance power—is at the heart of the partnership tension.
What does this mean for the broader AI market?
The OpenAI Microsoft partnership strain reflects a maturing AI market where no single vendor can monopolize infrastructure or innovation. OpenAI’s diversification strategy—Azure for APIs, Amazon and CoreWeave for non-API products and compute—is becoming the industry template. Other AI labs and startups will likely demand similar flexibility from cloud providers, eroding the exclusive partnerships that dominated 2023 and 2024.
For enterprise customers, the shift is positive: OpenAI’s multi-cloud approach reduces lock-in risk and encourages competition among cloud providers on price and performance. For Microsoft, the challenge is substantial. Maintaining a 27% stake and API exclusivity through 2032 is valuable, but it does not guarantee dominance in the AI era. The company must compete on merit, not contract terms, to retain OpenAI’s business beyond the current deal’s expiration.
Will the OpenAI Microsoft partnership survive beyond 2032?
The partnership will likely persist in some form, but on fundamentally different terms. By 2032, exclusive licensing of GPT-class models ends, and the revenue-sharing mechanism dissolves upon AGI confirmation. At that point, OpenAI and Microsoft will be negotiating as independent entities with no contractual obligation binding them together. If OpenAI has successfully diversified its customer base and infrastructure dependencies—the explicit goal of its Amazon and CoreWeave partnerships—it will have far less incentive to prioritize Microsoft’s terms.
Does the $250 billion Azure deal lock OpenAI into Microsoft?
Only for API products. The October 2025 restructuring explicitly permits OpenAI to develop non-API products with any cloud provider, including Amazon and CoreWeave. The $250 billion commitment ensures Microsoft captures API infrastructure revenue for years, but it does not prevent OpenAI from building consumer products, enterprise software, or hardware on competing platforms.
Why is OpenAI praising Amazon if it still needs Microsoft?
OpenAI’s public praise of Amazon signals strategic intent: to demonstrate that alternatives exist and that Microsoft’s restrictions are neither inevitable nor acceptable. By highlighting the Amazon relationship and infrastructure diversification, OpenAI is negotiating leverage for future renegotiations and signaling to the market that it will not be cornered by any single cloud provider. The partnership with Microsoft remains financially significant, but it is no longer exclusive or foundational in the way it was in 2023.
The OpenAI Microsoft partnership entered 2025 as the most valuable technology relationship in the world, but it is now defined by tension, constraint, and strategic hedging. OpenAI has the leverage to demand better terms, and it is using that leverage to break free from Microsoft’s cloud dominance. Microsoft, in turn, must accept looser control in exchange for maintaining API exclusivity and a significant financial stake. Neither party is happy, but both recognize that the old model—where Microsoft controlled OpenAI’s infrastructure—is no longer sustainable in a competitive AI market.
Edited by the All Things Geek team.
Source: TechRadar
