Microsoft Replaces OpenAI and Anthropic Models with MAI in Office Apps

Bloomberg reported on July 7, 2026, that Microsoft has quietly begun routing tens of thousands of weekly AI prompts in Excel and Outlook through its own internally built MAI models—reducing its dependence on OpenAI and Anthropic in its flagship Office products.

Key Points:

• Microsoft unveiled its MAI model portfolio at Build conference, including a reasoning model and image, voice, and transcription systems—seven models in total.

• Microsoft AI chief Mustafa Suleiman stated explicitly: 'We pay a lot of money to Anthropic—so our goal is to reduce and ultimately eliminate that cost.'

• The shift is incremental: OpenAI and Anthropic still handle most Copilot traffic, but the direction of travel is clear and accelerating.

• Microsoft's 2025 renegotiation of its OpenAI partnership gave it the contractual freedom to build and deploy competing models, which it is now actively exercising.

In plain terms: Microsoft spent billions partnering with OpenAI, then decided to start making its own models in-house to save money—much like a restaurant that starts buying from its own farm instead of a supplier.

For any organization whose AI roadmap depends on Microsoft's Copilot products, understanding that the underlying models may change without notice is critical to planning. This is the week's clearest lesson: vendor diversification is not optional.

Why It Matters: Microsoft—the world's most committed AI investor—is building its own models to replace its partners. If Microsoft sees vendor concentration as a risk worth billions to eliminate, every enterprise should evaluate their own AI vendor dependencies.