Meta and Microsoft Urge Staff to Cut Claude AI Usage to Reduce Costs
Meta and Microsoft are directing employees to reduce internal reliance on Anthropic's Claude AI models to cut enterprise expenses. Microsoft slashed its internal Anthropic budget by over one-third after annual costs were projected to hit $1 billion, while Meta halved its internal Claude Code user base by promoting internal tools like MetaCode and Muse Code. This shift highlights how tech giants are prioritizing cost discipline and self-hosted AI tools over expensive third-party subscriptions for software engineering. It also presents a major revenue risk for Anthropic, as Meta and Microsoft account for roughly a quarter of its revenue without long-term binding contracts. Despite internal cutbacks, Microsoft still offers Anthropic models to external cloud clients, and Meta spent over $105 million on Claude Code during a recent 28-day period. Meta's newly deployed Muse Code is a terminal-based coding agent powered by in-house LLMs, built specifically to displace Claude Code internally.
## BACKGROUND
Anthropic's Claude Code is an agentic terminal tool designed to automate complex software development tasks directly within a programmer's workspace. As large language models became critical for engineering productivity, tech companies initially allowed employees to use top-performing external models before developing proprietary alternatives to minimize API expenditure.