Microsoft and OpenAI want you to know they’re still very much together—and still very serious about each other—even as OpenAI starts seeing other people in the cloud.

The new joint statement, published on February 27, is essentially a public reassurance campaign aimed at customers, developers, regulators, and investors watching OpenAI’s freshly announced multi‑billion‑dollar deals with Amazon and others. On paper, it doesn’t change anything. In practice, it tells you a lot about how this AI power couple plans to navigate an increasingly crowded (and expensive) AI ecosystem.

If you strip away the legalese, the first message is simple: the Microsoft–OpenAI partnership is still “strong and central.” Both sides repeat that line like a chorus. They frame the relationship as one of the “most consequential collaborations in technology,” a familiar callback to language they’ve used since at least 2025, when they renegotiated the deal and set out the “next chapter” of the partnership.

The second pillar of the statement, and arguably the most important for the industry, is intellectual property. Microsoft says nothing has shifted: it still holds an exclusive license to OpenAI’s models and IP, and that exclusivity covers the core technology behind systems like GPT-5, o-series models, and Frontier. This is why the wording matters when they talk about OpenAI’s “new partners.” The companies stress that collaborations like OpenAI’s new Amazon deal “were always contemplated” in the contract. In other words, OpenAI can work with others, but the underlying crown‑jewel IP remains tied to Microsoft.

Then comes the money question. The joint note confirms that the commercial and revenue‑share structure also stays exactly where it was left in the October 2025 agreement. Microsoft and OpenAI still share revenue from OpenAI products and from deals that involve other cloud providers, which explicitly includes partnerships with companies like Amazon. That means when OpenAI makes money via AWS‑powered offerings, Microsoft still participates economically via the revenue‑share mechanisms they agreed on earlier.

Cloud exclusivity is where things get a bit more technical but also more nuanced. The statement underlines that Azure remains the exclusive cloud for “stateless” OpenAI APIs—the classic call‑and‑response model where you send a prompt, get a response, and nothing persists across calls. Those APIs can be purchased either from OpenAI or from Microsoft, but under the hood, they all run on Azure. The companies go a step further and spell out that any stateless API usage generated by third‑party collaborations, including the new OpenAI–Amazon partnership, will still be hosted on Azure.

This is where the Amazon deal threads the needle. OpenAI and Amazon are building a “Stateful Runtime Environment” for agents inside Amazon Bedrock and positioning AWS as the exclusive third‑party cloud distributor for OpenAI’s Frontier platform, aimed at enterprises running teams of AI agents. That stateful layer—long‑lived context, memory, tools, and orchestration—sits on AWS, while the underlying stateless calls into OpenAI’s core models are still anchored in Azure per the Microsoft agreement. It’s a layered architecture: Azure owns the core model serving; AWS gets to own a rich runtime and distribution channel for a specific enterprise platform.

OpenAI’s own products remain firmly parked on Microsoft’s side of the fence. The companies reiterate that OpenAI’s first‑party offerings—including its new Frontier platform—continue to be hosted on Azure. For Microsoft, that’s crucial: it keeps the flagship OpenAI experiences, from ChatGPT‑class products to frontier‑model experimentation, inside its data centers, even as OpenAI builds custom experiences elsewhere.

One interesting line in the joint statement is about AGI—artificial general intelligence. The contractually defined threshold for AGI, and the process for deciding when it has been reached, remain unchanged. That may sound abstract, but it matters because Microsoft’s exclusivity on IP and Azure’s API role are explicitly tied to the pre‑AGI era in the 2025 agreement. By stressing that this definition hasn’t been quietly tweaked, they’re signalling stability around some of the most sensitive long‑term terms in the deal.

The document also nods to OpenAI’s growing appetite for compute diversity. OpenAI, it says, retains flexibility to “commit to additional compute elsewhere,” including massive infrastructure initiatives like the Stargate project. Stargate itself has been described in reporting and analysis as a multi‑phase, potentially hundreds‑of‑billions‑of‑dollars build‑out of AI data centers and supercomputers, spread over years and multiple phases beyond the current Microsoft–OpenAI supercomputer efforts. The joint message here is that pursuing mega‑projects with other partners doesn’t break the underlying Microsoft deal, as long as the defined IP and API rules are respected.

If you zoom out, this statement is almost defensive in tone, but intentionally so. OpenAI has just lined up an enormous strategic partnership with Amazon, including a promised $50 billion investment, expanded use of AWS Trainium chips, and deep integration into Amazon’s own products and the Bedrock platform. That kind of move naturally raises questions: is Microsoft being sidelined? Are we witnessing an AI custody battle between hyperscalers? The answer the statement offers is: not really. Microsoft is still the primary home of OpenAI’s frontier models and the sole provider of stateless APIs, and it still benefits directly when OpenAI expands elsewhere.

At the same time, the companies are honest about the fact that the partnership was consciously designed to let each side “pursue new opportunities independently” while still collaborating. For OpenAI, that means it can cut deals with Amazon, NVIDIA, SoftBank, Oracle, and others to secure compute, distribution, and capital. For Microsoft, it means it keeps a privileged technical and commercial position—exclusive IP rights, Azure as the backbone of model delivery, and a share of revenues—even as OpenAI’s ecosystem broadens.

For developers and customers, the practical takeaway is less dramatic than the headlines. If you’re already using OpenAI APIs or Azure OpenAI Service, nothing changes: same model access, same cloud, same economic structure. If you’re an AWS-first shop, the Amazon deal means you’ll soon see richer OpenAI‑powered agent runtimes and Frontier‑based tools on Bedrock, but those experiences will quietly route back to Azure for stateless model calls under the hood.

The politics are more subtle. Microsoft wants the world to see itself as OpenAI’s anchor partner, not its ex. OpenAI wants the freedom to tap every major cloud and capital source on the planet without triggering panic about its core alliances. This joint statement tries to square that circle: OpenAI is going multi‑cloud and multi‑partner at the edges, while the center of gravity—core IP, model serving, and a big chunk of the economics—still orbits around Microsoft.