Industry Leaders Criticize ‘Tokenmaxxing’ as Inefficient AI Metric Executives across tech and finance are turning against “tokenmaxxing,” the push to use as many AI tokens as possible, arguing it inflates costs and distorts what real productivity looks like.

Over the past year, Silicon Valley embraced tokenmaxxing as a sign of aggressive AI adoption. But enterprise users quickly began questioning whether soaring AI bills were producing meaningful returns. Uber’s COO publicly wondered about the link between rising AI spend and useful outcomes, crystallizing wider doubts in the industry.

By early June, Anthropic president Daniela Amodei was distancing her company from hard‑pressure tactics. Speaking at a Bloomberg tech conference, she noted there is “a lot more distance to go” in what models will be able to do over the next several years and said Anthropic does not run AI leaderboards or mandate usage, instead hoping AI becomes “incorporated into the day-to-day of how humans do our work” in ways that feel good to employees.

Financial institutions soon added their own critique. At BNP Paribas CIB, AI chief Charles Holive labeled tokenmaxxing a “vanity metric,” saying his team “try to make sure that what we track is an outcome, not a vanity metric,” focusing instead on questions like “What did you do that you didn’t do before? How much faster did you do it?”

Big tech employers also began to reverse course. Amazon dropped an internal AI-usage leaderboard after discovering employees were inflating token consumption, shifting to tracking deployments that show “real code shipping.”

Criticism has sharpened in recent days. Palantir CEO Alex Karp dismissed tokenmaxxing culture as people “just sitting there all day,” likening it to a kind of compulsive behavior that confuses raw AI usage with real value creation. Legora CTO Jacob Lauritzen went further, calling tokenmaxxing “a really stupid way to do anything” and urging companies to reward employees “for being effective and efficient and having more output, not for necessarily using AI.”

Together, these voices signal a shift from measuring AI by how much is used to how much it actually improves work.

Continue reading https://foxvector.com/stories/019ea83b-14c1-16e0-72f4-326d2bfd59a7