Sam Altman Questions AI Token-Based Pricing, Calls Tokens ‘Fake’ Business Units

Speaking with Stripe, co‑founder John Collison on the Cheeky Pint and Stripe podcast Altman said tokens are not a way to measure the real business value that AI models deliver. He explained that while tokens show how much of an AI model is used, they do not always show how useful work the model actually completes.

“Tokens are a kind of unit. I mean Tokens are real in some sense. Tokens are not a good unit. Tokens are fake, from a business value point of view,” Sam Altman said.

His comments show there is more talk, about how AI companies charge money. This is happening because more and more companies and people use AI tools every day. Altman thinks that people should pay for what they get done instead of paying based on something that is not real like how much they use it.

Why customers still prefer token-based pricing

Interestingly, Altman acknowledged that customers are not necessarily comfortable moving away from token-based pricing. Even though he questions its usefulness, he said users prefer having a clear unit to understand what they are paying for.

"I would rather not make up a fake unit,” Sam Altman said, explaining that customers often feel uncomfortable receiving bills without a clearly defined measurement.

“But the customer response on that is so negative. They're like, 'I need to know what I'm paying for somehow," he said.

Altman also shared an example of asking customers whether they would prefer paying more for identical AI output if the pricing was calculated using a consistent number of tokens. Surprisingly, customers preferred the familiar measurement, even if the token count was artificially increased.

"I asked a few people, 'Would you rather pay more for the same output, but have it priced in some number of consistent tokens?' Even if we inflated the tokens, they're like, 'Oh, I'd much rather you inflate the tokens and pay more, because I know there's a real thing there.”

He further explained that customers are concerned about unpredictable costs when there is no clear pricing structure.

"People don't just want us to give them a price, because they feel like over the long term they're going to get screwed somehow."

Jensen Huang has a different perspective

Altman's comments contrast with Nvidia CEO Jensen Huang's earlier proposal to provide engineers with AI tokens as an additional workplace benefit.

Huang suggested offering engineers token allowances worth around half their base salary, arguing that access to AI tools could significantly improve productivity.

“I’m going to give them probably half of that on top of [their base pay] as tokens ... because every engineer that has access to tokens will be more productive,” Huang said.

The contrasting views highlight how AI tokens are becoming an important part of conversations around pricing, employee productivity and the business value of artificial intelligence.