Meta and Microsoft are pulling back from Claude at a moment when enterprise AI budgets, internal product strategy and competitive pressure are starting to collide. Both companies have been major Anthropic customers, but the source article says each is now sharply reducing internal use of Claude.
The move is not only about trimming spending. It also reflects a larger change in the AI market: Anthropic is no longer just a supplier of powerful tools to large technology companies. It is increasingly competing with them.
Microsoft pushes staff toward its own AI stack
At Microsoft, Claude spending had been projected at over $1 billion annually, according to The Information. That figure helps explain why leadership attention has turned toward how Claude is being used inside the company.
The source article says Microsoft executives Scott Guthrie and Jay Parikh told employees to use in-house tools like GitHub Copilot and OpenAI's models instead. That is a significant instruction because it moves internal AI work closer to products and partnerships Microsoft already controls.
Spending was cut by more than a third. In the cloud division, the monthly per-employee budget dropped from $100,000 to about $10,000.
That change suggests a practical reset. If employees can complete similar work using GitHub Copilot and OpenAI's models, Microsoft has a clear reason to reduce dependence on Claude. It can also keep more usage inside systems aligned with its own AI strategy.
Meta reduces Claude Code users
Meta is also reducing its use of Claude. The source article says Claude Code users at Meta fell from roughly 60,000 to 30,000.
That drop is attributed partly to layoffs and partly to Meta pushing its own tools, including Muse Code and MetaCode. The direction is clear: Meta wants employees using internal AI coding tools rather than relying so heavily on Anthropic's Claude Code.
The spending involved was large. In one 28-day stretch, Meta spent over $105 million on Claude Code alone.
Meta had already introduced AI cost-cutting measures in June. The Claude pullback fits that broader effort, but the source makes clear that cost is only part of the explanation.
The cost issue is only one layer
AI tools can become expensive quickly when thousands of employees use them every day. For companies the size of Meta and Microsoft, even small changes in per-user behavior can become major budget items.
But the source article points to another concern: competitive alignment. Meta has little reason to keep feeding Anthropic when it wants to sell competing products. The article also says Meta reportedly wanted to restrict Anthropic's access to its training data.
That turns a purchasing decision into a strategic one. Buying Claude access may help employees work faster, but it can also strengthen a company that is trying to win in the same market.
For Microsoft, the same issue appears in a different form. Claude is useful, but Microsoft already has GitHub Copilot and access to OpenAI's models. Steering employees toward those tools reinforces Microsoft's existing AI platform rather than expanding Claude's role inside the company.
Anthropic is becoming harder to treat as just a partner
The source article says Anthropic has grown extremely fast and is becoming a headache for incumbents. That growth changes how large technology companies view the relationship.
When a supplier is small, buying from it can be a straightforward productivity decision. When that supplier grows into a rival, every purchase looks different. The customer has to consider not only whether the product works, but also whether the spending helps a competitor gain strength.
The article notes that even Microsoft has reason to worry because tools like Claude Cowork and ChatGPT Work are increasingly turning into Office alternatives. That matters because it places AI assistants closer to the everyday productivity software market, not just developer tooling or research workflows.
In that context, reducing Claude use is not simply a budget cut. It is also a way to control where employee workflows, usage data and internal habits settle over time.
What this signals for enterprise AI
The Meta and Microsoft pullbacks show how quickly the enterprise AI market can shift. A tool can be adopted widely because it is useful, then face restrictions once the provider becomes too strategically important.
Several pressures appear to be acting at once:
- Cost control: Claude spending at Microsoft and Meta reached levels that made reductions attractive.
- Internal tools: Microsoft is directing staff toward GitHub Copilot and OpenAI's models, while Meta is promoting Muse Code and MetaCode.
- Competitive risk: Anthropic is increasingly positioned as a rival, not only as a vendor.
- Platform strategy: Companies want employees to build habits around tools that support their own product ecosystems.
For Anthropic, the situation cuts both ways. The reductions from Meta and Microsoft show the risk of depending on giant technology companies as customers. But the same decisions also underline why Anthropic matters: its tools became important enough, expensive enough and competitive enough to force a strategic response.
The clearest takeaway is that enterprise AI adoption is entering a more selective phase. Large companies are no longer just testing the best available model or coding assistant. They are asking who owns the tool, who benefits from usage, and whether today's productivity gain strengthens tomorrow's competitor.