AI too expensive: Uber restricts employees from using tools like Claude, puts hard limit of $1500

Uber has restricted employee access to AI coding tools, enforcing a $1,500 monthly spending cap per tool after burning through its annual AI budget in just four months of 2026. The new policy applies to agentic coding platforms like Claude Code and Cursor, requiring approval for exceeding limits. Employees will track usage via dashboards, while requests for additional access must be formally approved. The move follows Uber’s earlier push to maximize AI adoption, including internal leaderboards ranking employees by tool usage. Chief Technology Officer Praveen Neppalli Naga revealed the company had already consumed its full AI budget by April, despite AI now generating around 10% of Uber’s code. Beyond engineering, legal and marketing teams have also integrated AI tools to boost productivity, though quantifying long-term benefits remains challenging. Uber’s Chief Operating Officer, Andrew Macdonald, acknowledged difficulties in directly linking AI spending to customer-facing improvements, despite internal productivity gains. The company is not alone in facing AI cost pressures, as Microsoft has also restricted access to Claude Code, shifting engineers to GitHub Copilot CLI by June 30, 2026, partly due to rising operational expenses. Microsoft’s transition affects engineers in its Experiences + Devices division, which oversees products like Windows and Surface devices. The shift follows Claude Code’s popularity among developers after its late-2025 rollout, though its high token consumption contributed to cost concerns. Both companies are balancing AI’s productivity benefits against escalating financial burdens as adoption scales. Uber’s spending cap applies separately to each AI tool, with employees receiving individual budgets. The company emphasizes responsible experimentation while controlling costs amid growing AI integration across departments. The policy reflects broader industry challenges as firms move from AI pilot projects to large-scale deployment.