AI is getting expensive. Chamath Palihapitiya says he thinks CFOs may be about to discover those costs the hard way.
Palihapitiya, a billionaire venture capitalist and one of the co-hosts of the “All In” podcast, said Tuesday on CNBC that businesses encouraging employees to consume more AI could face surprise operating expenses — and even earnings misses — as AI-use bills add up.
“The CEOs and the CFOs, in my opinion, probably have no idea how much tokenmaxxing is going on inside of their organizations,” Palihapitiya said. He predicted top executives across the US could miss earnings expectations.
That rising price pressure, Palihapitiya said, is coming as premium models from OpenAI and Anthropic face growing competition from lower-priced options that are narrowing gaps in quality.
Tokenmaxxing is Silicon Valley shorthand for using as many AI tokens as possible — the theory is that AI will produce more work, faster. Tokens are units of data processed by AI models and commonly used to calculate customer bills.
Some firms that embraced AI dove headfirst into the AI-optimist-deep end: several introduced company-visible leaderboards and incentives to show which workers were using the tools the most.
Now, the bill is becoming harder to ignore, and Palihapitiya says leaders are adjusting. For example, Uber’s CTO Praveen Neppalli Naga said in April that the company had already exhausted its full-year Claude Code budget, while Instagram CEO Adam Mosseri said the company shut down “the silly things” that were burning tokens.
“It is not that hard to build a token incinerator,” Mosseri said during a recent episode of “Lenny’s Podcast.”
At the same time, Palihapitiya told CNBC that tech mainstays like Meta, Google, and SpaceX — which he said have fewer computing-capacity constraints — are closing the gap. They’re releasing cheaper models that are “80 to 95% as good,” he said.
“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.’ So of course, I’m going to go to jet fuel.”
Instead, he likened recent AI updates to successive generations of the iPhone: They’re still improving, but are no longer introducing massive technological step changes.
When asked whether lower-cost models from Grok and Meta were as good as Anthropic’s, he said: “For most use-cases, the answer is a screaming yes.”
Palihapitiya didn’t immediately respond to a request for comment from Business Insider.
Read the full article here
