Microsoft is beginning a large-scale transition to using its own internal MAI (Microsoft AI) models instead of third-party solutions from OpenAI and Anthropic. These changes will affect products such as Excel, Outlook, and GitHub Copilot, as well as transcription features in Microsoft Teams.
What Happened
Microsoft is implementing its own specialized MAI models into a number of its key services, including the office applications Excel and Outlook, as well as GitHub Copilot. Additionally, internal models are planned for transcription tasks in Microsoft Teams. This transition is aimed at radically reducing operating expenses associated with paying for tokens to third-party providers and decreasing dependency on external AI labs.
Context
Previously, Microsoft actively used OpenAI and Anthropic technologies to power its AI functions. The current strategy involves moving away from using general-purpose, but expensive, LLMs toward specialized models optimized for specific tasks (domain-specific models), which allows for better control over both the software stack and the computational cost of inference.
Why It Matters for the Industry
For the industry, this is a significant signal of major players' drive toward full vertical integration of the AI stack—from managing computing power to creating specialized models and applications. Microsoft is setting a trend toward creating efficient, highly specialized solutions instead of relying on third-party universal APIs, which may force competitors to also accelerate the development of their own internal models to ensure autonomy.
Why It Matters for Users
Users may notice increased speed in Copilot functions within office applications and reduced latency due to tighter integration of models with Microsoft's infrastructure. Furthermore, reducing the cost of using AI for the company itself may help stabilize subscription costs for end customers.
Sources
Author
Look at AI, Editorial Team