Tech giant Microsoft is quietly building its own artificial intelligence reasoning models in a strategic move that could reshape its relationship with OpenAI, according to a recent report from The Information.
The Redmond-based company has begun developing a family of models internally called “MAI,” which reportedly perform nearly as well as leading models from OpenAI and Anthropic on standard industry benchmarks. This development marks a significant shift in Microsoft’s AI strategy, potentially reducing its heavy reliance on OpenAI’s technology.
Microsoft’s AI division, under the leadership of Mustafa Suleyman, is also training specialized reasoning models that employ chain-of-thought techniques.
These advanced models generate answers with intermediate reasoning abilities when solving complex problems, directly competing with similar offerings from OpenAI.
Despite being a major financial backer of OpenAI, Microsoft appears to be hedging its bets in the increasingly competitive AI landscape. The company has reportedly begun testing models from other AI companies including xAI, Meta, and DeepSeek as potential alternatives to OpenAI’s technology in its Copilot product.
MAI Takes Center Stage: Microsoft Plans to Replace GPT-4 in Copilot
When Microsoft launched 365 Copilot in 2023, one of its main selling points was integration with OpenAI’s powerful GPT-4 model.
Now, according to the report, Suleyman’s team is experimenting with replacing OpenAI’s models with Microsoft’s own MAI models in Copilot. These MAI models are described as significantly larger than Microsoft’s earlier Phi family of models.
This move aligns with reporting from Reuters last December, which revealed Microsoft was working to diversify the AI models powering its flagship Microsoft 365 Copilot product. This diversification strategy aims to reduce dependence on OpenAI’s technology while potentially lowering costs.
Microsoft has stated it plans to ship its MAI models in the second half of this year as an API, or application programming interface, to let other software developers bring those models into their applications, offering the prospect of an added stream of income to Microsoft in the very crowded AI space.
Microsoft’s In-House AI Push
In-house AI reasoning capacity is a strategic move for Microsoft. Microsoft spent billions on OpenAI and added its technology to many products, but creating internal AI models offers Microsoft more control over its AI strategy and fewer business risks of depending on a single outside partner.
This move comes in the backdrop of the rapid evolution of AI technology and increasing competition among technology giants. Google, Meta, and Amazon are among the companies that have all geared up their AI development initiatives, and smaller niche AI companies are still emerging with new ideas.
For Microsoft’s artificial intelligence software and the firms using them, it will eventually mean more options and even lower costs. But how this will affect Microsoft’s partnership with OpenAI, which has been profitable for both companies until now, remains to be seen.
Microsoft and OpenAI have not yet made a public announcement about the report. The public release timeline for Microsoft’s new models is unknown, though the report suggests we may see them materialize later this year.
While the AI landscape continues to evolve at an incredibly rapid pace, Microsoft’s move helps highlight the strategic importance big players in tech possess in maintaining their own AI tech stack, yet continuing to collaborate with specialized AI research centers.
