Silicon Valley’s backlash against tokenmaxxing has officially begun.
Uber Chief Operating Officer Andrew McDonald said in an interview published last week that he has not observed any direct productivity gains from increased usage of AI tokens.
“You don’t have that link yet, right?” McDonald said in a comment that went viral online, racking up more than 2 million views on X. “I think there’s probably implicitly more stuff being shipped, but it’s very hard to draw the line between one of these statistics and, ‘Okay, now we’re actually producing 25% more useful features for the consumer.'”
AI tokens are the basic building blocks processed by AI chatbots, and each make up about 3/4 of a word. “Tokenmaxxing” refers to using as many tokens as possible to increase productivity and show off.
MacDonald’s comments came as a shock as companies across the U.S. are aggressively implementing AI within their companies. Meta now calls some of its employees “AI builders” and expects them to work in AI-native “pods.” Major companies like Disney and JPMorgan are tracking their employees’ use of AI. Visa rewards teams that build faster with AI and boasts monthly token spend of nearly 2 trillion.
As companies race to implement AI, a growing number of technology experts say it’s creating massive amounts of waste. There are concerns that the push for AI will cause companies to quickly exhaust their budgets. The Information reported that Uber used up its annual AI budget in the first four months of this year.
Akshat Bubna, co-founder and CTO of AI startup Modal, wrote in a post on
“Tokens burned millions of dollars without really showing any significant ROI,” engineering manager Kartik Hariharan posted, referring to the return on investment.
Google CEO Sundar Pichai recently said the problem is only getting worse. Pichai said last week at Google’s flagship developer conference, I/O, that he heard from chief information officers who were “very concerned about how much their budgets are being blown.”
“I think the problem will only get worse over the course of this year,” Pichai added.
The tokenmaxxing debate also led to concerns that the AI bubble might burst. On Monday, prominent “Big Short” investor Michael Burley called token maxing a “crazy, rushed, temporary step” and said Nvidia stock was at high risk of a “severe” decline.
However, Tokenmaxxing has its defenders.
Garry Tan, CEO of Y Combinator, a well-known investment firm in San Francisco, accepts this saying, saying, “We’ve been doing token maxing longer than most people.”
There are also ways to balance the benefits of AI without spending too much, according to a report from engineering intelligence company Jellyfish.
According to the report, the top 10% of Claude Code users spent about 10 times as many AI tokens as the median developer and only generated about twice as many. output.
To solve this, companies should not reward or penalize the consumption of raw tokens. Instead, costs should be tied to tangible metrics such as pull requests, a way for developers to propose code changes to a shared project, the report said.