Microsoft Drops Claude Code After Budget Overrun (aiweekly.co)

Microsoft has canceled its internal Claude Code pilot, which launched in December 2025, after consuming its entire annual AI budget within months due to token-based billing. This abrupt decision stems from the structural challenges of predicting costs under flat seat licenses that initially obscured token consumption until the switch to usage-based pricing revealed unsustainable expenditures. Affected developers are now being redirected to GitHub Copilot, which may strengthen Microsoft's internal tools at the expense of Anthropic, whose $900 billion valuation round hinges on enterprise adoption of Claude. This incident underscores a crucial lesson for the AI/ML community regarding enterprise pricing models. Token-based billing has become a standard mode of operation for frontier AI APIs, but this case exemplifies the budgetary risks enterprises face, as common procurement practices lack frameworks to manage such unforeseen spending. If other large enterprises encounter similar issues, it could lead to increased demand for fixed pricing models or spend controls, potentially slowing down AI deployments like Anthropic's. Moreover, this development might encourage competitors, including GitHub Copilot and others, to emphasize predictable pricing as a key selling point in future negotiations with enterprise clients.