AI Token Spending Is About to Blindside Corporate Bosses, Warns Investor Chamath Palihapitiya

Key points

• Investor Chamath Palihapitiya warned in May 2025 that hidden AI spending will cause companies to miss their profit targets.
• Palihapitiya's own AI company, 8090, projects it will spend more than $10 million a year on AI, which he described as "very scary" for a small startup.
• 8090 raised $135 million in a funding round led by Salesforce in June 2025.
• Palantir CEO Alex Karp separately criticised OpenAI and Anthropic this month for pricing AI by the token, a model he says is wasting enterprise money.

Imagine your phone bill arriving and it is ten times what you expected, because everyone in the house was streaming 4K video all day without telling you. That is roughly the situation Chamath Palihapitiya thinks is heading for corporate boardrooms across America.

Palihapitiya, the founder of investment firm Social Capital and CEO of AI startup 8090, told CNBC Tech that company bosses have no idea how much "tokenmaxxing" is happening inside their own organisations. Tokenmaxxing means encouraging staff to use AI tools as heavily as possible, where the cost is charged per token, a small unit of text or data that AI systems process. More usage equals more tokens equals a bigger bill.

"CEOs and CFOs probably have no idea how much tokenmaxxing is going on inside of their organizations," he said. His prediction: one quarter, profits will come in lower than expected, and leadership will scramble to find out why.

For ordinary employees, the picture is almost upside-down. Many workers have been actively pushed to use AI tools more, often without anyone tracking the cost carefully. The bill lands on a different desk.

Should companies be worried about their AI spending?

Yes, especially smaller ones. Palihapitiya said his own startup's AI costs are heading past $10 million a year, and he found that alarming. He wrote in a post on X earlier this year that many companies are "feeding this revenue ramp without getting any meaningful ROI," meaning return on investment, the basic question of whether the money spent is actually earning more back.

He is not alone in sounding this alarm. Palantir CEO Alex Karp, whose company builds data-analysis software for governments and large businesses, pointed to the same problem this month. Karp criticised the token-based pricing models used by OpenAI and Anthropic, the two biggest AI labs, saying the general attitude among large American businesses has become one of paying for tokens without seeing real results.

The concern matters beyond the stock market. If companies start cutting AI budgets sharply, that could slow the rollout of AI tools that some workers genuinely find useful, from writing assistants to customer-service bots.

Palihapitiya's 8090 is building a platform where teams work alongside AI agents, software that can handle multi-step tasks on its own, to build business software. The startup raised $135 million in a Salesforce-led funding round in June 2025, so there is clearly still appetite for AI investment.

But appetite and discipline are different things. The spending wave is real. The returns are still being counted.