The CEO of Palantir, Alex Karp, publicly criticized the sales of the AI industry, stating that many corporate clients were tired of model services based on token fees. Businesses fear spending money without receiving a clear return and may also hand over their data and business advantages to model providers.
Business starts questioning returns.
Karp, in an interview with CNBC, said that a growing number of corporate executives had expressed similar concerns in private: on the one hand, AI continued to increase its service expenditures; on the other hand, the actual business value was not clear. He summed up the situation as a “waste of time and token” by enterprises, without the corresponding output.
He also mentioned that clients were most concerned not only with costs but also with data control. For many enterprises, internal data, processes and industry knowledge are in themselves competitive advantages. Once these elements are exposed to external modelling platforms for long periods, enterprises may lose their original market leadership.
Palantir stresses the sovereignty of AI.
Karp's statement echoes a nine-point proposition recently published by Palantir. This document emphasizes “AI sovereignty” and, at its core, means that the enterprise should, to the extent possible, retain control over the way in which data and models are deployed and avoid sunking key data into third-party platforms when using external AI services.
In his view, what the AI industry currently needed most was confidence-building, not continued outreach. According to him, some of the model products are “totally and irresponsibly oversold” in terms of capacity, landing speed and commercial returns, which is weakening the patience of business clients.
Military model disagreement is raised again.
The issue of proximity was also addressed this month by the Chief Executive Officer of Microsoft Satya Nadella. He warned that different industries should not, in the course of adopting AI, quickly commercialize their own knowledge systems and eventually lose their original barriers.
Karp also extended the topic to national security and military applications in interviews. He mentioned that it was a serious question whether key battlefield capabilities should be left to the Silicon Valley mainstream. This statement also echoes the recent disagreement between Anthropic and the United States Government over the way the model was used in some military scenarios.
There has also been some reflection on token spending in recent times. A number of companies that had previously procured AI services had begun to re-examine their budgets, requiring clearer efficiency indicators and proof of return. As enterprises become more cautious in their procurement, the future pressures on AI model producers may no longer be just technical competition, including pricing, deployment modalities and customer trust.