In order to optimize operational costs and build a more competitive in-house AI technology matrix, Microsoft has recently initiated a significant strategic adjustment within its core software products. According to insiders, Microsoft is gradually introducing its in-house MAI artificial intelligence model into Excel and Outlook, replacing the previously deeply relied-upon OpenAI and Anthropic models.

Currently, these two major office software applications have already seen tens of thousands of AI tasks completed independently by the MAI model each week. Although this accounts for a small portion of Microsoft's overall AI usage volume, it sends a clear signal: Microsoft has achieved a phased breakthrough in the development of its own high-level AI models. Mustafa Suleyman, head of Microsoft's AI models, publicly stated in June this year that the company is increasing the frequency of MAI usage to reduce high costs associated with third-party models.

For a long time, the "AI computing bill" behind office assistant products such as Copilot has been a major expense for Microsoft. Although Microsoft can enjoy certain technology purchase discounts due to its close partnership with OpenAI, this is not a sustainable solution. By developing in-house model technology, Microsoft aims to take control of the future AI landscape, avoiding being constrained by pricing strategies set by leading AI laboratories.

The current strategic layout goes beyond this. In addition to Excel and Outlook, Microsoft's MAI model has already been deployed in GitHub Copilot. According to Suleyman, in the coming months, the company will further expand the application scope of its in-house transcription models, gradually integrating this technology into Teams video conferencing and more product lines, thereby building a more self-controlled AI ecosystem.