“Something Went Wrong” in AI. Alex Karp Says Businesses Are Paying to Lose Their Competitive Edge

Quick Learning

• Karp argues token-based AI exposes business IP while delivering little value, a view that Palantir’s 85% revenue growth and 46% operating margin seem to confirm.

• Karp puts AI’s strong profits only in the compute (NVIDIA) and application sectors, yet PLTR trades at a forward P/E of 80 compared to NVDA’s 23.

• Act now: the analyst who called NVIDIA in 2010 just named his top 10 AI stocks — and Palantir didn’t make the cut. Pick up FREE words today.

Palantir Technologies (NASDAQ:PLTR) founder and CEO Alex Karp argued this week that the commercialization of productive AI in enterprises is structurally broken. In his words, “something went wrong” industry, and customers who purchase token-based access to major language models paying to expose their intellectual property and their “alpha” while receiving a small amount in return.

Karp is hardly a neutral observer. As the CEO of an AI infrastructure company that competes with enterprise spending, you will benefit if enterprises move away from token-based AI services. But his criticism shows a a broader debate is playing out across the industry about whether long-term value will be found in the supply chain model or to companies that help businesses safely deploy AI while maintaining control over their data, models, and calculations.

Karp’s says AI’s advantage lies in the compute and application layers

Karp’s main claim is that long-term profit pools in the AI ​​business reside at two ends of the stack: computer layer, where NVIDIA sells accelerators, and application layer, where Palantir sells its ontology and AIP platform. Token-based access to border models is not one of themhe argues, because clients refuse to pay the actual cost so the labs bear it “bad funds.”

Take action now: the analyst who called NVIDIA in 2010 recently named his top 10 AI stocks — and Palantir didn’t make the cut. Pick up FREE words today.

His vision for enterprise CIOs is that Palantir’s relevance is also NVIDIA (NASDAQ:NVDA) is about letting customers control their computing, models, and data stacks, and “owner of the means of production” instead of renting token recognition from a third party. He credits Palantir’s five-year start to the years he spent designing for the warfighter’s needs and building an ontology application layer that sits on top of asset models.

The ontology acts as a safeguard by preventing LLMs from storing customer data and replicating business secrets, and he says competitors are now copying the design.

Why Karp Says AI Is a National Security Problem