Uber caps AI coding tool spending at $1,500 per engineer per month

Simon Willison flags new reporting from Bloomberg that Uber will cap engineer spending on AI coding tools at $1,500 per tool per month, applied to products like Claude Code and Cursor. With most teams adopting two tools, the effective annual ceiling is around $36,000 per engineer. Willison does the math against Uber's reported median engineer compensation of $330,000 and notes the cap works out to about 11 percent of total comp before benefits.

Willison calls this a sensible response to an emerging problem: token-priced coding agents do not have a natural ceiling, and a small number of heavy users can run usage that surprises a CFO. He shares his own consumption as a reference point, around $1,000 a month per provider across Anthropic and OpenAI subscriptions, which would leave him roughly $500 of headroom under Uber's policy. The point is that the cap is high enough to support real work, not a punitive number.

The deeper signal is in the price tag itself. Uber is putting a public number on what AI tools are worth to it on a per-engineer basis. That is useful data for anyone trying to forecast their own AI tool budgets or to push back when procurement asks for one.

Why it matters

If you run an engineering org, this gives you a concrete anchor when you set your own tool budget. If you are an engineer running up a large API tab, the conversation about caps is coming to your company too. Build a personal sense of what your usage costs now, before someone else does the accounting for you.