Uber’s CTO Says the ‘Tokenmaxxing’ Era Is Ending as AI Costs Fall

Uber’s chief technology officer, Praveen Neppalli Naga, says the company’s AI bills are falling even as far more of its employees use the tools, and he thinks that points to the end of what the industry has been calling “tokenmaxxing.”

“As our CFO mentioned at earnings today, we’re seeing some very interesting trends on AI costs,” Naga wrote on X on Aug. 5, after Uber’s second-quarter results. “I think it’s another signal that we’re coming to the end of the so-called ‘tokenmaxxing’ era.”

Tokenmaxxing is the habit of spending ever more on AI, measured in the tokens models consume, on the assumption that more usage means more value, as The Next Web put it. Uber knows it from the inside. The company burned through its entire 2026 budget for Anthropic’s Claude Code by April.

Usage up, costs down

According to Naga’s post, as reported by India Today, the number of employees using advanced AI tools has more than quadrupled since the start of the year, yet the company’s AI costs are coming down. “You might expect costs to rise as adoption accelerates. We’ve seen the opposite,” he wrote.

He said the savings didn’t come from restricting access. Uber cut unnecessary AI requests, improved prompt reuse, gave engineers better visibility into their own usage and spending, and tested different models to find the most cost-effective one for each task. “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,” Naga wrote.

The earlier doubts

The budget overrun followed an internal leaderboard that ranked teams by total AI tool usage, Fortune reported in May. Uber president and chief operating officer Andrew Macdonald said on the Rapid Response podcast that he couldn’t yet tie the higher usage to better products. “That link is not there yet,” he said. “Maybe implicitly there’s 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 like 25% more useful consumer features.'”

CEO Dara Khosrowshahi said on an earlier earnings call that about 10% of the company’s committed code is built by autonomous agents, according to Fortune.

Uber isn’t alone in rethinking AI spend. Atlassian has started putting its engineers on AI budgets, per The Next Web. At Microsoft, Jay Parikh, the executive who runs the CoreAI organization, sent a memo telling engineers to review their individual AI spend, Alex Heath reported at Sources. “Tokenmaxxing is not what we are optimizing for,” the memo said.

Uber is also moving AI beyond engineering. It has started embedding AI engineers in finance, legal, HR, marketing and customer support teams to redesign workflows, Business Insider reported, and Naga said some of those processes have gone from hours or days to minutes.