NEW YORK, June 10, 2026, 12:02 EDT
Palantir CEO Alex Karp on CNBC Wednesday criticized frontier AI labs, saying many enterprise clients are privately dissatisfied with those companies. Karp told CNBC’s Sara Eisen that discontent isn’t just limited to everyday users. He said most of the Anthropic projects discussed in public are “running on Palantir.” Let’s Data Science
Corporate AI buyers are starting to set real budgets, not just run pilots. Business Insider said Wednesday that OpenAI, Anthropic and GitHub are steering more customers to token-based billing, where users pay for the text AI models handle. Coinbase and Walmart added usage caps, and Amazon dropped its internal token leaderboard.
The stakes are higher now as top model makers look to show they can make big corporate use pay off. Axios said Anthropic put in pre-IPO paperwork while major customers hit an AI “sticker shock” period. The report pointed to Bain numbers: 40% of almost 1,000 companies saw AI deliver cost savings of less than 10% after spending on it. Axios
Karp is pitching that unease to customers. At Palantir’s AIPCon 10, according to remarks posted by the company, he told customers to try a large language model company first and see “they don’t care about you at all.” After that, he said, when they “pay a lot in tokens,” they should come back. Palantir, he said, wants customers to “own the means of production.” LinkedIn
Karp took aim at “token maxing”—when companies show off AI activity but don’t fix actual business issues. According to Benzinga, Karp said Palantir has an internal tool that is supposed to stop enterprises from doing that. He argued more AI calls don’t mean operational change. Benzinga
Palantir CTO Shyam Sankar didn’t mince words at the event, saying, “More tokens means more slop.” CEO Alex Karp’s main argument was that big enterprise work still relies on steady, exact processes, and large language models are there to assist, not take over. Business Insider
Other AI execs have also raised cost issues, but they don’t always use Palantir’s language. Replit’s Michele Catasta, the company’s president and head of AI, described token leaderboards as “very dystopian.” Catasta said token use isn’t “proportional” to business impact and cautioned that heavy AI demand drives up energy and compute usage. Business Insider
Anthropic moves to address costs with Claude Fable 5. Reuters said Monday the new model is more powerful and charges $10 per million input tokens and $50 per million output tokens. Anthropic said lower token usage could mean cheaper tasks for some users. The model also brings new safeguards to curb cybersecurity misuse.
Analysts bullish on Palantir argue the company’s edge comes from both its software and the people behind it. Wedbush’s Dan Ives and his team reaffirmed their Outperform rating and $230 target after AIPCon, pointing to Palantir’s forward-deployed engineers—specialists who work inside customer firms—as the reason the company is “irreplaceable” for AI rollouts. BigGo Finance
The case isn’t bulletproof. If Anthropic, OpenAI or others manage to bring down the price per task quickly, Palantir’s position as a check on spending could seem less important. Governments might also see Palantir as a risky dependency, holding up contract approvals. UK officials are examining a 330 million pound NHS data deal with Palantir now, after lawmakers raised concerns about patient confidentiality, trust, and overreliance on a U.S. supplier.
Palantir traded just under 1% higher at $132.66 late Wednesday morning in New York. The stock’s small gain valued the company near $341 billion. Karp’s latest comments didn’t appear to move the shares, with the AI spending story not turning into a new spark for Palantir in this session.
