$MSFT and $META Curb Internal Claude Usage to Rein in Surging AI Token Costs

Tech giants Microsoft ($MSFT) and Meta ($META) are clamping down on internal usage of Anthropic's Claude as enterprise AI bills escalate, signaling a broader industry pivot toward strict token budget discipline. Microsoft has dramatically cut internal developer allowances—slashing monthly per-employee AI spending limits from $100,000 down to $10,000—while steering thousands of engineers away from standalone Claude Code licenses and back toward GitHub Copilot CLI. Concurrently, Meta is actively promoting its in-house Muse Code toolchain to contain external model expenses, even as aggregate spending on frontier models remains substantial.
The clampdown underscores how pay-per-token pricing models are straining enterprise IT budgets once developer adoption surges. As companies begin treating API tokens like conventional operational line items, usage continues to expand, but pricing leverage is shifting. The transition toward capped internal budgets and in-house proprietary tooling is mounting margin and pricing pressure on leading frontier labs like OpenAI and Anthropic.