OpenAI and Anthropic Built AI Around Tokens. Palantir’s Karp Says ‘Something Has Gone Completely Wrong’ With The Model.

Palantir CEO Alex Karp is escalating his criticism of Silicon Valley’s biggest artificial intelligence companies, accusing OpenAI and Anthropic of promoting a business model that encourages companies to spend heavily on AI tokens without delivering meaningful value.

Speaking during an appearance on CNBC’s Squawk Box on Wednesday, Karp argued that enterprise customers are growing increasingly frustrated with token-based pricing, in which businesses are charged according to the amount of text processed by large language models rather than the outcomes those systems produce.

“Something has gone completely wrong,” Karp said. “The basic view among enterprises in this country is, ‘I’m going to chillax and waste my time with tokens, I’m going to get no value, and they’re going to get my IP.'”

The remarks mark Karp’s latest and most direct attack on OpenAI and Anthropic, two of the dominant players in generative AI. Although Palantir partners with several AI model developers, including OpenAI, it increasingly competes with them for lucrative enterprise contracts by offering software that helps governments and corporations deploy AI systems securely within their own operations.

The Palantir CEO also accused leading AI developers of exaggerating what today’s large language models can realistically accomplish for businesses. “The reason this is happening is because these models have been completely, irresponsibly, oversold,” he said.

That changing sentiment over tokens has fueled growing criticism of what some industry observers call “tokenmaxxing,” the practice of maximizing token consumption because it increases revenue for AI providers, regardless of whether customers receive proportional business benefits.

Karp also reiterated concerns about the growing influence of major AI companies in national security, warning against allowing Silicon Valley’s prevailing views to shape military decision-making.

“Are we really going to outsource the battlefield of this country to the consensus view in Silicon Valley? That is effing insane,” Karp said during the interview. The comments align with Palantir’s broader strategy of promoting what it describes as “AI sovereignty,” encouraging organizations to maintain control over their own data instead of relying heavily on outside AI providers.

The company has argued that proprietary information is one of an organization’s most valuable assets and should not become the cost of accessing artificial intelligence.
Karp has previously praised OpenAI CEO Sam Altman and Anthropic CEO Dario Amodei as thoughtful leaders, but his public criticism has intensified in recent months.

Earlier this year, he argued that many AI companies “don’t understand how unlikeable they are” and suggested some competitors believe customer problems will disappear simply by building larger, more capable models. He has also dismissed attempts to replicate Palantir’s enterprise approach as “a complete farce.”