Uber's $1,500 monthly AI limit signals a real price for coding agents

Uber has begun limiting how much its employees can spend on AI coding tools, capping token usage at $1,500 per month for each tool. A company spokesperson confirmed the policy in response to an inquiry from Bloomberg News, in a report by Natalie Lung that developer and writer Simon Willison highlighted and analysed.

What the cap covers

The limit applies to agentic coding software such as Cursor and Anthropic's Claude Code, the spokesperson said. Spending on one tool does not affect the budget available for another, so the ceiling works per product rather than as a shared pool. According to the report, the restrictions were put in place over recent months.

The policy follows earlier reporting that Uber had used up its entire 2026 AI budget within four months. Willison notes that this is not surprising, since such budgets are typically set in 2025, before anyone could predict how popular token-burning coding agents would become.

A dollar value for AI-assisted work

The $1,500 figure is interesting because it puts a concrete price on what these tools are worth inside a large engineering organisation. Assuming an engineer actively uses two tools, the cap comes to $3,000 per month, or $36,000 per year. Levels.fyi lists the median yearly compensation package for Uber software engineers in the United States at $330,000, which means the AI spending ceiling is roughly 11 percent of that median package.

For comparison, Willison says his own token usage runs to about $1,000 per month against each of Anthropic and OpenAI, which currently costs him only $100 per provider thanks to subsidised plans for individual subscribers. Those plans are no longer offered to larger companies such as Uber. Under the new policy, he estimates he would still have about $500 per month to spare for each tool.

A signal for AI tool pricing

Willison describes the monthly cap as a rational response to overspending and a more sensible approach than internal leaderboards that encourage employees to compete for the highest possible AI usage. The numbers also offer a rare public data point for teams trying to forecast their own AI budgets: a large employer is willing to spend thousands of dollars per engineer per year on coding agents, but not without a ceiling.