OpenAI and Microsoft Tear Up Their Exclusivity Pact, Unlocking $50B Amazon Deal

The document that has quietly governed the AI industry since 2019 was rewritten on Monday. Microsoft and OpenAI announced a sweeping renegotiation of their partnership, replacing the exclusive licensing arrangement that had bound them together — and generated enormous friction over the past several months — with a new framework designed to let OpenAI grow beyond any single cloud provider.

The practical effect was immediate. Amazon CEO Andy Jassy posted on X within hours: “We’re excited to make OpenAI’s models available directly to customers on Bedrock in the coming weeks.” OpenAI, for its part, said it would expand its existing $38 billion agreement with AWS by an additional $100 billion over the next eight years, creating one of the largest cloud infrastructure commitments in the industry’s history.

How the Old Deal Broke Down

The original Microsoft-OpenAI partnership, first struck in 2019 and significantly expanded in 2023 with a $10 billion investment, gave Microsoft an exclusive license to deploy OpenAI’s models and technology — along with a provision tied directly to whether OpenAI achieved artificial general intelligence (AGI). Under the original terms, the commercial arrangement would have changed materially the moment OpenAI’s own leadership declared that AGI had been achieved, a clause Microsoft’s lawyers reportedly viewed as an unusually subjective and potentially catastrophic trigger.

The trouble began in February 2026, when Amazon announced an investment of up to $50 billion in OpenAI. The deal included provisions giving AWS exclusive rights to serve OpenAI’s new frontier agent runtime — the underlying infrastructure that lets AI agents persist context and carry out multi-step tasks over extended periods. Microsoft objected, arguing that granting such exclusivity to a competitor violated the terms of their existing agreement. In March, the Financial Times reported that Microsoft was weighing legal action.

The standoff created a genuine crisis for OpenAI. With its IPO on the horizon and its valuation recently reaching the $300 billion range, the prospect of protracted litigation with its largest cloud partner and shareholder was untenable. Months of negotiation followed.

What Changed

Under the new agreement, announced jointly by both companies on April 27, the core changes are:

Non-exclusive license through 2032. Microsoft retains the right to use OpenAI’s models and technology for its products and services, but that right is no longer exclusive. OpenAI is free to license the same technology to Amazon, Google, or any other partner without restriction.

Revenue share capped and restructured. Microsoft will stop paying OpenAI a revenue share on products built with OpenAI technology. OpenAI will continue paying Microsoft a revenue share on its own revenue through 2030, but that obligation is now capped at a fixed dollar ceiling — eliminating the open-ended liability OpenAI had previously carried. The AGI milestone trigger that would have automatically restructured the deal has been removed entirely.

Microsoft remains primary cloud partner. Despite the non-exclusive language, the two companies specified that Microsoft Azure will remain the “primary” provider for OpenAI’s cloud infrastructure for the duration of the agreement. Microsoft’s 27% ownership stake in OpenAI’s for-profit entity is unchanged.

AWS exclusivity for Frontier runtime. In exchange for resolving Microsoft’s objections, OpenAI confirmed that AWS will have exclusive rights to host its stateful agent runtime — the technology that allows AI agents to remember long-running tasks, maintain context across sessions, and operate with greater autonomy. Amazon Bedrock, the AWS service that aggregates AI models for enterprise customers, will add OpenAI models “in the coming weeks,” according to Jassy’s statement.

Why This Matters for the Broader AI Market

The restructured partnership reflects a new reality: OpenAI is no longer a startup dependent on a single corporate benefactor. With annualized revenue approaching $25 billion and an IPO reportedly under active preparation for late 2026, the company has the leverage to negotiate as a peer with its partners rather than a dependent.

The deal also resets the competitive dynamics between the three largest cloud providers at a critical moment. Microsoft had used its exclusive relationship with OpenAI as a differentiator for Azure — a reason for enterprises considering cloud providers to lean toward Microsoft’s infrastructure. That advantage erodes as OpenAI models become available natively on AWS Bedrock and, presumably, on Google Cloud through OpenAI’s existing research relationship with Google. Enterprise customers gain optionality; Microsoft loses a moat.

For Amazon, the expanded AWS relationship is a significant trophy. AWS has lagged behind Azure and Google Cloud in AI-native workloads despite being the overall leader in cloud infrastructure revenue. Having direct access to OpenAI’s frontier models — and exclusive access to the stateful agent runtime that is increasingly central to enterprise AI deployments — gives AWS a meaningful capability upgrade at precisely the moment when enterprise AI spending is accelerating toward $700 billion in annual capex across the hyperscalers.

The removal of the AGI clause may prove to be the most consequential change in the long run. As AI systems become more capable, the question of what constitutes AGI has grown from a philosophical curiosity into a live legal risk. By decoupling the commercial relationship from any milestone that OpenAI’s own leadership might declare, both companies have reduced one of the most unpredictable variables in their partnership.

OpenAI’s Path to Independence

The renegotiated deal is one of several moves OpenAI has made in the past year to reduce its dependency on any single partner. The company has struck cloud deals with Oracle and SoftBank as part of the Stargate initiative, built direct infrastructure relationships with sovereign wealth funds in the Middle East, and is reportedly in conversations with additional hyperscalers about future model distribution agreements.

The IPO preparation — confirmed by multiple sources familiar with the process as targeting late 2026 — will require OpenAI to present a capital structure that can be explained to public market investors without a labyrinthine set of exclusive provisions with its biggest competitor. Monday’s deal cleans up the most problematic element of that narrative.

What remains to be seen is whether the partnership’s restructuring changes the pace at which Microsoft integrates OpenAI models into its own products. Azure AI and Microsoft 365 Copilot have been deeply built on OpenAI technology; Satya Nadella has described OpenAI as Microsoft’s most important strategic partner. Whether that description survives a transition from exclusive to non-exclusive is a question that Microsoft’s Q3 earnings call on April 29 will likely begin to answer.