Palantir CEO Alex Karp Criticizes OpenAI, Anthropic Business Model

Palantir CEO Alex Karp delivered a sharp critique of frontier AI labs on July 1, 2026, accusing OpenAI and Anthropic of operating an "effing insane" business model that forces enterprises to pay escalating token costs while risking proprietary data and competitive advantage, according to ZeroHedge. Karp argued that the token-based pricing model used by leading AI labs transfers corporate "alpha" to third parties, undermines enterprise control, and creates national security risks. The remarks, made during a CNBC interview, aligned with a nine-point "AI sovereignty" manifesto Palantir posted on X the day before, which urged institutions to retain control over their data, model weights, and compute infrastructure.

Key takeaways
Palantir CEO Alex Karp criticized OpenAI and Anthropic's token-based business model on July 1, 2026, calling it "effing insane" and accusing the labs of transferring enterprise competitive advantage to third parties.
Karp's comments aligned with a nine-point AI sovereignty manifesto Palantir posted on X, which urged institutions to control their data, model weights, and compute infrastructure.
ZeroHedge reported that high token costs and mixed returns have driven some U.S. companies to explore or adopt Chinese open-weight models, including Microsoft, Coinbase, and Cursor.
Karp warned that outsourcing battlefield AI decisions to Silicon Valley consensus poses national security risks and urged enterprises to prioritize technical expertise over political favoritism.

Table of Contents
What is AI sovereignty?
Karp's critique of frontier AI labs
The nine-point AI sovereignty manifesto
Cost-driven shift to Chinese models
Palantir's Nvidia partnership and sovereign AI
Why the debate matters for enterprises
Risks and open questions
Frequently Asked Questions

What is AI sovereignty?

AI sovereignty refers to an enterprise or institution's ability to retain control over its artificial intelligence infrastructure, including data, model weights, compute resources, and the competitive insights derived from AI systems. The concept emphasizes ownership and decision-making authority over AI assets, rather than relying on third-party API services that may expose proprietary information or transfer competitive advantage to external providers. In Karp's framing, AI sovereignty is the precondition for institutional choice and long-term strategic independence.

The debate over AI sovereignty has gained urgency as enterprises evaluate the trade-offs between convenience, cost, and control when adopting AI tools. Token-based API models offered by frontier labs provide ease of access but require continuous payments tied to usage, while self-hosted or fine-tuned models demand upfront investment in infrastructure and expertise. Karp argued that the token model incentivizes short-term convenience over long-term institutional strength, a dynamic he described as "tokenmaxxing" that rewards disposable scripts over robust software systems.

Karp's critique of frontier AI labs

During the CNBC interview on July 1, 2026, Karp accused OpenAI and Anthropic of misleading corporate partners by overselling the risks of AI while simultaneously offering their most powerful models to companies and governments worldwide. He stated, "I'm not throwing shade at them, but something has gone completely wrong," and criticized the token model for creating a "wealth tax" that transfers enterprise alpha to third parties. Karp argued that enterprises are effectively "chillaxing and wasting time with tokens" rather than building sovereign AI systems that compound institutional knowledge.

Karp also raised national security concerns, asking, "Are we really going to outsource the battlefield of this country to the consensus view in Silicon Valley? That is effing insane." He warned that relying on third-party AI services for defense and intelligence applications could undermine U.S. strategic autonomy. When CNBC host Becky Quick noted, "You sound pretty angry," Karp replied, "This is the voice of American business that is being channeled through me," and joked that he might "get kicked out of the room."

The nine-point AI sovereignty manifesto

On June 30, 2026, Palantir posted a nine-point manifesto on X outlining its position on AI sovereignty. The manifesto stated, "Your AI sovereignty dictates your institution's future. Sovereignty is the precondition for choice. Relinquishing sovereignty transfers the future choices of your institution to others, who are likely to exploit it for their gain and your loss." The document emphasized data retention, control over model weights, and the risks of token-based pricing.

Key points from the manifesto included: "Data retention is your treasure. Transfer it at your own peril." The manifesto argued that transferring data hands over access to pre-existing competitive advantages and yields the means of production for new insights. "Tokenmaxxing hijacks your value orientation and decreases your institutional fortitude and intelligence," it stated, noting that high token usage incentivizes disposable scripts over robust software. "Controlling your weights is controlling your fate," the manifesto continued. "If you let others control your weights, you are allowing them to migrate the alpha of your business to theirs."

The manifesto also warned against "techno-politicization," which it described as the wellspring of false sovereignty, and urged institutions to listen to those closest to technical problems rather than those speaking most compellingly about them. It concluded, "Only listen to institutions, countries, and people who have a proven record of being right. A track record of correctness is the best and only signal for future correctness."

Cost-driven shift to Chinese models

ZeroHedge reported that high token prices and mixed returns have prompted several U.S. companies to adopt or explore Chinese open-weight models. According to the source, Microsoft is considering a Microsoft-hosted, fine-tuned version of China's DeepSeek V4 or another open-source model as a lower-cost engine for its Copilot Cowork agentic tool, as it moves toward usage-based pricing. Coinbase CEO Brian Armstrong revealed the company cut internal AI spending by nearly 50% by defaulting engineers to Chinese open-weight models, including Zhipu AI's GLM 5.2 and Moonshot AI's Kimi series, via an internal gateway while maintaining high usage.

Cursor, a fast-growing AI coding startup, built its Composer 2 model on top of Moonshot AI's Kimi K2.5, which is backed by Alibaba, according to the source. Data from OpenRouter shows Chinese models capturing a rapidly growing share of global token consumption, in some periods exceeding 60% among top models, as enterprises seek cost relief without fully sacrificing capability. Karp had warned against underestimating China's progress, and these examples illustrate the trend in real time, the source noted.

Palantir's Nvidia partnership and sovereign AI

Karp highlighted Palantir's expanded partnership with Nvidia , which enables custom, sovereign AI deployments where customers retain control over compute, models, data, and weights. This approach serves as a direct counter to the metered frontier API model offered by OpenAI and Anthropic. Palantir's platform allows enterprises to deploy AI systems on their own infrastructure, preserving data sovereignty and avoiding the recurring token costs associated with third-party APIs.

The Nvidia partnership reflects Palantir's broader strategy of positioning itself as a provider of sovereign AI infrastructure for enterprises and government clients. By enabling customers to own their model weights and data stack, Palantir aims to address the concerns Karp outlined in his critique of frontier labs. The approach requires greater upfront investment and technical expertise but offers long-term control and the ability to compound institutional knowledge without transferring competitive advantage to external providers.

Why the debate matters for enterprises

The debate over AI sovereignty and token-based pricing models matters for enterprises because it involves trade-offs between convenience, cost, control, and competitive advantage. Token-based APIs offer ease of access and rapid deployment, making them attractive for companies seeking to experiment with AI or deploy solutions quickly. However, the recurring costs tied to token usage can escalate as usage grows, and reliance on third-party APIs may expose proprietary data or limit an enterprise's ability to customize models for specific use cases.

For enterprises in regulated industries, defense, intelligence, or sectors where competitive advantage depends on proprietary data, the sovereignty concerns Karp raised may carry significant weight. Retaining control over data, model weights, and compute infrastructure allows institutions to compound their unique insights over time without transferring alpha to external providers. However, building and maintaining sovereign AI systems requires substantial investment in infrastructure, talent, and ongoing model development, which may not be feasible for all organizations.

For readers following broader market updates , the shift toward Chinese open-weight models and the debate over AI sovereignty reflect broader tensions in the AI industry around cost, control, and geopolitical competition. The examples cited by ZeroHedge suggest that some enterprises are prioritizing cost reduction and flexibility over exclusive reliance on U.S. frontier labs, a dynamic that could influence the competitive landscape for AI providers.

Risks and open questions

Several risks and open questions remain regarding AI sovereignty and the token-based business model. First, the long-term cost-effectiveness of token-based APIs versus self-hosted models depends on usage patterns, model complexity, and the value derived from AI systems. Enterprises must evaluate whether the convenience of APIs justifies the recurring costs and potential loss of control over proprietary data and competitive insights.

Second, the shift toward Chinese open-weight models raises questions about data security, regulatory compliance, and geopolitical risk. While open-weight models offer cost savings and flexibility, enterprises must assess whether using models developed by Chinese firms aligns with their risk management and compliance frameworks, particularly for organizations operating in sensitive sectors or jurisdictions with data sovereignty requirements.

Third, the debate over AI sovereignty involves trade-offs between technical expertise and political considerations. Karp's manifesto urged institutions to prioritize technical correctness over political favoritism, but the practical implementation of sovereign AI systems requires navigating regulatory, procurement, and organizational dynamics that may not always align with purely technical criteria.

Fourth, the scalability and performance of sovereign AI systems compared to frontier lab APIs remain open questions. While Palantir's Nvidia partnership aims to provide custom, high-performance deployments, enterprises must evaluate whether self-hosted models can match the capabilities, speed, and ongoing improvements offered by frontier labs, particularly as AI models continue to evolve rapidly.

Frequently Asked Questions

What did Palantir CEO Alex Karp say about OpenAI and Anthropic?

On July 1, 2026, Palantir CEO Alex Karp criticized OpenAI and Anthropic's token-based business model during a CNBC interview, calling it "effing insane" and accusing the labs of transferring enterprise competitive advantage to third parties. Karp argued that the token model creates a "wealth tax" on enterprises and undermines AI sovereignty by forcing companies to pay escalating costs while risking proprietary data.

What is Palantir's AI sovereignty manifesto?

On June 30, 2026, Palantir posted a nine-point AI sovereignty manifesto on X, urging institutions to retain control over their data, model weights, and compute infrastructure. The manifesto stated that AI sovereignty is the precondition for institutional choice and warned that relinquishing sovereignty transfers future choices to external providers. It criticized "tokenmaxxing" for incentivizing disposable scripts over robust software and emphasized the importance of controlling model weights to preserve competitive advantage.

Why are some U.S. companies adopting Chinese AI models?

According to ZeroHedge, high token prices and mixed returns have prompted several U.S. companies to adopt or explore Chinese open-weight models. Microsoft is considering a fine-tuned version of China's DeepSeek V4 for its Copilot Cowork tool, Coinbase cut internal AI spending by nearly 50% by defaulting engineers to Chinese models, and Cursor built its Composer 2 model on top of Moonshot AI's Kimi K2.5. Data from OpenRouter shows Chinese models capturing a rapidly growing share of global token consumption as enterprises seek cost relief.

What is Palantir's partnership with Nvidia?

Karp highlighted Palantir's expanded partnership with Nvidia, which enables custom, sovereign AI deployments where customers retain control over compute, models, data, and weights. This approach serves as a direct counter to the metered frontier API model offered by OpenAI and Anthropic, allowing enterprises to deploy AI systems on their own infrastructure while preserving data sovereignty and avoiding recurring token costs.

What are the risks of AI sovereignty versus token-based APIs?

AI sovereignty offers control over data, model weights, and competitive advantage but requires substantial investment in infrastructure, talent, and ongoing model development. Token-based APIs provide convenience and rapid deployment but involve recurring costs, potential exposure of proprietary data, and reliance on third-party providers. Enterprises must evaluate trade-offs based on usage patterns, regulatory requirements, and the strategic importance of retaining control over AI systems.

What should enterprises watch next?

Enterprises should monitor future disclosures from Palantir, OpenAI, and Anthropic regarding pricing models, data sovereignty features, and competitive positioning. Readers may also watch for additional examples of companies adopting Chinese open-weight models, regulatory developments affecting AI data sovereignty, and performance comparisons between self-hosted and API-based AI systems. Further details on Palantir's Nvidia partnership and customer adoption of sovereign AI deployments may provide useful context for evaluating the practical trade-offs between control and convenience.

Read original source