OpenAI and Microsoft restate that Azure hosts all stateless OpenAI model calls

What the statement actually pins down

The statement exists to answer a single question raised by OpenAI's simultaneous announcements of new funding and new partners: does any of it change the Microsoft relationship? The answer given is no. The two companies say nothing about the day's announcements alters the terms shared in their October 2025 joint blog.

Four things are declared unchanged: Microsoft's exclusive license and access to IP across OpenAI models and products, the revenue share arrangement, the AGI definition and the process for determining whether it has been achieved, and Azure's status as exclusive cloud provider of stateless OpenAI APIs.

The document is notable for what it confirms rather than what it introduces. It names the OpenAI–Amazon partnership specifically and frames it as something the agreements always contemplated, rather than a departure from them.

The stateless API boundary is the load-bearing clause

For anyone consuming OpenAI models through an application, the operative detail is the definition of what Azure exclusively hosts: stateless APIs that provide access to OpenAI's models and IP. The statement says these can be purchased from Microsoft or directly from OpenAI, but the underlying inference runs on Azure either way.

The clarification extends to new partners. Any stateless API calls to OpenAI models that result from a collaboration between OpenAI and a third party — Amazon is named explicitly — would be hosted on Azure. So a new cloud partnership does not mean OpenAI model calls move to that partner's infrastructure.

Any stateless API calls to OpenAI models that result from a collaboration between OpenAI and any third party—including Amazon—would be hosted on Azure.Montana Labs

The word doing the work is 'stateless.' The statement draws the exclusivity around API calls that carry no session state, which leaves room to read where the boundary lies for stateful workloads without saying so directly.

Frontier and OpenAI's own products stay on Azure

The statement separately commits OpenAI's first-party products to Azure, naming Frontier among them. This is a distinct point from the API exclusivity: it covers what OpenAI ships to its own customers, not only what other builders call through an interface.

At the same time, the companies describe headroom for OpenAI to commit additional compute elsewhere, citing large-scale infrastructure initiatives such as Stargate. The picture is a fixed floor — model APIs and first-party products on Azure — with expansion capacity allowed above it.

The implication: new partners do not move OpenAI inference off Azure

For teams building on OpenAI, the practical takeaway from this statement is that the roster of OpenAI's cloud and capital partners can grow without changing where the model runs. A collaboration announced with Amazon or another provider does not relocate stateless OpenAI model calls; those stay on Azure.

That means procurement and latency assumptions tied to Azure hosting hold even as OpenAI diversifies its funding and compute. The choice the statement preserves for developers is where to buy the API — Microsoft or OpenAI directly — not which infrastructure serves it.