Microsoft and Meta Limit Internal Use of Anthropic’s Claude Models

Microsoft and Meta are reportedly reducing their internal reliance on Anthropic’s Claude models to control rising computing costs and encourage the use of in-house AI tools. The move highlights a significant challenge for Anthropic: its biggest corporate customers are also its fiercest competitors. Despite this, both companies continue to offer Claude to external clients through enterprise products and cloud infrastructure.

At Microsoft, top leaders like Azure Cloud head Scott Guthrie and developer tools leader Jay Parikh have instructed teams to limit their use of Anthropic’s systems. The company has scaled back its projected spending on Anthropic tools by over a third, with individual monthly spending ceilings dropping from $100,000 to about $10,000 per person. Internal managers now receive daily tracking metrics on employee expenditures and model usage. However, Microsoft’s commercial partnership with Anthropic remains intact, with Claude integrated into Copilot productivity applications.

Meta has also reduced its reliance on Anthropic’s developer assistant, Claude Code. Usage within the company has halved, from roughly 60,000 users earlier this year to around 30,000 recently. Part of this decline is attributed to spring layoffs, but the primary reason is the rapid rollout of internal alternatives like MetaCode and Muse Spark. Meta is directing personnel toward these proprietary models, which have gained significant traction among employees.

These internal pullbacks by Meta and Microsoft contrast with Anthropic’s overall commercial success. The AI startup’s annualized revenue surged to over $65 billion by the end of July, up from approximately $9 billion at the end of 2025, according to reports from The Information and Reuters.