Key Takeaways
- • Palantir CEO Alex Karp publicly criticized the token-based pricing model used by leading AI firms (OpenAI, Anthropic), calling the industry "effing insane" and declaring "something has gone completely wrong" with how AI is being sold to enterprises.
- • Karp argues that token-based AI models expose enterprise intellectual property while delivering insufficient value, and that skyrocketing token costs are forcing companies to abandon this approach in favor of open-weight models and efficiency-focused alternatives.
- • Palantir announced a new partnership with Nvidia to develop frontier AI models, positioning the company as an alternative to the prevailing token-maximization strategy in the AI industry.
- • Enterprise customers are reportedly fed up with AI labs that have oversold models and prioritized "tokenmaxxing" over genuine utility, according to Karp's characterization of market sentiment.
- • The critique suggests structural problems in how major AI providers monetize their services, with Karp contending that token-based economics are fundamentally misaligned with enterprise needs and security requirements.
Summary
Palantir Technologies CEO Alex Karp delivered a forceful public critique of the AI industry's dominant business model during a CNBC appearance on July 1, 2026, characterizing the token-based pricing strategies employed by OpenAI and Anthropic as fundamentally broken. Karp declared that "something has gone completely wrong" with how leading AI companies charge for their services, and described the broader industry approach as "effing insane," signaling deep frustration with current market practices.
The core of Karp's argument centers on the structural flaws of token-based pricing. He contends that this model forces enterprises to expose sensitive intellectual property while providing limited tangible value in return. Additionally, Karp highlighted that escalating token costs are driving corporate customers away from token-dependent systems toward open-weight models and efficiency-first alternatives. This shift reflects growing enterprise dissatisfaction with the economics of current AI offerings, wherein companies pay per token consumed regardless of actual business outcomes.
Karp characterized enterprise sentiment as increasingly skeptical, noting that customers feel "oversold" by AI labs that have prioritized what he terms "tokenmaxxing"—maximizing token consumption to increase revenue—over delivering genuine utility. The CEO positioned this critique as evidence that the industry has lost sight of practical value creation for end users.
In tandem with this public criticism, Palantir announced a strategic partnership with Nvidia focused on developing frontier AI models. This initiative appears designed to offer an alternative approach to the prevailing token-maximization model, signaling Palantir's intent to differentiate itself by addressing the structural inefficiencies Karp identified. The partnership announcement, coupled with Karp's unfiltered assessment of industry problems, suggests Palantir is positioning itself as a corrective force in enterprise AI adoption, targeting companies frustrated with incumbent providers' pricing and delivery models.
The core of Karp's argument centers on the structural flaws of token-based pricing. He contends that this model forces enterprises to expose sensitive intellectual property while providing limited tangible value in return. Additionally, Karp highlighted that escalating token costs are driving corporate customers away from token-dependent systems toward open-weight models and efficiency-first alternatives. This shift reflects growing enterprise dissatisfaction with the economics of current AI offerings, wherein companies pay per token consumed regardless of actual business outcomes.
Karp characterized enterprise sentiment as increasingly skeptical, noting that customers feel "oversold" by AI labs that have prioritized what he terms "tokenmaxxing"—maximizing token consumption to increase revenue—over delivering genuine utility. The CEO positioned this critique as evidence that the industry has lost sight of practical value creation for end users.
In tandem with this public criticism, Palantir announced a strategic partnership with Nvidia focused on developing frontier AI models. This initiative appears designed to offer an alternative approach to the prevailing token-maximization model, signaling Palantir's intent to differentiate itself by addressing the structural inefficiencies Karp identified. The partnership announcement, coupled with Karp's unfiltered assessment of industry problems, suggests Palantir is positioning itself as a corrective force in enterprise AI adoption, targeting companies frustrated with incumbent providers' pricing and delivery models.
Go deeper than the headline
Research every firm in this story, track the ones that matter and get the daily briefing — free for 7 days.