Microsoft’s quiet Claude Code retreat and the real cost of enterprise AI

Microsoft is cancelling most Claude Code licences inside its Experiences and Devices division, redirecting engineers to GitHub Copilot CLI by June 30. The move is framed as toolchain unification but the real driver is cost: agentic AI coding tools consume tokens at a rate that breaks enterprise budget models. Uber's CTO revealed his company burned through its entire 2026 AI coding budget in four months, with individual engineers spending $500–$2,000/month on tokens. The core problem is that token-based pricing behaves like a metered utility, not a per-seat licence — and agentic systems consume more tokens per task with each generation, not fewer. Gartner places generative AI in the trough of disillusionment, with 25% of 2026 AI budgets expected to slip. The analysis argues this signals the end of the experimental 'absorb arbitrary costs for learning' phase, with enterprises moving toward capped budgets, tiered access, and AWS-style billing rather than flat-seat models.