Meta and Microsoft cut internal use of Claude
The Decoder
Meta and Microsoft are cutting back on Claude inside their companies. They’re pushing their own AI tools instead, and that’s a bad sign for Anthropic.
Based on reporting by The Decoder — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
Two of Anthropic’s biggest enterprise customers are dialing down their own use of Claude, and the shift says a lot about how quickly “partner” can turn into “competitor.” Meta and Microsoft are both steering employees toward in-house or allied tools instead of Anthropic’s models.
At Microsoft, the pullback was steep. The company had been projecting more than $1 billion a year in Claude spending, according to The Information, but executives Scott Guthrie and Jay Parikh told employees to use GitHub Copilot and OpenAI’s models instead. Internal Claude spending fell by more than a third, and in the cloud division the monthly per-employee budget dropped from $100,000 to about $10,000.
Meta’s cut was also sharp. Claude Code usage fell from roughly 60,000 users to 30,000, a drop tied partly to layoffs and partly to Meta’s own push behind Muse Code and MetaCode. In one 28-day stretch, Meta spent more than $105 million on Claude Code alone. The company had already started AI cost-cutting measures in June.
Cost control is clearly part of the story, but not the whole thing. Meta has every incentive to funnel workers toward its own products rather than keep paying Anthropic, and reports say it also wanted to limit Anthropic’s access to its training data. Anthropic’s rapid growth is now creating a familiar problem for the giants feeding it: the better it gets, the more it looks like a rival. And Microsoft has its own reason to flinch, with Claude Cowork and ChatGPT Work increasingly poking at Office territory.
My take — AI-written commentary, not fact-checked reporting
This is the oldest story in tech: the platform will happily rent your AI until it decides to build its own. Meta and Microsoft are not making a moral statement here; they’re doing business, which is usually less romantic and more brutal. The only surprise is that Anthropic got this far before the choke chain got pulled.
Read more about this at: The Decoder