Uber Technologies has imposed mandatory monthly spending caps on the AI coding tools used by its engineers after the company exhausted its entire 2026 AI budget in the first four months of the year.

Under the new rules, which went unreported until this month, each employee is limited to $1,500 in monthly token spending for every AI coding tool. The policy applies specifically to agentic coding software, including Cursor and Anthropic’s Claude Code. Uber has also provided a centralized dashboard so staff can track usage and a formal approval process to request exceptions.

Uber Chief Technology Officer Praveen Neppalli Naga confirmed in April that the annual AI budget had already been overspent, and Chief Executive Dara Khosrowshahi said in late May that roughly 10 percent of the company’s code is now written and submitted by AI agents. The ride-share company also said it expects AI-driven efficiency gains to slow its overall hiring pace this year.

The limits offer a concrete snapshot of how quickly generative AI costs can scale across a large engineering organization. Executives said adoption inside legal and marketing teams is also accelerating, adding that no single customer-facing benefit has yet been directly tied to the internal AI spending, a point Uber Chief Operating Officer Andrew Macdonald described as hard to measure at this stage.

The move also underscores a broader tension in corporate America. Companies are racing to embed AI into their operations, but they are also struggling to forecast and control the associated costs. Uber’s decision to cap usage rather than expand the budget sends a signal that even the most enthusiastic adopters are beginning to treat AI spend as a line item that requires discipline.

The cap does not apply to non-coding AI tools, leaving open the question of whether other departments will eventually face their own limits as the company’s overall AI footprint continues to grow.

Image source: v3b.fal.media