AI costs are climbing fast, requiring companies to stay measured in their approach, according to Jamie Dimon.

Speaking on CNBC on Wednesday, the JPMorgan CEO stressed the need to track token usage and evaluate returns on investment carefully.

"They all see the costs going up rapidly," Dimon said. "So, of course, we're all going to be rational about it like any other resource we use."

Dimon, who has led JPMorgan since 2006, explained that the bank routinely assesses the value delivered by AI tools and works with vendors to manage expenses.

He observed that organizations are already applying smarter strategies throughout the AI ecosystem.

"I've already seen tons of systems. People are going to send queries to the cheapest token, the cheapest thing," he said. "That's already happening, and that's going to happen with power. It's going to happen with data centers."

The bank remains committed to safeguarding its information assets, he added, responding to rising executive concerns on the topic.

"We are very protective of our data and our IP," Dimon said. "You should assume that JPMorgan will do everything they can to protect its own data, its own IP, to protect our customers."

Modelmaxxing all the way

Dimon joins other executives urging greater restraint on AI expenditures. Recent reporting has described modelmaxxing, in which companies skip defaulting to the priciest frontier models for routine work.

As bills mount, many are stepping back from tokenmaxxing, the habit of deploying numerous high-cost tools such as Claude, Codex, and Cursor without clear gains.

Palantir CEO Alex Karp has voiced strong reservations about tokenmaxxing.

In a recent CNBC discussion, Karp stated that AI offerings have been overhyped and that many firms quietly recognize payments for tokens yielding little benefit.

"Something has gone completely wrong," Karp said. "The basic view among enterprises in this country is 'I'm going to chillax and waste my time with tokens, I'm going to get no value, and they're going to get my IP,'" he said.

Last month Karp likened excessive AI reliance to a form of addiction.

"Really, we call it the demastibatory, like get off masturbation thing internally," Karp said. "It's like people are just sitting there all day, kind of like a porn addiction."

Cerebras Systems CEO Andrew Feldman has offered a similar critique of unchecked token use.

At a Bloomberg gathering last month, he called unlimited employee token access a flawed idea from the start.

"You don't need a Ferrari to go to the grocery store, right? Use a lower-cost open source model," he said. "What we're learning is how to shop at Costco."