Uber Caps Employee AI Coding Tool Spending

According to Bloomberg, Uber Technologies Inc. has instituted a $1,500 monthly cap on each employee's token spending for every agentic AI coding tool, including Claude Code and Cursor. The caps apply per tool, so usage of one does not reduce another's allotment, and were rolled out in recent months. The move followed disclosure, reported across Bloomberg, TechCrunch, and the Los Angeles Times, that Uber had used up its full-year AI budget within the first four months of 2026. Employees can monitor usage on an internal dashboard and may request approval to exceed the cap. An Uber spokesperson framed the limits to Bloomberg as a way to responsibly encourage agentic AI adoption and experimentation at scale.

Agentic coding tools bill per token or per inference, so costs can climb quickly when engineers run many iterations, parallel agents, or large-context prompts. Internal leaderboards and encouragement to experiment can amplify usage without clear visibility into return on investment. Per-user dashboards and approval workflows are a common governance pattern to contain outliers while preserving experimentation.

Multiple outlets place Uber's caps within a wider trend of companies constraining AI variable spend after rapid internal adoption. Reporting notes related signals, including statements that a meaningful share of Uber's code is now produced with AI agents and ongoing uncertainty about tying that usage directly to new product value.

• •Adoption of per-tool and per-user telemetry, dashboards, and approval workflows.
• •Whether teams link token usage to feature velocity or defect reduction.
• •Shifts in tool mix, or moves toward negotiated enterprise plans or in-house agents to stabilize costs.

Details come from Bloomberg's reporting and corroborating coverage by TechCrunch, the Los Angeles Times, Inc., Yahoo Finance, and GuruFocus.