Venture capitalist Chamath Palihapitiya says CEOs and CFOs have no idea how much AI token spending is accumulating inside their companies
Craig T Fruchtman / Getty Images
Tech investor Chamath Palihapitiya warned on Tuesday that AI token spending inside large companies has grown so far out of view that executives could be caught off guard by earnings misses they did not see coming. The venture capitalist and Social Capital founder made the remarks in an interview with CNBC.
"CEOs and the CFOs, in my opinion, probably have no idea how much tokenmaxxing is going on inside of their organizations," Palihapitiya told CNBC. "I suspect what'll happen is one day you're going to have a miss, and EPS will be off by a few pennies, and the CEO will say to the CFO, 'What happened?'"
Tokenmaxxing is the term for corporate policies that push employees toward maximum AI usage, premised on the idea that higher consumption translates into greater productivity. AI vendors typically charge by the token — discrete chunks of data that models consume when generating a response — making token volume a direct driver of enterprise costs.
Palihapitiya also said that competitive pressure on premium AI models is intensifying. Cheaper alternatives from companies including Meta $META and Google $GOOGL are now "80 to 95% as good" as leading models for most use cases, he said, and the performance gap between successive AI model releases now resembles the incremental improvements of iPhone generations rather than step-change advances.
"I think that you are seeing a convergence; it used to be the case that when a model dropped, it was so superior to everything else," Palihapitiya said. "You're like, 'Oh my God. We went from kerosene to jet fuel.'"
Palihapitiya is also the CEO of 8090, an enterprise software company that announced a $135 million funding round led by Salesforce $CRM in June, and a co-host of the "All-In" podcast. He said in March that his own company's AI spending was trending toward more than $10 million a year.
His comments land amid a broad pullback from the tokenmaxxing era, during which some of the largest companies in the world prioritized burning AI tokens with little attention to returns. Uber $UBER burned through its entire annual Claude Code allocation well ahead of schedule and subsequently imposed a $1,500-per-developer spending ceiling on individual tools. Microsoft $MSFT pulled back employee access to Claude Code. Meta's CTO Andrew Bosworth told staff in an April memo that token usage alone is not a measure of impact.
Palantir $PLTR Technologies CEO Alex Karp made similar criticisms earlier this month, arguing that OpenAI and Anthropic had fundamentally mispriced their AI services and that enterprise customers were generating little value from token spending. Palihapitiya's warning echoes that view, extending it to the financial risk now accumulating on corporate income statements.
Join 500,000+ readers who start their day with Quartz.
By subscribing, you agree to our Terms of Service and Privacy Policy.
