Meta and Microsoft Scale Back Internal Use of Anthropic’s Claude Models

Meta and Microsoft are scaling back their internal use of Anthropic’s Claude models, opting instead for homegrown coding tools. The shift, reported by The Information and summarized by other outlets in October 2026, does not signal an end to their commercial relationships with Anthropic but reflects internal cost-control measures and tool preferences.

Meta’s use of Claude Code reportedly dropped by half, with the number of active employees using the tool falling from around 60,000 to 30,000. Despite this decline, Meta spent over $105 million on Claude Code during a single 28-day period this year. The company is now directing engineers toward internal alternatives like MetaCode, Muse Code, and the Muse Spark model line.

Microsoft, meanwhile, cut its projected 2026 internal spending on Anthropic by more than a third, reducing a forecast of at least $1 billion. Within its Cloud and AI organization, monthly AI spending limits per employee reportedly dropped from around $100,000 to $10,000. Microsoft is steering employees toward its own tools, particularly GitHub Copilot and GitHub Copilot CLI, which rely on OpenAI’s models.

Both companies continue to offer Claude to their customers, with Microsoft integrating it into Microsoft Foundry, Azure, and parts of the Copilot family. The reduction in internal usage highlights a strategic pivot rather than a full withdrawal from Anthropic’s technology.