Microsoft, a major investor in OpenAI, appears to be building its own AI reasoning models, potentially setting the stage for direct competition with its partner.
The tech giant is reportedly developing a suite of advanced AI reasoning models internally called MAI. This could be a strategic effort to diminish its dependence on OpenAI’s technology, according to multiple reports. These new models are being designed to tackle complicated problem-solving tasks. They are already undergoing testing as potential replacements for OpenAI’s systems within Microsoft products, including Copilot (the AI assistant integrated into Microsoft 365, Microsoft Office), Bing Search, and GitHub.
Mustafa Suleyman, Microsoft’s AI Chief, is leading the development of the MAI models.
Since 2019, Microsoft and OpenAI have worked together closely, with Microsoft having invested over $13 billion in the ChatGPT maker. However, there are signs of a shift in the partnership. Under the leadership of Mustafa Suleyman, co-founder of DeepMind and Inflection, Microsoft’s AI division is expanding its internal capabilities. The Information reports that Suleyman’s team has trained a family of AI models that perform almost as well as OpenAI’s most advanced models, including the o1 reasoning model. These MAI models are said to excel in reasoning tasks, where AI not only provides answers but also explains its thought process. This feature could be especially useful in fields like healthcare, finance, manufacturing, and scientific research.
Microsoft is also exploring models from other companies. They are evaluating models from Elon Musk’s xAI, Meta, Anthropic, and DeepSeek as potential backups or alternatives for OpenAI’s models in Copilot. This multi-model strategy may signal that Microsoft is heading towards a more diverse AI infrastructure rather than relying solely on OpenAI.
The MAI models are being tested as potential replacements for OpenAI’s models within Copilot. If successful, Microsoft could release them through a public API later this year, allowing third-party developers to integrate them into their applications, much like how OpenAI’s models are used now.
Despite these developments, the partnership between Microsoft and OpenAI remains in place. Recent reports suggest some tension, particularly after OpenAI reportedly declined to share technical details about its latest reasoning models with Microsoft. The companies renegotiated their agreement in January, allowing OpenAI to use cloud services from providers other than Microsoft’s Azure, provided Microsoft does not want the business. This change reflects a more flexible approach, even as Microsoft continues to invest in its own AI capabilities.
Microsoft CFO Amy Hood recently emphasized the strength of the partnership: “We’re both successful when each of us is successful,” she said at a recent investor event. With Microsoft already offering smaller models like Phi and testing the waters with MAI, the company is positioning itself for a more autonomous role in the AI sector. If successful, this could reduce Microsoft’s dependence on OpenAI while still benefiting from the partnership. Neither Microsoft nor OpenAI has publicly commented on these reports.
