The tokenmaxxing backlash has begun

• Uber’s COO went viral after saying rising AI costs haven’t led to proportional productivity gains.
• Others now say the same thing as tech companies blow through their AI budgets early.
• Tokenmaxxing is the practice of burning through as many AI tokens as possible.

Silicon Valley’s debate over tokenmaxxing is reaching new heights.

Uber COO Andrew Macdonald said in an interview released last week that he hasn’t observed direct productivity improvements from increased AI token usage.

“That link is not there yet, right?” Macdonald said in comments that went viral, racking up over 2 million views on X. “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, ‘OK, now we’re actually producing 25% more useful consumer features.'”

AI tokens are the basic building blocks processed by AI chatbots, making up roughly ¾ of a word each. “Tokenmaxxing” refers to using as many tokens as possible to boost productivity — and show off.

Macdonald’s remark struck a nerve because companies across the US are aggressively adopting AI internally. Meta now calls some employees “AI builders” and expects them to work in AI-native “pods.” Corporate giants like Disney and JPMorgan track employees’ use of AI. Visa rewards teams that build faster with AI and has bragged that its monthly token spend is almost 2 trillion.

As corporations race to embrace AI, an increasing number of tech professionals say it’s leading to massive waste. The AI push has sparked concerns about companies blowing through their budgets early. Uber used up its annual AI budget in the first four months of the year, the Information reported.

“Pretty sure 50% of internal token spend is completely useless, but right now it’s hard to know which 50%,” posted Akshat Bubna, cofounder and CTO of AI startup Modal, on X.

“Tokens got burned for millions of dollars without any real significant ROI to show for it,” engineering manager Karthik Hariharan posted, referring to return on investment.

The issue is only getting worse, Google CEO Sundar Pichai recently said. At Google’s flagship developer conference, I/O, last week, Pichai said he’s heard from chief information officers that they’re “so concerned about how much their companies are blowing through budgets.”

“I think the problem is going to get worse as we go through the year,” Pichai added.

The tokenmaxxing debate has also led to concerns that the AI bubble is about to pop. On Monday, famed “Big Short” investor Michael Burry called tokenmaxxing a “crazy, rushed, temporary phase” and said that Nvidia stock has a high risk of an “aggressive” fall.

Tokenmaxxing has its defenders, though.

Garry Tan, the CEO of fabled San Francisco investment firm Y Combinator, has embraced the term, saying, “we’ve been tokenmaxxing longer than most people.”

There’s also a balanced way to take advantage of AI’s benefits without spending too much, according to a report from engineering intelligence company Jellyfish.

The report found that the top 10% of Claude Code users consumed about 10 times as many AI tokens as the median developer and produced only about twice the output.

To resolve this, companies should not reward or penalize raw token consumption. Instead, they should make sure costs are tied to concrete metrics, such as pull requests — how developers propose code changes to shared projects — the report said.