Uber has capped all employee spending on agentic coding tools at $1,500 per month per tool after exhausting its entire 2026 AI budget within four months, according to reporting published June 3, 2026. The limits apply to tools including Cursor and Anthropic's Claude Code.

What Happened

The rideshare company limited all employees to $1,500 monthly in token spending per AI coding tool. The caps apply individually per tool, meaning an engineer using both Cursor and Claude Code faces a combined $3,000 monthly ceiling. TechCrunch reported that Uber's CTO revealed in April that the company had depleted its full annual AI budget in four months. Before the cap, engineers were spending between $500 and $2,000 per month each on AI tool access.

The company is not eliminating access to AI tools. An internal dashboard lets each employee track usage, and engineers can request permission to exceed the cap in specific cases. Uber is treating AI tool usage as a managed benefit with a ceiling, similar to software license budgets.

Why It Matters

Uber's situation reflects a pattern emerging across larger enterprises in 2026: organizations that moved quickly to adopt agentic coding tools are now encountering real budget pressure as usage scales with team size. The step from piloting AI tools to running them across hundreds of engineers produces costs that require active management.

For individual engineers, the $1,500 monthly cap covers substantial usage. Simon Willison noted his own monthly spending runs around $1,000, leaving buffer room under the Uber ceiling. The constraint matters primarily for heavy agentic workflows on large codebases, extended autonomous tasks, or multi-step review pipelines that generate high token volumes.

Key Details

  • Cap: $1,500 per month, per AI coding tool (Cursor, Claude Code, and equivalents)
  • Scope: All employees, all agentic coding tools
  • Trigger: Uber's 2026 AI budget fully consumed within four months
  • Pre-cap spend: Engineers were individually spending $500 to $2,000 per month
  • Exception process: Caps can be exceeded with manager approval

What to Do Next

For developers facing token budget constraints, one practical option is routing Claude Code through a cost-reducing proxy. A Claude Code Proxy can direct requests through lower-cost providers including Kimi and ChatGPT subscriptions, significantly reducing per-token spend on standard coding tasks without changing the Claude Code interface.

For teams managing AI tool budgets, per-tool individual caps are simpler to enforce than team-level allocations and avoid cutting access entirely. The approach also creates natural feedback: engineers who hit the ceiling regularly are likely getting high value and are candidates for a higher limit, while those who never approach it may need less access than initially provisioned.

The AI Insider noted that Uber's experience is likely to influence how other large enterprises structure AI tool policies in the second half of 2026. Budget exhaustion within a fiscal year is a forcing function that transforms AI tools from an experimental expense into a line item that requires governance like any other software spend.

If you manage AI tool budgets at any scale, Uber's approach offers a replicable model: cap per tool, not per team, allow individual exceptions with approval, and give employees visibility into their own usage. That combination controls cost without eliminating the productivity gains that made the tools worth adopting.