Alex Karp, CEO of Palantir, recently launched a harsh critique of leading AI companies OpenAI and Anthropic during a national television appearance. In two minutes, he essentially accused them of basing their business model on "stealing" intellectual property from Fortune 500 corporations under the guise of a subscription.

What is the essence of the complaints?

Karp's key blow was aimed at the token-based pricing model. He posed a reasonable question: if a product is so valuable that it can bring a client a billion dollars, why not take 30% of the result? Why charge for tokens if the value is so great?

According to Karp's logic, if OpenAI and Anthropic's models truly delivered the claimed productivity gains, the labs would take an equity or profit share. Instead, they sell access by the million tokens. Paying per token is, in essence, an admission that the product cannot deliver reliable value at scale. If it could, it would be valued based on that, not on the volume of computations.

Karp went further, calling the entire scheme a "wealth tax that doesn't help the poor, only punishes." In his view, American business is handing over the alpha of its operations—workflows, customer data, strategic notes—directly into the training pipelines of a few labs.

The mechanism Karp warns about

Karp's key thesis is that after retraining, the client's advantage turns into a new corporate product sold to their competitors. A company feeding confidential documents through a model is potentially teaching it how to replace itself.

This is why Karp uses the word "theft": the provider gets both the token fees and the accumulating insights into the client's business. The Palantir CEO explains this as the reason executives are furious in private conversations but remain silent publicly.

This raises a fundamental question: what are corporations actually paying for—computations or value? The answer directly impacts valuations across the entire industry. If the position described by Karp reflects the real sentiments of big business, it calls into question the sustainability of the subscription model for leading AI companies.

My expert opinion: Karp's criticism is not just an emotional outburst but a precise hit on the most painful point of the current AI monetization paradigm. The market is still valuing labs based on potential, but if corporate clients begin to massively renegotiate terms, we will see a serious correction in valuations across the entire sector. It's a matter of time, not probability.