Skip to content

Anthropic

Meta trims Claude Code use, yet spent over $105M in 28 days

2 min read

Written by an AI model from the linked sources and checked against them by a second model. How we work

All about: Claude Code versions and changes →
Image: rswebsols.com

Meta and Microsoft are moving their own engineers off Anthropic’s Claude and onto in-house coding tools, The Information reported on October 5. Meta’s Claude Code user count has roughly halved, and Microsoft’s projection for internal Anthropic spending is down by more than a third. The cuts cover the two companies’ employees only. Customer spending on Anthropic models through Microsoft’s enterprise platforms reportedly keeps rising, and the Microsoft figures are spending caps and an internal projection rather than what was actually spent.

Microsoft’s caps and projections

Microsoft had expected internal spending on Anthropic technology to run above $1 billion a year. That estimate reportedly fell by more than a third after management told staff to use less Claude and more Microsoft tools, including GitHub Copilot and OpenAI models.

The sharper cut is per person. In Microsoft’s cloud and AI unit, the monthly AI spending cap per employee dropped from $100,000 to approximately $10,000 in most cases. Because these are ceilings, the drop tells you nothing about what any given engineer actually spent. Some engineers who used to have more room to try out models are reportedly frustrated by the change.

Where Meta’s users went

At Meta, Claude Code users went from around 60,000 earlier this year to about 30,000, according to The Information. Spring layoffs, which hit about 10% of Meta’s 78,000-person workforce, account for part of that drop. The main reason, Yahoo Finance reports as well, is Meta’s push into its own tools. MetaCode is internal-only and has more than 30,000 users. Muse Code, a Claude Code competitor, has over 6,000 employee users and began external testing with customers in August. Both run on Meta’s own models and are tied to its Muse Spark model line.

Over one 28-day period during this shift, Meta reportedly still spent more than $105 million on Claude Code, so fewer users does not necessarily mean a smaller bill.

Both companies remain major Anthropic customers while also competing with it. Anthropic’s annualized revenue pace is reported at $65 billion.

Same risks on the new tools

Switching assistants still leaves an agent with access to sensitive files, commands and credentials. Cybersecurity News has previously described Claude Code flaws that could have allowed unauthorized execution and theft of API keys; Anthropic fixed them before they were disclosed. Microsoft’s tools have had similar problems: the now-patched RoguePilot vulnerability showed that malicious instructions embedded in a GitHub Issue could lead to a repository takeover.

No comments yet. Be the first.

Join the conversation

Sign in to leave a comment. It appears after moderation.