Altman: AI token count has limits as a metric for corporate performance; customers prefer token-based billing

OpenAI CEO Sam Altman said AI token counts are an inadequate measure of the value enterprises derive from artificial intelligence, speaking in a "Cheeky Pint" interview with Stripe co-founder John Collison, who characterized tokens as "fake" from a business-value perspective. Altman responded that tokens are "real in some sense but not a good unit," noting that tokens—small data units models process when reading inputs and generating outputs, distinct from cryptocurrency tokens—measure usage and cost but don't reveal business outcomes. The comments matter because token-based pricing and consumption metrics are widely used to gauge AI monetization across the industry, and skepticism from a leading AI executive could fuel debate over usage-based billing models. The article also notes customers reportedly prefer token-based billing, suggesting pricing structures may persist despite the measurement critique. Watch for whether AI providers shift toward outcome-based or value-based pricing models and how enterprise clients respond to alternative billing frameworks.