Uber has instituted a monthly cap of $1,500 per employee on agentic coding tools including Anthropic’s Claude Code and Cursor, Bloomberg reported on June 2, as the company moves to control AI spending that ran well beyond plan this year. Employees can track usage via an internal dashboard, and the caps can be exceeded with approval in certain cases. The cutback follows a disclosure from Uber’s CTO in April that the ridesharing giant had burned through its entire annual AI budget in just four months — an outcome that stemmed partly from the company’s own earlier posture: Uber had told staff to use AI “as much as possible” and ranked internal usage competitively on internal leaderboards, The Information previously reported.
The scale-back has arrived alongside growing leadership skepticism about whether the spending is generating measurable returns. COO Andrew Macdonald said during a recent podcast that “it’s very hard to draw a line” between AI usage and new consumer features — a notably cautious assessment from a company that until recently championed maximum AI adoption. The case reflects a wider tension now building across enterprise technology: after a wave of aggressive AI tool rollouts and expanding budgets, companies are beginning to ration access as costs outpace demonstrable productivity gains. A Bain survey published this week found AI is delivering less cost reduction than many firms predicted, and GitHub Copilot’s recent shift to token-based billing has already prompted significant pushback from developers. TechCrunch noted the ROI on enterprise AI spending “has so far remained a largely theoretical phenomenon.”