Meta and Microsoft sharply cut internal Claude usage

Meta and Microsoft, two of Anthropic's largest enterprise customers, are moving to cut their employees' internal use of Claude, according to a report from The Information that 华尔街见闻 dates to October 5 and that Techmeme and The Decoder picked up on October 6.

At Microsoft, internal Claude spending had been projected at more than $1 billion a year; that budget has now been cut by over a third. The Decoder reports that the cloud division's monthly per-employee budget fell from $100,000 to roughly $10,000. Executives Scott Guthrie and Jay Parikh told employees to use in-house models, and The Decoder describes the push as steering staff toward GitHub Copilot and OpenAI models. Microsoft began reducing its reliance on Claude Code as early as May, aiming to curb employees' token consumption and promote internal models.

At Meta, the number of employees using Claude Code has dropped from about 60,000 earlier this year to roughly 30,000 — a decline of nearly 50% — while the company promotes its own Muse Code and MetaCode.

The pullback matters for Anthropic's revenue. According to 华尔街见闻, the company disclosed in a recent IPO prospectus that the two customers together contribute about 25% of its revenue. Both Meta and Microsoft say demand from external customers using Claude through its enterprise platform is still growing. The Decoder frames the shift as a strategic risk for Anthropic, casting the relationship as one that has gone from partnership to competition, and notes that large technology companies are moving faster to bring AI capabilities in-house. As of these reports, neither company has given a further timetable or end goal.