Microsoft Pulls the Plug on Claude Code Over Runaway AI Costs
Microsoft is pulling back from one of its own AI experiments, and the reason is as straightforward as it is telling: the bill got too big. The company began revoking internal licenses for Claude Code for most employees on May 14, 2026, with a final deadline of June 30, which also marks the end of Microsoft’s fiscal year.
The Microsoft Claude Code ban affects the Experiences and Devices division, the group responsible for some of the company’s biggest products. Engineers working across Windows, Microsoft 365, Outlook, Teams, and Surface have been instructed to stop using Anthropic’s coding assistant and shift instead to GitHub Copilot CLI, Microsoft’s own AI-powered command-line coding tool.
The story behind the Microsoft Claude Code ban is one of a tool that worked almost too well. Back in December 2025, Microsoft made Claude Code available to thousands of employees, including engineers, product managers, and designers, encouraging everyone to reshape their workflows using the tool. Employees loved it. Then the token costs started arriving.
Microsoft’s Executive Vice President Rajesh Jha acknowledged the tool’s value in an internal memo. “Claude Code was an important part of that learning,” he wrote, before adding that GitHub Copilot CLI offered something more strategically valuable: “a product we can help shape directly with GitHub.”
The cost problem is not unique to Microsoft. Uber deployed Claude Code to around 5,000 engineers and saw monthly usage rates climb to between 84 and 95 percent by April 2026. Individual engineers were spending between $500 and $2,000 a month on tokens. The result was that Uber burned through its entire $3.4 billion 2026 AI budget in four months, with its CTO saying the annual budget had already been exhausted before the year was half over.
Fortune framed the broader issue bluntly, reporting that using AI is now more expensive than paying human employees, a conclusion that is gaining ground across the industry. Amazon is pushing employees to “tokenmaxx,” Meta has an internal leaderboard called “Claudeonomics” to track AI usage, and companies across the board are discovering that enthusiasm for AI tools does not come with a predictable price tag.
Importantly, canceling Claude Code licenses will not affect Microsoft’s broader Foundry deal with Anthropic, which includes investing up to $5 billion in the company and giving Foundry customers access to Claude models, as well as Anthropic’s $30 billion commitment to purchase Azure compute capacity. This is a budget decision, not a breakup.
What the Microsoft Claude Code ban signals, more than anything, is that the honeymoon phase of enterprise AI adoption is over. The question now is not which AI tools are most impressive, but which ones companies can actually afford to run at scale.
