Microsoft and Meta Platforms are reducing employees’ internal use of Anthropic’s Claude as both companies promote proprietary artificial-intelligence tools, raising questions about the AI startup’s customer concentration as it moves toward an initial public offering. Microsoft’s earlier projection of at least $1 billion in internal Anthropic spending has been cut by more than a third, although customer purchases of Anthropic models through Microsoft’s enterprise platforms continue to grow. At Meta, estimated Claude users fell to about 30,000 from roughly 60,000 earlier in 2026, based mainly on token consumption and spending rather than confirmed headcounts. From late June 2026, Meta restricted access to Claude Code and OpenAI’s Codex, particularly for applied AI engineers, over concerns that their outputs could influence Meta’s training data through model distillation. Meta continues to spend hundreds of millions of dollars a month on Anthropic models while developing MetaCode and Muse Code as in-house alternatives.