Meta and Microsoft Cut Back on Claude: What Does It Mean for Anthropic’s IPO?

Microsoft and Meta are reducing their internal use of Anthropic’s Claude while steering employees toward their own AI tools. Microsoft reportedly cut its projected Anthropic spending by more than a third from an earlier estimate of at least $1 billion. At Meta, Claude Code users fell from roughly 60,000 to 30,000 as its alternatives gained adoption. Source

Microsoft’s cloud and AI division reportedly lowered monthly per-employee AI allowances from around $100,000 to $10,000. These are spending limits, so the reduction does not necessarily translate into an equivalent revenue loss for Anthropic. Microsoft is also encouraging GitHub Copilot adoption, although Copilot itself supports Claude models, meaning a change of tool does not automatically eliminate Anthropic usage. GitHub documentation

Meta’s shift presents more direct competition. Its internal MetaCode tool reportedly has over 30,000 users, while Muse Code has exceeded 6,000. Their growth suggests that Meta can redirect development work toward products it controls, although user counts alone do not establish whether those tools match Claude’s performance or reduce total costs.

Anthropic has pushed back, saying payments from these customers remain in the hundreds of millions and that broader customer growth offsets concentration concerns. Its annualized revenue run rate reportedly exceeded $65 billion by late July. That figure extrapolates recent sales over a year; it is not revenue already collected across twelve months. Revenue reporting

Ahead of its proposed IPO, Anthropic faces a test of how much work Claude retains when customers tighten budgets. Companies may reserve premium models for demanding tasks while moving routine work to cheaper alternatives. Anthropic’s position will depend on whether Claude’s performance justifies continued spending—and whether growth across other customers outweighs reductions at its largest accounts.