Meta and Microsoft Step Back from Claude as Anthropic Evolves from Ally into Rival

Meta description: Meta and Microsoft sharply reduced internal use of Anthropic's Claude as AI budgets tighten and rivals push in-house tools and models for enterprise teams.

Tags: Anthropic Claude, Meta, Microsoft, AI cost management, enterprise AI

Featured image: abstract technology style, blue circuit patterns, no real people, no company logos, no text.

Two major buyers of Anthropic's Claude have started pulling back inside their own companies, turning a tool many teams adopted into a budget line under scrutiny. Meta and Microsoft, two of Anthropic's biggest enterprise customers, are sharply cutting internal use of Claude, and the change matters because it shows how enterprise AI spending can shift when cost, competition, and control over data meet.

Background: Claude became a large internal spend

Anthropic's tools had already become large enough to affect internal budgets at major technology companies. The Information reported that Microsoft's Claude spending had been projected at over $1 billion annually. That number placed Claude inside a tier of enterprise AI tools with budgets that could affect cloud division economics.

Meta's exposure was visible in a different product: Claude Code. Meta had roughly 60,000 Claude Code users, and in one 28-day stretch, Meta spent over $105 million on Claude Code alone. The company had already introduced AI cost-cutting measures in June, signaling that heavy model usage was becoming a management issue before the latest reductions.

Microsoft redirects budgets to GitHub Copilot and OpenAI models

One clear change came inside Microsoft. Executives Scott Guthrie and Jay Parikh told employees to use in-house tools like GitHub Copilot and OpenAI's models instead of Claude, according to The Information. The directive reduced Claude spending by more than a third.

The budget cut was most concrete in the cloud division, where the monthly per-employee budget dropped from $100,000 to about $10,000. The shift changed internal spending, but it was not described as ending Microsoft's relationship with Anthropic. For a company already selling GitHub Copilot and using OpenAI models, the move also reduced the risk of a third-party tool becoming a preferred internal standard.

Meta halves Claude Code use while pushing Muse Code and MetaCode

Meta's adjustment centered on developer workflows. Claude Code users fell from roughly 60,000 to 30,000. Part of the decline came from layoffs, but Meta was also pushing its own tools like Muse Code and MetaCode.

For Meta, the decision is not only about saving money. The company wants to sell competing products, and it reportedly also wanted to restrict Anthropic's access to its training data. The irony is easy to see: Meta is worried about another company using its data.

The wider pressure on Anthropic's fast growth

Anthropic has grown extremely fast, and that growth is becoming a headache for incumbents. Meta's move shows how a customer can become a competitor. Microsoft's move shows how an enterprise can use one model provider while still steering internal teams toward tools that support its own product roadmap.

The pressure also extends into productivity software. Tools like Claude Cowork and ChatGPT Work are increasingly turning into Office alternatives. That gives Microsoft another reason to limit internal reliance on Claude, even when Claude may be useful in isolated workflows.

Current stage: reduced use, not a clean break

The current phase is a contraction, not a public divorce. Both companies still have reasons to use Anthropic's models selectively, and both have reasons to reduce Claude's footprint when internal tools can cover the same work. The most important numbers now are the drop in Microsoft's projected spending, the cut in Meta's Claude Code users, and the new internal budget ceilings. What comes next will depend on whether Muse Code, MetaCode, GitHub Copilot, and OpenAI models can hold the workloads that Claude had been handling.