In the first half of 2026, Uber was a central driver of the tokenmaxxing phenomenon in the workplace. But now the company's chief technology officer, Praveen Neppalli Naga, argues that the tokenmaxxing era is winding down.
On Wednesday, Naga pointed on X to "very interesting trends on AI costs" and described them as "another signal that we're coming to the end of the so-called 'tokenmaxxing' era."
What Is Tokenmaxxing, Exactly?
Tokenmaxxing refers to a workplace AI movement that appeared during the first six months of 2026: businesses encouraged employees to use AI tools heavily in their daily routines. At some firms, AI adoption was even built into performance reviews.
Uber ignited the trend. In April, Naga said that Uber had already spent its full 2026 budget for Anthropic's Claude Code. A March LinkedIn post from Naga said its internal coding agent had autonomously composed 1,800 code changes per week.
But by May, Uber's chief operating officer, Andrew Macdonald, was telling interviewers that Uber was finding it harder to justify AI investment trade-offs. The executive added that the productivity gains from higher AI costs were not proportionate.
Get the free Always Be Buying eBook and learn the simple system for building wealth on any income
The trend was never just about engineering. But costs quickly became the dominant concern.
Uber's own budget overrun for Claude Code was an early warning sign, and Macdonald's later comments highlighted the gap between usage and value. The move toward efficiency, in that light, is less a retreat from AI than a maturation of it.
The Numbers Behind the Pivot
The count of Uber employees using leading-edge AI tools has quadrupled since the start of 2026, Naga said, while the cost per AI token has fallen. Naga credited the decline to measures such as improved prompt caching, stronger default model choices, better AI usage visibility for engineers, and trials with open-weight models.
"The next phase, whatever we call it, will not be characterized by who spends the most tokens, but about how people use them as efficiently as possible," he added.
Balaji Krishnamurthy, Uber's CFO, delivered a similar message on Wednesday's second-quarter earnings call.
"On AI, we are very early, but what we are seeing is that we are able to cost-efficiently deliver some productivity lifts with developers," he said. "And for the measurement that we are looking at right now, we are seeing doubling in the code output for engineers."
What It Means for Investors
The move from raw token volume to output-based value is part of a broader cost-control push. Uber's own experience shows that usage can climb quickly, but that better model selection and infrastructure measures can bring per-token costs down. For investors, the key question is whether these productivity gains continue to show up in metrics like code output while AI expenses stay under control.
Uber is hardly alone. Across the tech sector, companies have struggled to show stronger returns on massive AI budgets. Coinbase, for instance, has said it is trying different models.
Download the free Always Be Buying eBook and start putting your money to work today
