Meta and Microsoft are both scaling back how much their own employees use Anthropic's Claude, according to a report from The Information published on October 5, 2026, as the two tech giants steer their workforces toward in-house AI coding tools and cheaper internal budgets.

The cutbacks do not mean either company is abandoning Claude for customers. Both remain significant Anthropic clients — and both compete with the startup at the same time. But the report offers a detailed look at how two of Anthropic's biggest enterprise customers are tightening the leash on internal AI spending, even as Anthropic's business reportedly runs at a $65 billion annualized revenue pace. For continuous coverage of stories like this, AI Buzz Wire tracks the latest AI developments as they land.

Microsoft's $1 Billion Estimate Shrinks by a Third

At Microsoft, management had previously projected that internal spending on Anthropic technology would exceed $1 billion per year. According to The Information, that estimate has since fallen by more than a third after executives directed employees to curb their use of Claude and lean harder on Microsoft's own tools, including GitHub Copilot and OpenAI-based frameworks.

The mechanics are budgetary. Within Microsoft's cloud and AI division, monthly AI spending ceilings per employee have reportedly been slashed from $100,000 to roughly $10,000 in most cases. The report is careful to note these are spending limits, not what engineers actually spend — but the effect on culture is real. Engineers who previously enjoyed broad latitude to experiment with outside models now find that latitude sharply reduced.

Microsoft's own published numbers underline why the shift matters to its bottom line. The company's stewardship of OpenAI's models through Azure, and its push to make GitHub Copilot the default coding assistant, only pay off if its own developers use them.

Importantly, The Information's reporting indicates that client-facing spending on Anthropic models through Microsoft's enterprise platforms continues to rise even as internal usage gets squeezed. Microsoft sells what its customers want — and many enterprise customers still want Claude.

Meta Halves Its Claude Code User Base

The picture at Meta is one of substitution. The number of employees using Claude Code reportedly plummeted from around 60,000 earlier this year to about 30,000.

Layoffs account for part of the decline, but The Information attributes the primary cause to Meta's strategic pivot toward its own AI development tools, built on Meta's proprietary models. The company's internal coding assistant, MetaCode, now counts more than 30,000 internal users, while Muse Code — a newer tool Meta began testing externally with clients in August — has more than 6,000 employee users.

The halving of Claude Code usage is striking given what Meta still spends. The report notes Meta allocated more than $105 million to Claude Code over a 28-day period during this transition, a figure suggesting that fewer users does not immediately mean less money flowing to Anthropic.

Competing With Your Supplier

The dynamic at both companies is a textbook case of what happens when AI buyers become AI builders. Meta and Microsoft are simultaneously:

  • Customers — paying Anthropic for model access used internally and, in Microsoft's case, resold through enterprise platforms
  • Competitors — building coding assistants on proprietary and partner models that substitute for Claude
  • Cost managers — imposing internal budgets that push engineers toward in-house tools

That triple role explains the seemingly contradictory signals: Microsoft cuts its internal Claude budget by a third while customer spending on Claude through its platforms grows, and Meta cuts its Claude Code user base in half while spending nine figures on the tool in a single month.

For Anthropic, the trend cuts both ways. Internal retrenchment at two of the world's largest technology companies is a leading indicator of where substitution pressure will come from — every MetaCode user is a seat that might otherwise have been Claude's. But the reported $65 billion annualized revenue pace suggests demand from the broader enterprise market remains strong enough to absorb it, at least for now.

The Broader Context: Coding Tools Become Strategic Infrastructure

The report lands at a moment when coding assistants have become core software engineering infrastructure rather than experiments. That shift is precisely why Meta and Microsoft are rationalizing their toolchains: when every engineer's daily workflow runs through an AI assistant, the choice of model becomes a strategic decision about cost, control, and data — not a developer preference.

It also explains the timing of Anthropic's own moves this week. On Tuesday, the company launched Claude Haiku 5.5 with price cuts of up to 90% on small-model API calls, halved Sonnet 5.5 cache-read pricing, and introduced monthly API credits for subscribers — a package clearly designed to make Claude the economical default for high-volume agentic work.

Whether those cuts win back cost-sensitive internal buyers like Meta and Microsoft, or merely slow the migration to in-house tools, will be one of the more closely watched storylines in enterprise AI over the coming quarters.

Neither Meta, Microsoft, nor Anthropic has publicly commented on the specifics of The Information's report.

Stay Ahead of AI Read more AI news on AI Buzz Wire