Uber’s COO questions AI tokenmaxxing value

AI tokenmaxxing – Uber’s operations chief Andrew Macdonald says the company is finding it harder to justify the money it spends on AI token-heavy work, pointing to a growing gap between token usage and the consumer features delivered. The comments come after Uber’s CTO sparked

For Uber’s engineering teams, more AI tokens used to sound like progress. Now the message coming from the top is getting harder to ignore: the extra cost isn’t automatically translating into extra customer value.

In a Rapid Response interview released on Saturday. Uber operations chief Andrew Macdonald said it was becoming “harder to justify” the company’s AI costs. The blunt assessment landed in the middle of a wider corporate debate over tokenmaxxing — the push to use AI as much as possible. often measured through how many tokens systems consume.

Macdonald’s comments drew a direct line to a moment that had already unsettled internal conversations. He said Uber CTO Praveen Neppalli Naga went viral after telling The Information in an April interview that Uber had already blown through its Claude Code budget for 2026. Macdonald described that reaction inside the company as a “head-exploding moment. ” and it triggered fresh discussion about AI token consumption and the trade-offs it can force — including effects on head count.

As Macdonald spoke with Uber’s senior engineering leaders. he said he came away with the same uncomfortable conclusion: higher token usage did not produce a proportional rise in useful consumer features. He challenged the internal narrative with a question that sounded more like frustration than theory.

“That link is not there yet, right?” he said. “I think maybe implicitly there is more that is getting shipped, but it’s very hard to draw a line between one of those stats and, ‘Okay, now we’re actually producing 25% more useful consumer features.’”

He added that the trade-off costs from AI were harder to justify because he could not point to a direct, measurable connection between what the company is spending and what customers actually receive.

The timing matters. Earlier this month. CEO Dara Khosrowshahi said in an earnings call that Uber was slowing hiring to counter its investments in AI. Taken together. Macdonald’s remarks suggest that as token-driven spending rises. the cost pressure is reaching beyond engineering budgets and into workforce decisions.

Macdonald also addressed the illusion that AI costs are invisible. AI can look “free” to the people using it — “just a user sitting there coming up with interesting use cases” — but the bill ultimately lands with the company.

Outside Uber, the debate is playing out in starkly different directions. While Big Tech pushes tokenmaxxing and even evaluates employees based on their AI usage. some companies are pulling back on the incentive structure. Duolingo. for instance. walked back its decision to include AI usage in performance reviews after employees asked whether they had to use AI for the sake of using it.

Duolingo CEO Luis von Ahn said in a podcast interview in April, “It felt like, rather than being held accountable for the actual outcome, we were trying to just push something that in some cases did not fit.”

The tension across these stories is hard to miss: token-heavy AI can be framed as momentum. but when teams cannot clearly connect those token counts to customer-facing gains. the spending starts to feel negotiable — and the consequences start to show up in the most human parts of a company. from hiring plans to how performance is judged.

Uber Andrew Macdonald Praveen Neppalli Naga AI tokenmaxxing Claude Code token usage Dara Khosrowshahi AI costs hiring slowdown Duolingo Luis von Ahn performance reviews consumer features