Microsoft pushes model‑agnostic Azure

Microsoft’s latest AI positioning is less about a public break with OpenAI than about how Azure is being built to host, route and govern many models at once. On April 27, Microsoft said a reworked partnership with OpenAI would leave Microsoft as OpenAI’s primary cloud partner, while also making Microsoft’s license to OpenAI intellectual property non-exclusive and allowing OpenAI to serve products across other cloud providers. Two days later, Microsoft said its AI business had surpassed a $37 billion annual revenue run rate, up 123% year over year, and CNBC reported the company had more than 20 million paid seats for Microsoft 365 Copilot. Those figures gave investors fresh evidence that Microsoft’s AI growth story now rests on both its own software distribution and Azure infrastructure, not only on privileged access to one model provider. (blogs.microsoft.com) ### What changed in Microsoft’s relationship with OpenAI? Microsoft said on April 27 that OpenAI can now serve all of its products across any cloud provider, while Microsoft will continue to license OpenAI models and products through 2032 under a non-exclusive arrangement. Microsoft also said OpenAI products will ship first on Azure unless Microsoft “cannot and chooses not to support the necessary capabilities.” (microsoft.com) That language matters because it replaces a simpler exclusivity narrative with a more conditional one. Microsoft still keeps a privileged commercial relationship with OpenAI, but the company’s own statement shows Azure is no longer framed as the only place OpenAI can run. ### What does Azure now offer beyond OpenAI models? Microsoft’s Azure AI Foundry pages describe a model catalog that includes offerings from Azure OpenAI, Anthropic, DeepSeek, xAI, Meta, Mistral AI, Cohere, Hugging Face, NVIDIA and other providers. (blogs.microsoft.com) Microsoft says Foundry Models includes more than 11,000 models packaged for use in the platform. Microsoft Learn describes the Foundry catalog as a hub for discovering and using models from providers including Azure OpenAI, Mistral, Meta, Cohere, NVIDIA and Hugging Face. (blogs.microsoft.com) That product design gives enterprise customers a Microsoft-managed layer for evaluating and deploying multiple external and first-party models inside one Azure environment. ### Why are investors focusing on routing and control layers? (azure.microsoft.com) Microsoft’s own documentation now includes a “model router” service that routes prompts to what it calls the most suitable large language model in real time. Microsoft says the tool is designed to reduce cost and latency while maintaining comparable quality, and that it can invoke supported models through a single deployment. That is the clearest product-level evidence for the broader thesis around a model-agnostic Azure. (learn.microsoft.com) If customers can swap or combine models behind a common interface, the value shifts toward the control plane — routing, governance, deployment, cost management and telemetry — rather than toward dependence on a single upstream lab. That is an inference from Microsoft’s product design and partnership terms. (learn.microsoft.com) ### How do the revenue numbers fit this story? Microsoft said in its fiscal third-quarter earnings release that AI revenue had reached a $37 billion annual run rate, up 123% from a year earlier. CNBC separately reported that Microsoft 365 Copilot had passed 20 million paid commercial seats. Those numbers do not by themselves prove that Azure customers are choosing multi-model architectures. They do show that Microsoft is monetizing AI across both cloud infrastructure and application layers at scale, which gives the company room to sell Azure as a broader AI platform rather than as a resale channel for one partner’s models. (blogs.microsoft.com) That is an inference based on Microsoft’s reported figures and current product lineup. (microsoft.com) ### What should readers watch next? Azure AI Foundry’s catalog pages and Microsoft Learn documentation are the clearest places to track whether Microsoft keeps adding outside model providers, expands router support and deepens governance features. Microsoft’s next earnings report will show whether the company updates the $37 billion AI run-rate figure and the 20 million Copilot-seat count. (azure.microsoft.com) (microsoft.com)