The deal that lit the fuse was announced quietly on February 27. Amazon Web Services would invest $50 billion in OpenAI — becoming the "exclusive third-party cloud provider" for Frontier, OpenAI's new enterprise platform for AI agents. Andy Jassy and Sam Altman were all smiles.
Microsoft, which had already invested over $13 billion in OpenAI, was not smiling. It had committed to buying 250 billion dollars of Azure cloud services over six years — and the Amazon deal threatened the return on both.
Within days, a person close to Microsoft's position told the Financial Times exactly how the company felt about it: "We know our contract. We will sue them if they breach it. If Amazon and OpenAI want to take a bet on the creativity of their contractual lawyers, I would back us, not them."
A cold war had become a legal threat. And both sides were digging in.
The Contract at the Center of Everything
To understand why Microsoft is furious, you need to understand what it believed it had locked in.
When OpenAI completed its conversion from a nonprofit to a public benefit corporation in October 2025, Microsoft struck a new deal. It received a 27% equity stake in OpenAI, worth roughly $135 billion at the time. In exchange, OpenAI committed to buying 250 billion dollars of Azure cloud services over an estimated six years. Microsoft retained exclusive IP rights to OpenAI's models through 2032.
Most critically, one clause survived every renegotiation: all stateless API calls to OpenAI's models — the basic queries that power ChatGPT, Copilot, and thousands of third-party applications — must be routed exclusively through Microsoft's Azure platform. No Azure, no access.
That clause was Microsoft's insurance policy. No matter how big OpenAI grew, Azure would sit at the center of every AI workload.
Then Amazon's lawyers got creative.
The Workaround That Started a War
OpenAI and Amazon didn't just sign a distribution deal. They co-designed a technical architecture to go with it.
The argument goes like this: Microsoft's exclusivity covers "stateless" API calls — requests where the AI model processes a single question and forgets it immediately after. But OpenAI's new Frontier platform runs AI agents, which are different. Agents maintain memory across dozens or hundreds of interactions. They book your travel, manage your inbox, run your sales pipeline. They need context that persists.
So OpenAI and Amazon built what they're calling a "Stateful Runtime Environment," or SRE, hosted on Amazon Bedrock. Because the SRE maintains persistent memory — making it "stateful" rather than "stateless" — OpenAI and Amazon argue it doesn't touch Microsoft's exclusivity clause at all. Different category, different rules.
Microsoft's response was immediate: that argument is sophistry. You cannot build a stateful enterprise AI system without underlying stateless API calls at some level. The SRE is built on top of the same models Microsoft has exclusive rights to. Running them on AWS, under any label, violates the agreement.
"If Amazon and OpenAI want to take a bet on the creativity of their contractual lawyers," the Microsoft source said, "I would back us, not them."
The three companies are currently in talks to resolve the dispute before Frontier's public launch. But no resolution has been announced.
What OpenAI Actually Wants
The legal dispute is the symptom. The disease is something larger: OpenAI wants to be independent.
When Amazon agreed to invest $50 billion, it wasn't just buying a stake in a promising company. It was buying preferential access to the most valuable AI platform in the world for enterprises. OpenAI and AWS expanded their existing cloud agreement — worth 38 billion dollars — extending it by another 100 billion dollars over eight years. OpenAI committed to consuming roughly 2 gigawatts of Amazon's Trainium AI chips — hardware that directly competes with Microsoft's Azure AI infrastructure.
The Frontier platform itself is the clearest sign of where OpenAI is going. Launched in February 2026, Frontier is an enterprise system for deploying teams of AI agents inside large organizations. Think of it as the operating system for how companies will run AI workers. HP, Intuit, Oracle, State Farm, Uber, and Thermo Fisher are already in early access. BBVA, Cisco, and T-Mobile are piloting it.
That platform, running on AWS infrastructure, sits at the heart of what enterprise AI will look like for the next decade. Losing it to Amazon — after betting everything on being OpenAI's gateway — would be a serious blow to Microsoft's cloud ambitions.
Sam Altman has acknowledged tensions publicly, telling a podcast: "Obviously in any deep partnership, there are points of tension and we certainly have those. But on the whole, it's been really wonderfully good for both companies."
That statement was made before the lawsuit threat.
Microsoft's Two-Front Response
Microsoft isn't just threatening lawyers. It's building its own escape route.
In early 2026, Microsoft's AI chief Mustafa Suleyman announced the company's ambition for "true self-sufficiency" in AI — a direct signal that relying on OpenAI forever was never the plan. Microsoft has been quietly developing its own MAI (Microsoft AI) family of foundation models, including MAI-1-preview and MAI-Voice-1. The October 2025 restructuring of the OpenAI partnership explicitly permitted Microsoft to pursue frontier-scale AI models independently or with other partners — a clause that Microsoft exercised quickly.
Microsoft also built the Maia 200 AI accelerator chip and is expanding the Fairwater data center network to train its own models at scale.
The message is clear: Microsoft is not going to let OpenAI walk out the door without a fight. But it's also not betting its entire AI future on a single partner anymore.
The Other Side of the Argument
Not everyone thinks Microsoft has the stronger position.
OpenAI and Amazon's lawyers have constructed the stateful/stateless distinction carefully. There is a legitimate technical argument that persistent agentic workflows are architecturally different from single-shot API calls. If that argument holds in court, Microsoft's exclusivity clause may be narrower than it thought.
Some analysts also point out that Microsoft's $250 billion Azure commitment from OpenAI remains intact regardless of how this dispute resolves — OpenAI is still going to spend enormous sums on Azure for its own core products, including ChatGPT. The Frontier deal with AWS doesn't erase that.
There's also a competitive reality that cuts against Microsoft: enterprise customers are demanding multi-cloud flexibility. A clause that forces all AI workloads through a single provider, no matter how capable, creates vendor lock-in that large companies increasingly refuse to accept. OpenAI's argument is essentially that letting AWS run Frontier makes the product more deployable — and therefore more valuable — for everyone.
Amazon, for its part, has been strengthening its position. The launch of the Nova 2 Pro model and the wide availability of Trainium 3 processors have allowed AWS to offer AI training and inference at roughly 40% lower costs than competitors. AWS has been winning back enterprise AI workloads it lost to Azure when GPT-4 made OpenAI synonymous with Microsoft's cloud.
The Bottom Line
Microsoft invested over 13 billion dollars in OpenAI when no one else would. It got Azure exclusivity as the return on that bet. Now OpenAI — valued at $730 billion and backed by Amazon, NVIDIA, and SoftBank — is using a technical distinction between "stateful" and "stateless" compute to route 50 billion dollars' worth of cloud business to its biggest rival.
Microsoft's response is what you'd expect from a company that feels a founding partner is walking away: threaten litigation, accelerate in-house alternatives, and remind everyone publicly that the contract still exists.
What happens next matters for every enterprise buying AI. If OpenAI wins the legal argument, it establishes that AI platforms can distribute through multiple cloud providers regardless of legacy exclusivity deals — a green light for the multi-cloud AI era. If Microsoft wins, it may preserve Azure's position but further poison a partnership that already shows visible cracks.
Either way, the era of Microsoft and OpenAI presenting a unified front to the market is over. The question now is how messy the divorce gets.
Sources
- Microsoft is threatening to sue OpenAI over its $50 billion Amazon deal (Neowin, March 18, 2026)
- Microsoft Considers Suing to Halt Amazon-OpenAI Cloud Deal (PYMNTS, March 18, 2026)
- Filings: How Amazon's $50B OpenAI deal actually works (GeekWire, 2026)
- OpenAI Secures AWS Distribution for Frontier Platform in $110B Multi-Cloud Deal (InfoQ, March 2026)
- OpenAI completes for-profit move, Microsoft given 27% stake and $250bn Azure contract (Data Center Dynamics, 2025)
- The next chapter of the Microsoft-OpenAI partnership (OpenAI, October 2025)
- Microsoft–OpenAI Cloud Dispute: Tensions Rise Over $50 Billion Amazon AI Deal (Influencer Magazine, March 2026)
- The Microsoft-OpenAI Files: Internal documents reveal the realities of AI's defining alliance (GeekWire, January 2026)
- Microsoft's AI Chief Targets AI Self-Sufficiency and OpenAI Independence (WinBuzzer, February 2026)
- The Great Cloud Divorce: Microsoft and OpenAI Clash as Amazon Charges into the AI Supremacy Vacuum (FinancialContent, March 18, 2026)
- OpenAI and Amazon announce strategic partnership (OpenAI, February 2026)
- OpenAI Sam Altman breaks silence on Microsoft feud (Windows Central, 2026)