clauding.de
DE
• 2 min read AI-generated

Microsoft Kicks OpenAI and Anthropic Out of Excel and Outlook

Copy article as Markdown

Microsoft is now handling tens of thousands of AI prompts a week in its Office apps with its own MAI models. The stated goal: cut the cost of outside models — and eventually eliminate it entirely.

Microsoft Anthropic OpenAI Business
Featured image for "Microsoft Kicks OpenAI and Anthropic Out of Excel and Outlook"

For years Microsoft was OpenAI’s biggest distribution channel — and lately an important Anthropic customer too. Now the company is pulling the plug in some of its most-used apps. According to a Bloomberg report, Microsoft is increasingly replacing OpenAI’s and Anthropic’s models in Excel and Outlook with its own, internally built MAI models.

What’s happening

Tens of thousands of AI prompts in those two applications are now completed each week with Microsoft’s own models. Previously, Excel and Outlook leaned more heavily on outside providers. The reason is unglamorous: money.

At its Build conference in June, Microsoft unveiled seven new AI models. One of them, the company claims, matches the coding ability of a prior-generation but still very popular Anthropic model — Opus 4.6 — at a substantially lower cost.

A Microsoft staffer summed up the strategy unusually bluntly: “We pay a lot of money to Anthropic — so our goal is to reduce and ultimately eliminate that cost.”

Why this is more than a cost footnote

That one line says a lot about the state of the industry. The big platforms don’t want to depend forever on a handful of labs controlling the price of access to frontier models. Whoever owns distribution — and with Office, Microsoft owns one of the largest distributions on earth — wants to own the model too.

For Anthropic it’s a double-edged signal. On one hand, it shows how deeply Claude is embedded in enterprise workflows when Microsoft actively names replacement as a goal. On the other: every prompt that shifts from Claude to MAI is revenue walking out the door.

My take

This isn’t an overnight break, it’s a slow melt. And it fits a pattern I’ve covered here several times in recent weeks: the cost of frontier models has become a real issue — Tesla capping its AI spending, companies switching to cheaper Chinese models, and now Microsoft building its own alternatives for high-volume use.

The interesting question isn’t whether Microsoft’s models can keep up. For autocomplete in an email, they don’t need to beat Opus — they just need to be good enough. And “good enough at a fraction of the price” is a dangerous combination for anyone who lives off API revenue.

Sources: