Microsoft Quietly Routes Excel and Outlook AI Prompts to Its Own MAI Models
Microsoft has begun silently redirecting AI prompts inside Excel and Outlook from OpenAI and Anthropic to its in-house MAI models, aiming to cut what AI chief Mustafa Suleyman called a very large bill to Anthropic.
Microsoft has quietly started routing tens of thousands of weekly AI prompts inside Excel and Outlook away from OpenAI and Anthropic toward its own in-house MAI models, Bloomberg reported July 7 — the first production deployment of Microsoft’s homegrown frontier-model stack.
No UI, no announcement
The switch happens entirely server-side. Users typing an AI-assisted formula in Excel or drafting a reply in Outlook see no visible change — same interface, same response format — but a growing share of those requests now hit Microsoft’s own models instead of GPT or Claude. External models still process the bulk of Copilot traffic across Microsoft’s product line; this is a deliberate, incremental expansion of the in-house share, not a wholesale replacement.
The quote that explains the strategy
Microsoft AI chief Mustafa Suleyman put the motive plainly: “We pay a lot of money to Anthropic — so our goal is to reduce and ultimately eliminate that cost.” That’s an unusually direct admission from a company that has spent years positioning its OpenAI partnership as the centerpiece of its AI strategy, and it confirms Microsoft has been paying Anthropic directly for Claude access to serve some Copilot workloads alongside its OpenAI dependency.
The MAI model lineup
Microsoft unveiled seven MAI (Microsoft AI) models at Build 2026, including MAI-Thinking-1, MAI-Code-1-Flash, MAI-Image-2.5, MAI-Voice-2, and MAI-Transcribe-1.5. MAI-Code-1-Flash is reportedly matching Anthropic’s Opus 4.6 on coding benchmarks at a lower operating cost — if that holds under real Excel/Outlook production load rather than curated benchmarks, it’s the model most likely to keep expanding its routing share first.
Why it matters
This is the clearest signal yet that Microsoft’s multi-year OpenAI partnership and its emerging in-house model program aren’t complementary — they’re on a collision course over the same budget line. Every prompt routed to MAI instead of GPT or Claude is a direct subtraction from what Microsoft pays OpenAI and Anthropic, and Suleyman’s comment makes clear that subtraction is the explicit goal, not a side effect. For a company that has bankrolled OpenAI’s compute for years, quietly building the muscle to walk away from that spend is the more important story than any single model benchmark.