Microsoft Pushes Switch to In-House AI Models

Microsoft has revealed that it is gradually switching the AI embedded in its products such as Excel and Outlook from models provided by OpenAI and Anthropic to its in-house developed 'MAI' model. Tens of thousands of queries per week are already being processed through the MAI model, and the transition is proceeding in stages.

Leading this initiative is Mustafa Suleyman, who heads Microsoft's AI division. He has set a policy to ultimately bring the cost of external models to zero, and the switch to in-house models is positioned as part of this cost reduction strategy. With the rapid proliferation of generative AI, infrastructure costs for processing large-scale inquiries have become a common challenge for all companies, and Microsoft is no exception.

Until now, Microsoft has deployed Copilot (a general term for AI assistant functionality) centered on a deep partnership with OpenAI. Through large-scale investments in OpenAI, it has secured access to cutting-edge models and integrated them into business applications such as Excel, Outlook, and Teams. However, this policy shift indicates an intention to reduce such external dependence and gain control over cost structures in-house.

From a user perspective, there are important considerations regarding this switch. It remains unclear whether the MAI model possesses performance equivalent to OpenAI's and Anthropic's cutting-edge models. According to reports, there is a view that Copilot users may experience "performance degradation at the same price point," and concerns have been raised about the risk that cost reduction could impact user experience.

As an industry-wide trend, the 'internalization' of AI models is not unique to Microsoft. Many major technology companies are recognizing the dependency costs and competitive risks of external models and investing heavily in developing and deploying their own models. In this sense, Microsoft's current initiative represents an example of the industry entering a phase where the decision axis of "in-house development versus external procurement" when integrating AI into business is being reconsidered from a cost perspective.

The key points to watch going forward are how close the MAI model's performance can come to external models, and whether users will actually perceive changes in service quality. Since Copilot has been widely adopted for enterprise use, how performance changes impact operational efficiency in the field becomes a critical question for corporate users. It will be necessary to continue monitoring how Microsoft balances cost reduction with quality maintenance.