Chamath Palihapitiya warns soaring AI token spend will trigger unexpected earnings misses

CEOs Face Blind Spot on AI Spending

Tech investor Chamath Palihapitiya issued a stark warning this week that soaring AI token spend could blindside corporate executives and trigger unexpected earnings misses 1

1

. The Social Capital founder told CNBC that CEOs and CFOs "probably have no idea how much tokenmaxxing is going on inside of their organizations," predicting a scenario where earnings per share falls short by a few pennies and leadership scrambles to understand what happened 2

2

.

Tokenmaxxing refers to corporate policies that push employees toward maximum AI usage under the assumption that higher consumption drives greater productivity. AI vendors typically charge by the token—discrete chunks of data that AI models consume when generating responses—making token volume a direct driver of enterprise costs 2

2

. Palihapitiya, who leads the AI company 8090 and co-hosts the All-In podcast, suggested that unchecked AI spending has grown so far out of view that executives will have to contend with ballooning costs tied to AI model usage "they didn't know existed inside of their organization" 1

1

.

Companies Already Pulling Back from Token Burn Era

Palihapitiya's comments arrive amid a broad industry retreat from aggressive tokenmaxxing strategies. 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 2

2

. Microsoft pulled back employee access to Claude Code, while Meta's CTO Andrew Bosworth told staff in an April memo that token usage alone does not measure impact 2

2

.

Palihapitiya disclosed in March that his own company 8090, which raised $135 million in a funding round led by Salesforce in June, was trending toward more than $10 million a year in AI spending 2

2

. His warning echoes recent criticism from Palantir Technologies CEO Alex Karp, who argued earlier this month that OpenAI and Anthropic had fundamentally mispriced their AI services and that enterprise customers were generating little value from token spending 2

2

.

Premium AI Models Lose Edge as Alternatives Converge

Adding pressure to the cost-efficiency equation, Palihapitiya noted that competitive dynamics around premium AI models are shifting rapidly. Cheaper alternatives from companies including Meta and Google are now "80 to 95% as good" as leading models for most use cases, he said 2

2

. The performance gap between successive AI model releases now resembles the incremental improvements of iPhone generations rather than transformative advances. "It used to be the case that when a model dropped, it was so superior to everything else," Palihapitiya explained. "You're like, 'Oh my God. We went from kerosene to jet fuel'" 2

2

.

This convergence suggests companies may face mounting scrutiny over whether premium AI subscriptions justify their costs when cheaper alternatives deliver comparable results. Palihapitiya's warning extends beyond immediate financial risk, pointing to a longer-term reckoning where AI tokenmaxxing will hurt earnings and force organizations to rethink deployment strategies. Palihapitiya, who launched a new SPAC called American Exceptionalism Acquisition Corp. A (AEXA) last year targeting companies in AI, energy, defense and decentralized finance, remains a controversial figure in Silicon Valley due to his role promoting SPACs during the pandemic 1

1

.