Microsoft cuts Claude Code access as AI coding costs surge

Microsoft has scaled back access to Anthropic’s Claude Code for thousands of its employees, including developers, project managers, and designers, citing rapidly increasing operational expenses from AI-assisted coding tools. The company initially expanded access to encourage experimentation but now faces rising costs tied to token-based pricing, where usage volume drives expenses. Despite the rollback, Microsoft’s broader partnership with Anthropic remains intact under a multibillion-dollar Foundry agreement. This deal includes cloud infrastructure support and access to Claude models via Azure services, ensuring continued collaboration outside of Claude Code’s restricted access. The move reflects broader industry challenges, as companies like Uber have reported exhausting their 2026 AI budgets within months due to heavy employee adoption. Uber’s Chief Technology Officer, Praveen Neppalli Naga, highlighted the financial strain from token consumption, a pricing model that charges based on text processed and generated. Analysts project costs will rise further as businesses deploy agentic AI systems, which require significantly higher token usage per task. Goldman Sachs estimates token demand could surge by 2030, while Gartner notes that even as model costs decline, enterprise AI spending may grow due to increased complexity. Earlier this month, tech giants including Meta Platforms, Oracle, and Cloudflare announced job cuts and restructuring to shift focus toward AI-driven operations. Microsoft’s decision underscores the tension between innovation and cost management in scaling AI tools across large organizations.