Microsoft shifts model strategy
Reports say Microsoft is rolling out its own AI models and reducing dependence on OpenAI, marking a strategic move in its AI product stack. The coverage frames this as a broader change in where major cloud vendors are placing their model bets and how they might repackage or price AI features. (tradingkey.com)
Microsoft is no longer relying only on OpenAI for its artificial intelligence products. On April 2, Microsoft released three in-house models through Microsoft Foundry and its MAI Playground. (microsoft.ai) The new lineup includes MAI-Transcribe-1 for speech-to-text, MAI-Voice-1 for synthetic speech, and MAI-Image-2 for image generation. Microsoft said MAI-Transcribe-1 works across 25 widely used languages and runs 2.5 times faster than its existing Azure Fast transcription offering. (microsoft.ai) Microsoft also attached pricing to the launch. TechCrunch reported MAI-Transcribe-1 starts at $0.36 per hour, MAI-Voice-1 at $22 per 1 million characters, and MAI-Image-2 at $5 per 1 million text-input tokens plus $33 per 1 million image-output tokens. (techcrunch.com) This does not mean Microsoft has ended its OpenAI alliance. Microsoft and OpenAI said on October 28, 2025 that OpenAI remains Microsoft’s “frontier model partner” and that Azure keeps API exclusivity until artificial general intelligence, or AGI, under their revised agreement. (blogs.microsoft.com) The agreement also gave Microsoft more room to build on its own. Microsoft said it can now “independently pursue AGI alone or in partnership with third parties,” while keeping rights tied to models and products through 2032 under the updated deal. (openai.com) That shift changes how Microsoft can package artificial intelligence inside Azure. In Microsoft Foundry’s model catalog, the company says customers can test and switch between models inside one system, with unified billing, governance, and reserved-capacity pricing. (ai.azure.com) The internal team behind the new models is also new. TechCrunch reported the MAI Superintelligence team was formed in November 2025 under Microsoft AI chief executive Mustafa Suleyman, and these April releases are its first public model launch. (techcrunch.com) Microsoft is still selling OpenAI models alongside its own. The Foundry catalog includes OpenAI systems such as o1-preview, which Microsoft describes as a model customers can purchase and manage directly through Azure. (ai.azure.com) Suleyman has said the two tracks can coexist. TechCrunch reported that he reaffirmed Microsoft’s commitment to OpenAI even as the company uses the renegotiated partnership to build a separate in-house model stack. (techcrunch.com) The practical change is that Microsoft now has more control over what sits underneath Copilot, Bing, PowerPoint, Azure Speech, and developer tools. The company said the same MAI models are already powering some Microsoft products, and more Microsoft-built models are coming to Foundry and Microsoft services. (microsoft.ai)
