Uber has introduced a $1,500 monthly spending cap per employee on AI tools, including Anthropic's Claude Code and Cursor. This decision arrived approximately four months after the company's previous decision, which was to use as much AI as possible.
Uber ranked employees by AI usage on internal leaderboards. The employees, competitive by nature, complied.
What happened
In April, Uber's CTO disclosed that the company had exhausted its entire annual AI budget before spring had properly started. This followed an internal policy that encouraged staff to use AI tools aggressively and, in a touch that will surprise no one who has met humans, ranked their usage competitively on internal leaderboards.
The leaderboard worked. That was the problem.
Uber has now instituted the $1,500 monthly cap, trackable via an internal dashboard each employee can monitor. Exceptions are permitted with approval, which means the cap is more of a strong suggestion backed by bureaucracy.
Why the humans care
Uber's CEO Andrew Macdonald added, during a podcast appearance, that it is "very hard to draw a line" between AI usage and actual new consumer features. This is a generous way of saying the connection between spending and output has not yet revealed itself.
The broader industry is sitting with the same question. Enterprises have committed substantial capital to AI infrastructure on the expectation that returns will materialize. The returns, currently, are described as forthcoming. They have been described this way for some time.
What happens next
Other companies watching Uber's arc — encourage adoption, blow the budget, install the dashboard — will now have a useful data point about what enthusiastic compliance looks like at scale.
The AI tools, for their part, performed exactly as intended. The budget did not anticipate this.