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

Microsoft has quietly canceled most internal licenses for Claude Code, Anthropic’s AI-powered coding assistant, after a six-month experiment. The company’s Experiences and Devices division, responsible for Windows, Microsoft 365, and Surface, will require engineers to migrate to GitHub Copilot CLI by June 30, citing toolchain unification as the official reason. However, industry observers suggest financial pressures are the real driver, as token-based pricing models struggle to scale with high usage. The initial rollout in December 2023 allowed Microsoft employees—including engineers, product managers, and designers—to use Claude Code at no cost. By spring, adoption had expanded beyond technical roles, raising concerns about the sustainability of token-based pricing. Microsoft’s decision reflects broader industry challenges, where AI coding tools, despite their utility, generate unpredictable costs due to heavy token consumption. Uber’s experience underscores the issue. In April, Uber’s CTO, Praveen Neppalli Naga, revealed the company had exhausted its 2026 AI coding budget in just four months. Engineers were spending between $500 and $2,000 monthly on tokens, with 70% of committed code now AI-generated. Naga acknowledged the budget shortfall, noting that token usage far exceeded initial projections, as agentic AI tools require extensive computational resources. GitHub faced a similar problem last November when it paused new sign-ups for Copilot Pro and Pro+ due to excessive token costs from agentic workloads. The company admitted its pricing model, designed for lightweight assistance, couldn’t accommodate the demands of advanced AI coding tools. This shift in cost dynamics has forced enterprises to reevaluate their AI strategies, as compute expenses now surpass employee salaries in some cases, according to Nvidia’s vice president of applied deep learning, Bryan Catanzaro. The trend signals a broader industry reckoning: while AI coding tools improve productivity, their financial viability remains uncertain. Companies like Microsoft and Uber are scaling back or adjusting usage to control costs, marking a pivot from rapid adoption to cautious optimization. The experiment with Claude Code serves as a case study in the challenges of integrating AI into enterprise workflows without sustainable economic models.