Palantir CEO Alex Karp on Wednesday criticized the token model used by US artificial intelligence lab Anthropic and OpenAI amid rising costs.
“I’m not throwing shade at them, but something is completely wrong,” he told CNBC’s “Squawk Box.” “The basic idea of companies in this country is, I’m going to sit back and waste my time with tokens.”
As the cost of AI rises and new models prove more expensive than previous versions, companies are moving from a so-called “token max” mindset to one that prioritizes return on investment.
This setup has led some companies to adopt open-weight models that can perform similar tasks at a fraction of the price. The capabilities of Chinese models are also accelerating, raising concerns that AI rivals will soon catch up to the US Frontier Institute.
Karp told CNBC that the industry should not underestimate the speed of China’s progress in building AI models.
In this environment, many companies are also moving from using extensive AI models to building and training their own more efficient tools.
Earlier this week, Palantir announced an expansion of its partnership with. Nvidia The idea is to use the semiconductor manufacturing giant’s AI tools to build custom models for U.S. government agencies.
Karp sees the open-weight model as a potential solution for CEOs who are dissatisfied with their AI labs.
“I think where we align with Nvidia is what our technical customers want: control of the compute, the models, the data stack, the alpha,” Karp said. “They want to know that they own the means of production. The means of production are not transferred to someone else.”
—CNBC’s Seema Mody contributed to this article.
