Palantir CEO Criticizes Token-Based AI Models Amid Rising Costs
Palantir chief Alex Karp said the token pricing model used by leading U.S. AI labs is becoming unsustainable for enterprises, prompting a shift toward open-weight and proprietary models.
Palantir chief executive Alex Karp told CNBC's "Squawk Box" that the token-based pricing approach employed by U.S. AI firms such as Anthropic and OpenAI is no longer viable for many companies as costs continue to rise. He said enterprises are moving away from a focus on maximizing token usage toward solutions that deliver clearer return on investment.
Karp noted that some businesses are turning to open-weight models, which can perform comparable tasks at lower expense, and that several firms are developing their own custom models to retain control over compute, data and algorithms. He added that the rapid progress of Chinese AI developers should not be underestimated.
The comments came as Palantir announced an expanded partnership with Nvidia to leverage the chipmaker's AI tools for building tailored models for U.S. government agencies. Following Karp's remarks, Palantir's shares rose about 9%.
"What aligns me with Nvidia, and I think is what the technical customers want, which is control over their compute, their models, their data stack and their alpha," Karp said.
CNBC contributor Seema Mody also worked on the story.
