Palantir CEO Alex Karp harshly criticized the token-based business models of OpenAI and Anthropic, accusing them of unfairly taxing American enterprises while offering little value in return. Speaking in a CNBC interview, Karp positioned Palantir as a trustworthy alternative that prioritizes enterprise control and data security.
- Karp labels current AI token-based pricing 'effing insane' and harmful to US enterprises
- He stresses need for clear data ownership, secure prompts, and control over AI compute
- Palantir offers a model-compute-application stack aimed at protecting enterprise IP
What happened
In a recent televised interview on CNBC, Palantir CEO Alex Karp publicly criticized the business models used by major AI providers like OpenAI and Anthropic. Karp described these models as overly expensive and unfairly taxing enterprise customers by charging for tokens that he argues create no real tangible value. He conveyed significant dissatisfaction among American enterprises, stating they are 'livid' about paying substantial costs while having their intellectual property exploited.
Karp elaborated that enterprises feel they are paying a kind of 'wealth tax' that benefits neither their companies nor society broadly but instead mainly the AI labs. The criticism is focused on the large language model providers’ pricing strategy and the overhyped promises of value that do not materialize, causing erosion of trust. He positioned Palantir’s collaboration with NVIDIA on sovereign AI capabilities as a response to these concerns.
Why it matters
Karp’s remarks highlight growing unease among enterprises forced to adopt AI services based on token consumption pricing that lacks transparency and fails to safeguard their sensitive data and intellectual property. He emphasized the need for enterprises to understand who owns the data, where it is stored, and whether the AI service secures proprietary business information from being inadvertently shared or monetized by third parties.
This critique is significant because it underscores a shift in enterprise demand toward more control and trust in AI implementation. The prevailing models risk alienating the very businesses they aim to serve by transferring valuable AI-produced insights and business 'alpha' to external labs. Palantir’s stance represents a broader industry push for on-premise or sovereign AI solutions that ensure customers retain ownership and governance over critical AI resources.
What to watch next
Following Karp’s public challenge, attention will focus on how other AI providers respond to enterprise dissatisfaction over token pricing and data security. There may be increasing demand for transparent pricing models that tie directly to actual value creation rather than token usage alone. Enterprises will likely push for AI offerings that integrate secure application layers and maintain strict control over intellectual property to avoid inadvertent technology leakage.
Additionally, Palantir’s approach combining AI models, compute infrastructure, and an ontology-based application layer may attract more enterprise customers seeking trustworthy, sovereign AI solutions. The broader market could see innovation around hybrid deployment models that balance AI performance with comprehensive data governance, especially in highly regulated industries and governments concerned about AI sovereignty.