Microsoft Pulls Claude Code Licences as Token Costs Spiral Out of Control
Microsoft Pulls Claude Code Licences as Token Costs Spiral Out of Control
Microsoft is cancelling most internal licences for Anthropic’s Claude Code across its Experiences and Devices division — the team behind Windows, Microsoft 365, Outlook, Teams, and Surface — with a deadline of 30 June 2026. Engineers are being redirected to GitHub Copilot CLI, Microsoft’s own command-line AI coding tool. The decision, first reported by The Verge’s Tom Warren on 14 May, was not driven by poor performance. It was driven by the opposite.
Too Popular for Its Own Good
Claude Code launched inside Microsoft’s Experiences and Devices group in December 2025. Within six months, it had become the preferred tool among engineers, project managers, and designers who were given direct access to the AI coding assistant in their daily workflows. Warren described it as having become “perhaps a little too popular” inside the company.
The internal directive came from Rajesh Jha, Microsoft’s Executive Vice President. “Claude Code was an important part of that learning,” Jha wrote in an internal memo. “At the same time, Copilot CLI has given us something especially important: a product we can help shape directly with GitHub.”
The Token Pricing Problem
The cancellation exposes a structural tension at the heart of enterprise AI deployment. Unlike conventional software licences — which cost the same whether an employee uses the tool for one hour or eight — token-based AI pricing scales directly with usage. Every prompt, code generation request, code review, and debugging session consumes tokens. The more useful the tool, the higher the bill.
At the individual developer level, the numbers are manageable. At the scale of thousands of engineers using the tool daily, they compound into a line item that can catch finance teams off guard.
Uber’s recent experience illustrates how quickly costs can escalate. The company deployed Claude Code to 5,000 engineers, and monthly usage rates climbed to between 84% and 95% by April 2026. Per-engineer API costs reached between $500 and $2,000 per month. Uber’s head of engineering acknowledged the forecasting failure bluntly: “I’m back to the drawing board, because the budget I thought I would need is blown away already.”
A Counternarrative for Anthropic
The timing is awkward for Anthropic. The company is currently raising capital at a reported valuation of $900 billion, with enterprise adoption central to its pitch. Its most prominent internal enterprise user — Microsoft — finding the cost model unsustainable is not a fatal development, but it is a signal that enterprise buyers and investors will weigh carefully.
Microsoft is uniquely positioned to understand what enterprise-scale Claude usage actually costs. Inside Experiences and Devices, its engineers were among the heaviest users outside Anthropic’s own customer base.
Two Readings of the Same Decision
Microsoft’s official line is strategic: Copilot CLI is the intended destination, engineers will retain access to Claude models within Copilot, and the company simply wants a product it can develop directly with GitHub. The second reading is harder to dismiss — what changed was not the strategic logic, but the bill.
The broader implication is structural. The first wave of enterprise AI adoption was driven by enthusiasm about capability. The second wave, now underway, is being shaped by cost discipline. Enterprises will continue buying AI coding tools — the productivity gains are real and competitive pressure is unforgiving. But the procurement model is shifting.
What replaces the era of blanket Claude Code seats will look less like an Office licence and more like an AWS billing dashboard: usage caps, consumption monitoring, and a finance team in the room.
