Palantir CEO criticizes AI labs for ‘tokenmaxxing,’ backs enterprise control of AI
Karp calls out OpenAI and Anthropic by name, saying CEOs are 'livid' about token-based pricing models that drain budgets while exposing proprietary data
Palantir Technologies CEO Alex Karp said the era of “tokenmaxxing” is fading as enterprises seek lower-cost, higher-value AI deployments amid rising model expenses.
Speaking to CNBC, Karp criticized the token-based pricing approach used by leading AI labs, saying businesses are increasingly focused on generating returns instead of consuming more AI tokens.
He said enterprises are turning toward open-weight models and developing proprietary AI systems that offer greater efficiency, while maintaining control over their data, computing resources and AI models. Karp argued that ownership of the AI technology stack is becoming a strategic priority for corporate customers.
Karp also cautioned that China is making rapid progress in AI development and should not be underestimated.
His comments follow Palantir’s expanded collaboration with Nvidia to build customized AI solutions for US government agencies and the company’s newly released “AI sovereignty” framework promoting enterprise control over AI resources.