Palantir CEO Criticizes AI Industry, OpenAI and Anthropic Pricing Models

Palantir CEO Alex Karp publicly criticized the token-based pricing models used by AI labs OpenAI and Anthropic, telling CNBC that as costs climb 'something has gone completely wrong'. Karp argued the underlying models have been 'irresponsibly oversold,' and said some enterprise executives privately report that heavy token spending yields little useful return while exposing proprietary data.

His comments capture a shift in enterprise sentiment away from 'tokenmaxxing' toward return on investment, with some customers turning to cheaper open-weight models for comparable tasks. The timing is pointed: on June 29, PLTR and Nvidia NVDA unveiled a Sovereign AI Operating System that runs Nvidia's open Nemotron models on air-gapped Blackwell Ultra hardware for US government agencies, giving customers ownership of model weights and data.

Karp's critique doubles as positioning for Palantir's deployment-and-ownership pitch against per-token consumption pricing. For investors, the debate matters because it bears on the unit economics of the entire applied-AI stack. Markets will watch whether enterprise budgets keep migrating toward open-weight and on-premise approaches, a trend that would favor Palantir's model while pressuring the token-metered revenue of the large labs.