Meta and Microsoft have significantly reduced their internal usage of Anthropic’s Claude AI models, according to reports from The Information. Microsoft, which previously projected over $1 billion in annual Claude spend, slashed its cloud division’s per-employee monthly budget from $100,000 to approximately $10,000, directing staff to use GitHub Copilot and OpenAI models instead. Meanwhile, Meta saw its active Claude Code users halve from 60,000 to 30,000 as it shifts focus toward in-house alternatives like Muse Code and MetaCode.

The pullbacks stem from both cost-cutting initiatives and strategic misalignment. Meta recently spent over $105 million on Claude Code in a single 28-day window but is now prioritizing its own tools and seeking to protect its training data from external developers.

As Anthropic expands products like Claude Cowork into workspace territory, major tech incumbents increasingly view the startup as a direct rival rather than just a software vendor. The shift highlights growing friction between frontier model developers and enterprise platforms competing for software developer and workplace productivity markets.

Why it matters

  • Big tech enterprises are prioritizing internal proprietary models to cut soaring third-party AI software costs.

  • Anthropic faces growing enterprise churn from former allies as its product expansion triggers direct competition.

  • Data protection and model independence are becoming primary drivers in enterprise vendor selection decisions.

Source: the-decoder.com