Tens of thousands of AI prompts in Excel and Outlook are now handled each week by Microsoft's own MAI models, not OpenAI or Anthropic, the first concrete sign the company's in-house AI has moved beyond testing into real production.
Bloomberg reported the shift, citing a person familiar with the work. The scale of MAI model usage inside Microsoft's workplace apps had not been previously disclosed. A Microsoft spokesperson declined to comment.
The move is incremental, OpenAI and Anthropic still handle most Copilot traffic, but it marks a deliberate unwinding of the dependency Microsoft spent years and billions building. The driver is simple: cost.
Mustafa Suleyman, Microsoft's AI model chief, said in June the company was trying to reduce spending on Anthropic by routing more work to MAI models. "We pay a lot of money to Anthropic, so our goal is to reduce and ultimately eliminate that cost," he said at the time.
Microsoft currently gets discounted access to OpenAI's technology through their long-standing partnership. That arrangement doesn't last forever, and Suleyman's team is preparing for a future where Microsoft pays market rates for external AI.
At its Build 2026 conference in June, Microsoft unveiled seven MAI models covering reasoning, coding, image generation, speech, and transcription. The lineup includes MAI-Thinking-1 for complex reasoning, MAI-Code-1-Flash for software development, MAI-Image-2.5, MAI-Voice-2, and MAI-Transcribe-1.5.
One model, tuned for consulting firm McKinsey, reportedly beat OpenAI's GPT-5.5 on cost efficiency by a factor of ten. The company said one of its coding models can match the programming capabilities of Anthropic's popular Opus 4.6 at a lower cost.
Microsoft is not ending its partnerships with OpenAI or Anthropic. Instead, it is turning Copilot and Azure AI into multi-model platforms that select the best model per task.
Customers may use MAI, GPT, or Claude depending on the workload, all inside Microsoft's ecosystem.
MAI models are already available in GitHub Copilot. Suleyman has said a Microsoft-built transcription model will roll out to Teams and other products in the coming months. The strategic logic traces back to Satya Nadella, who reportedly feared Microsoft becoming "the next IBM" if it leaned too hard on a single AI partner.
The 2025 renegotiation of the OpenAI deal freed Microsoft to build competing models while keeping a license to OpenAI's technology through 2032. That door cut both ways, OpenAI also gained the ability to sell through rivals like AWS.
Microsoft does not need MAI to top every leaderboard. Its reach into hundreds of millions of Office and Teams seats means even shifting a small slice of traffic to first-party models moves real money. The bill for outside AI is not trivial, and every feature Microsoft brings in-house trims it.
