Uber COO Says It's Getting Harder to Justify the Cost of 'Tokenmaxxing'

Uber's COO Andrew Macdonald said it's becoming increasingly difficult to justify whether AI spending is delivering results proportional to its cost. In particular, after Uber's CTO revealed the company had already used up its entire 2026 Claude Code budget, the tradeoff between token consumption and headcount has become a major topic of discussion internally.

After discussions with engineering leaders, Macdonald concluded that there is still no clear link showing that more token usage leads to a proportional increase in useful consumer features. The less directly AI spending connects to actual product outcomes, the harder it becomes to justify that cost. As a result, Uber's CEO Dara Khosrowshahi said the company is slowing down hiring to offset its AI investment.

Currently, Big Tech companies are showing a trend of tokenmaxxing, tying AI usage to performance reviews and promotions, but responses vary across companies — for instance, Duolingo withdrew its decision to factor AI usage into performance evaluations after pushback from employees.