Uber's COO Says AI Spending Isn't Paying Off as 'Tokenmaxxing' Debate Heats Up

A Widening Gap Between Spending and Results

Uber is confronting an uncomfortable question that is rippling across Silicon Valley: what exactly is all that AI money buying? Two of the company's most senior executives have publicly aired concerns that escalating spending on artificial intelligence tools is not translating into proportional gains in productivity or customer-facing value.

In a Rapid Response interview, Uber's chief operating officer Andrew Macdonald said it was becoming harder to justify the company's rising AI expenditures. After speaking with senior engineering leaders, he concluded that higher token usage did not produce a matching increase in useful consumer features.

The Rise of 'Tokenmaxxing'

Macdonald's comments put a name to a growing anxiety: 'tokenmaxxing', or the drive to consume as many AI tokens as possible on the assumption that more usage automatically equals more output. His remarks suggest that even firms deeply committed to AI tooling are starting to question whether raw consumption actually correlates with business value.

Budget Blown by April

The scepticism follows a striking disclosure from Uber CTO Praveen Neppalli Naga, who revealed that the company had already exhausted its entire 2026 AI coding budget just four months into the year. "I'm back to the drawing board because the budget I thought I would need is blown away already," he said.

The overshoot was driven by aggressive adoption of Anthropic's Claude Code across roughly 5,000 engineers after the tool launched internally in December 2025. Monthly API costs per engineer ranged from $500 to $2,000 for heavy users. Today, around 95% of Uber's engineers use AI tools every month, roughly 70% of committed code is AI-generated, and about 11% of live backend code updates are written entirely by AI agents with no direct human input.

Balancing Innovation and Discipline

The budget blowout is already reshaping how Uber allocates resources. Reports in May indicated the company was slowing hiring partly to help fund its AI investment. The internal debate mirrors a wider industry reckoning: an analysis published the same month found that while heavier token usage boosts raw coding output, extreme consumption delivers diminishing returns.

The tension is clear. Uber's operations chief is flagging a lack of proportional consumer benefit, even as its engineering organisation enthusiastically embraces AI coding tools. That gap underscores the central challenge facing technology companies right now: turning enormous generative AI spending into a durable, measurable competitive advantage rather than just an ever-growing bill.