Microsoft rolls its own AI models

Reports say Microsoft has launched its own AI models and is positioning them as a way to reduce dependence on OpenAI, a move investors and analysts are watching amid debates over cloud and AI economics. The coverage frames the shift as part of Microsoft’s broader strategy to capture more AI infrastructure and customer relationships through Azure and Copilot investments. ( )

Microsoft has started selling its own artificial intelligence models through Azure, a step that reduces how much it needs to rely on OpenAI. (microsoft.ai) On April 2, Microsoft said three in-house models were available in Foundry: MAI-Transcribe-1 for speech-to-text, MAI-Voice-1 for synthetic speech, and MAI-Image-2 for image generation. TechCrunch reported the systems were built by Microsoft’s MAI Superintelligence team, which the company formed in November 2025 under Microsoft AI chief executive Mustafa Suleyman. (microsoft.ai; techcrunch.com) Microsoft said MAI-Transcribe-1 posted a 3.9 percent average word error rate in its benchmark chart, compared with 4.2 percent for OpenAI’s GPT-Transcribe and 4.3 percent for ElevenLabs Scribe v2. Microsoft also said MAI-Voice-1 can generate 60 seconds of audio in one second and is being offered through Foundry with custom-voice controls. (microsoft.ai) Azure AI Foundry is Microsoft’s storefront and toolkit for developers building artificial intelligence applications, so putting MAI models there lets Microsoft sell the model, the cloud capacity, and the surrounding software in one stack. Microsoft said at Build 2025 that Foundry was its “app and agent factory” for customers that need models, tools, and governance in one service. (azure.microsoft.com; microsoft.ai) The move comes after Microsoft and OpenAI rewrote parts of their relationship in late 2025. Microsoft said on October 28, 2025 that OpenAI had committed to buy an incremental $250 billion of Azure services and that Microsoft would no longer hold a right of first refusal to be OpenAI’s compute provider. (blogs.microsoft.com) OpenAI and Microsoft said in a joint statement last month that their partnership remained “strong and central” and that Microsoft still had exclusive license and access rights across OpenAI models and products. That leaves Microsoft in the position of both backing OpenAI and building competing model families of its own. (openai.com; blogs.microsoft.com) Mustafa Suleyman told Bloomberg, in an interview published last week and summarized by multiple outlets, that Microsoft is aiming for “state-of-the-art” models across several data types. Forbes said the new releases were positioned as a hedge against Microsoft’s dependence on OpenAI for key model capabilities. (forbes.com; msn.com) Analysts have been debating whether Microsoft captures enough of the economics from artificial intelligence when OpenAI supplies the flagship models and Copilot demand has been uneven. The company’s answer is increasingly to own more of the infrastructure layer through Azure and more of the application layer through Copilot and Foundry at the same time. (azure.microsoft.com; forbes.com) The immediate test is not whether Microsoft leaves OpenAI, but whether customers buy enough MAI services to make a dual-track strategy pay off. After years of funding the most important model maker in the field, Microsoft is now trying to prove it can be one too. (openai.com; microsoft.ai)

Get your own daily briefing

Scout delivers personalized news, insights, and conversations tailored to your role and industry.

Download on the App Store

Shared from Scout - Be the smartest in the room.