Palantir's billionaire founder says companies are using AI all day like a porn addiction

Palantir Technologies Chief Executive Alex Karp compared the corporate obsession with consuming as much artificial intelligence output as possible to pornography addiction, arguing that many companies are generating vast volumes of AI-produced content without delivering any meaningful business results.

Karp made the remarks June 6 during a live interview on the TBPN podcast, filmed on the sidelines of Palantir's AIPCon 10 event. The comments took aim at a practice known in technology circles as "tokenmaxxing," in which companies maximize the volume of text or data processed by AI language models — measured in units called tokens — often as a proxy for measuring AI adoption.

"People are just sitting there all day, kind of like a porn addiction," Karp said, describing what he sees as compulsive, productivity-free AI use inside enterprises. "Enterprises are like — okay, we believe this will create value, but we cannot have people just checking the weather with it, just rearranging deck chairs on their personal Titanic."

Karp said Palantir has developed an internal product specifically designed to break companies of the habit, which he referred to in characteristically unfiltered terms as a "demasturbatory, get off masturbation thing."

The remarks landed in a technology industry that has spent two years justifying enormous AI spending bills. Companies including Meta and Amazon have reportedly built internal scoreboards tracking employee token consumption, treating high usage as a measure of AI productivity. That approach has recently drawn scrutiny as rising AI infrastructure costs have failed to translate into clear returns.

Karp cited comments by Uber Chief Operating Officer Andrew Macdonald, who said the ride-share company was struggling to draw a connection between its growing AI bills and measurable productivity improvements. Karp said executives were until recently reluctant to raise such questions in public.

Palantir Chief Technology Officer Shyam Sankar had made a similar argument on a recent earnings call, describing Palantir as "a no slop zone" and warning that treating cheap AI model calls as inherently valuable was a path to low-quality outputs regardless of scale.

Karp drew a distinction between AI tasks where language models perform reliably — generating reports, summarizing information and handling routine queries — and more complex industrial and operational problems, including optimizing supply chains, managing specialized manufacturing processes and running enterprise-level decision systems. The latter category, he argued, still requires human expertise and ongoing judgment that current AI cannot replicate.

The comments reflect a position Palantir has held for some time. The company has built its business on the premise that AI must be integrated into specific operational workflows — through its AIP platform — rather than deployed as a general productivity tool. Karp has repeatedly argued that large language model providers are compelling to investors but have struggled to deliver comparable results inside actual enterprises.

Palantir's own finances have benefited from that positioning. The company's stock rose more than 141 percent in 2025, and its revenue has grown for six consecutive quarters. Karp's personal net worth stands at approximately $14.3 billion, according to Bloomberg, nearly doubling over 2025 as Palantir shares surged.

The broader technology industry is navigating a moment of reckoning on AI spending. Hyperscale companies including Meta, Microsoft, Amazon and Alphabet are projected to spend a combined $620 billion on AI-related infrastructure in 2026, according to Morgan Stanley estimates. Critics including short-seller Michael Burry and JPMorgan Chief Executive Jamie Dimon have warned that AI asset valuations have entered bubble territory.

Karp's remarks do not amount to a skeptical view of artificial intelligence overall — Palantir has staked its commercial future on AI adoption. What he is challenging is the assumption that volume of AI consumption is itself a meaningful business metric, a distinction that may become increasingly consequential as boards begin demanding clearer returns on what have in many cases been nine-figure technology investments.