Microsoft moves toward model independence
Reporting indicates Microsoft has launched its own AI models as part of a push to reduce dependence on any single external model supplier and deepen its platform control. The coverage frames the move as part of a broader incumbent strategy to buy optionality and protect margins around cloud and productivity offerings. (tradingkey.com)
Microsoft has started shipping more of its own artificial intelligence models, giving customers Microsoft-built options alongside OpenAI inside its cloud platform. (geekwire.com) On April 2, Microsoft announced MAI-Transcribe-1 and opened MAI-Voice-1 and MAI-Image-2 for broad commercial use through Microsoft Foundry and MAI Playground. GeekWire reported the release followed a March reorganization that pushed Microsoft Artificial Intelligence chief Mustafa Suleyman toward frontier model work. (geekwire.com) Suleyman told The Verge, as quoted by GeekWire, that the new transcription model runs at about half the graphics processing unit cost of other state-of-the-art systems, and he told VentureBeat the team plans to build a frontier large language model that can be “completely independent” if needed. (geekwire.com) A model marketplace is becoming part of Microsoft’s pitch. Microsoft says Foundry Models lets customers compare and switch among more than 11,000 models, and its documentation says the catalog includes Microsoft, OpenAI, Anthropic, Meta, DeepSeek, xAI and others. (azure.microsoft.com, learn.microsoft.com) Microsoft has also built software that picks among models in real time. Its model router is described in Microsoft documentation as a deployable chat model that routes prompts to the most suitable large language model to cut cost and latency while keeping similar quality. (learn.microsoft.com) That is a change from the tighter structure Microsoft described in January 2025. At that point, Microsoft said key parts of its OpenAI deal would run through 2030, including access to OpenAI intellectual property, revenue sharing, and exclusivity for OpenAI application programming interfaces on Azure, while new capacity shifted to a right-of-first-refusal model. (blogs.microsoft.com) The partnership was revised again on October 28, 2025. Microsoft said OpenAI would remain its frontier model partner and Azure would keep application programming interface exclusivity until artificial general intelligence, but Microsoft also said the new agreement let each company “independently continue advancing innovation and growth” and that Microsoft could independently pursue artificial general intelligence alone or with third parties. (blogs.microsoft.com) Microsoft’s financial filings show why owning more of the stack matters. In fiscal second-quarter 2026 results released January 28, Microsoft said gross margin pressure came from continued investment in artificial intelligence infrastructure and growing artificial intelligence product usage, even as revenue rose 17 percent. (microsoft.com) OpenAI is still central to Microsoft’s business. Microsoft said in March 2026 that “nothing” in OpenAI’s latest funding announcements changed the October 2025 terms and that the partnership remained “strong and central.” (openai.com) So the shift is not a clean break. Microsoft is keeping OpenAI as a core supplier while building its own models, adding rival models to Foundry, and giving customers tools to swap among them inside Azure. (openai.com, azure.microsoft.com, geekwire.com)
