Microsoft cuts Claude Code as AI costs spike

Microsoft is canceling most Claude Code licenses as AI usage costs rise faster than expected.

Microsoft has reportedly started canceling most direct Claude Code licenses and steering engineers toward GitHub Copilot CLI, according to The Verge. The move comes six months after Microsoft opened access to Claude Code for thousands of employees, and it lands as Fortune reports that AI coding budgets are getting burned through faster than companies planned.

What changed

Microsoft’s internal AI rollout is moving from broad access to tighter control. The company is canceling most direct Claude Code licenses, but that does not affect its broader Foundry deal with Anthropic, which includes Microsoft investing up to $5 billion and Anthropic committing $30 billion to buy Azure compute capacity.

Other companies are hitting similar limits. Uber’s CTO said in April that the company used its full 2026 AI coding tools budget in four months, after previously pushing adoption with internal leaderboards. Meta employees have also built a “Claudeonomics” tracker, while Amazon has urged staff to “toxenmaxx,” or use more AI tokens.

• Microsoft opened Claude Code access about six months ago.
• Engineers are being shifted to GitHub Copilot CLI.
• Microsoft’s Foundry deal with Anthropic remains in place.
• Uber exhausted its AI coding tools budget in four months.

Why it matters

The core issue is pricing. AI tools are often billed by token, so heavier use means higher bills even when the software saves time. That makes agentic systems look cheaper per unit while becoming more expensive at scale.

Goldman Sachs expects agentic AI to drive a 24-fold jump in token use by 2030, reaching 120 quadrillion tokens per month. Gartner says inference on a one-trillion-parameter model could cost nearly 90% less by 2030, but warns that enterprise AI may still get pricier because agents use far more tokens per task and vendors may not pass through all savings.

For developers and CIOs, the message is simple: AI adoption is no longer just a productivity question. It is also a compute-budget problem, and the bill can grow faster than the headcount it is meant to replace.

The open question is whether companies can keep expanding AI use without turning token consumption into a new operating expense line that rivals payroll.