
Karp delivers sharp criticism of AI companies' pricing model
During a live interview on CNBC's "Squawk Box" in early July 2026, Palantir's CEO Alex Karp made what the financial press quickly dubbed an "eleven-word bombshell": "I am paying for tokens that create no value."
According to Yahoo Finance, this was no offhand jab, but a precise critique of the business model used by major AI laboratories such as OpenAI and Anthropic, where enterprise customers are billed per token — that is, per unit of text processed by the AI model.

What is the problem with token-based pricing?
Karp argues that many businesses are accumulating ever-rising costs for AI tokens without being able to document a corresponding return. Even more serious, he contends, is that companies risk handing over sensitive internal data and competitive advantages to the very vendors they are paying.
To describe the phenomenon, he coined the term "tokenmaxxing" — borrowed from crypto slang — to characterize what he calls an irrational obsession with selling as many tokens as possible without guaranteeing customers any real results.
"Tokenmaxxing" is Karp's term for AI companies' obsession with selling tokens — regardless of whether customers actually get anything in return for their money.
He drew a parallel to what he calls a "wealth tax": companies pay not only in money, but effectively surrender their unique knowledge and IP to the operators running the models.

Palantir itself delivers record numbers
Karp's decision to go on the offensive is no coincidence. Palantir wrapped up its best quarter ever in Q2 2026, according to the company's own reports.
The US commercial segment grew particularly strongly — up 149 percent to $764 million — driven by demand for the company's AI platforms. According to Yahoo Finance, Karp has set a goal of maintaining or exceeding this growth rate over the next 18 months.
Palantir's model differs fundamentally from what Karp criticizes: the company does not sell API access per token, but integrated data platforms where customers retain control over their own data and models — a concept the company calls "AI sovereignty."
What does this mean for the stock market?
Karp's remarks should be read as more than internal PR. They represent a direct positioning maneuver in a market where many investors are still pricing in growth at token-based AI providers without questioning profitability.
In a risk-off environment — where the Fear & Greed Index currently sits at 30 out of 100 — this kind of fundamental criticism of growth companies' revenue models is especially sensitive. Investors with exposure to OpenAI-related stocks or upcoming IPOs in the AI segment should note that the market's underlying assumptions are being challenged from within the industry itself.
Palantir's own relationship with digital assets
It is worth noting that Karp himself leads a company with its own ties to the crypto sector. Palantir announced as early as May 2021 that it would accept Bitcoin as a means of payment, and CFO David Glazer stated at the time that holding crypto on the balance sheet was "absolutely something we are considering," according to available company information.
This means that Karp's criticism of "worthless tokens" must be viewed in light of the fact that Palantir itself operates at the intersection of AI, data, and digital assets — and that the company has a commercial interest in presenting its own approach as superior.
The key takeaway for investors
Karp's remarks have not been verified by independent analysts in the source material available to 24markets, and the claims about competitors' lack of ROI represent Palantir's own characterization. Nevertheless, it is a signal the market should take seriously: when one of the heaviest hitters in the enterprise AI segment asserts that token-based pricing fails to deliver value, it is a question that will increasingly be raised in boardrooms and by CFOs around the world.
For equity analysts, this means that the earnings quality of companies with purely token-based revenue may come under fresh scrutiny — regardless of how impressive the growth figures look on the surface.
This article was written using large language models under editorial supervision by Aprex. Content is source-verified and auditable. Read our method →