Artificial intelligence has become a key investment area for businesses, but the rising cost of deploying advanced AI models is prompting many organisations to reassess their spending. As enterprises allocate larger budgets to AI, questions are emerging over whether current pricing models offer sufficient value. Palantir CEO Alex Karp believes many companies are becoming dissatisfied with the token-based pricing adopted by AI firms such as Anthropic and OpenAI.
Speaking on CNBC's Squawk Box on Wednesday, Karp criticised the industry's reliance on charging customers according to token consumption, arguing that the approach is frustrating enterprise users.
"I'm not throwing shade at them, but 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."
His comments come as the cost of operating increasingly sophisticated AI models continues to climb. With every new generation of AI systems demanding greater computing resources, businesses are paying closer attention to whether their AI investments generate measurable returns rather than simply increasing token usage.
The rising costs have also fuelled interest in open-weight AI models, which allow organisations to customise models for their specific requirements without relying entirely on commercial AI providers. Because these models can perform many enterprise tasks at significantly lower costs, they are becoming an increasingly attractive option for businesses seeking to reduce AI expenditure.
According to Karp, enterprises are also demanding greater control over the technology powering their AI systems. Rather than depending exclusively on third-party providers, many organisations want ownership of their computing infrastructure, AI models and data.
"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. "They want to know they own the means of production. It's not being transferred to someone else."
Palantir has already begun pursuing that strategy. Earlier this week, the company announced an expanded partnership with Nvidia to use the chipmaker's AI technology for developing customised AI models for US government agencies. The collaboration is focused on building AI systems tailored to specific organisational needs rather than relying solely on general-purpose foundation models.
Karp also cautioned against underestimating China's progress in artificial intelligence, saying Chinese companies are advancing rapidly and intensifying competition in the global AI race.
At the same time, many enterprises are increasingly developing proprietary AI models designed for specific business functions instead of relying exclusively on large, general-purpose systems. These specialised models are often more efficient, easier to customise and less expensive to operate.
Shortly before Karp's television appearance, Palantir published a nine-point manifesto on X centred on what it described as "AI sovereignty." The company criticised the "tokenmaxxing" business model and encouraged organisations to retain ownership of their data rather than surrendering greater control to AI service providers.
Investors appeared to welcome Palantir's approach, with the company's shares rising 8 per cent on Wednesday.
Karp's remarks add to the growing debate over the future of enterprise AI, as businesses increasingly weigh the capabilities of expensive frontier AI models against customised alternatives that offer greater control, flexibility and lower operating costs.
