Palantir Claims AI Dominance Bid as Anthropic Outspends It 43-to-1

Anthropic spent 43 times as much on research and development in the third quarter as Palantir, a gap CEO Alex Karp framed as a strength on his company's earnings call this week. Palantir's strategy, he said, is to post-train open-weight models such as NVIDIA's Nemotron on customers' operational data — building on research that AI labs and open-weight developers already paid for. He described Anthropic as a symbiotic commercial partner, not a rival in spending.

Karp also declared a "moral imperative to be the dominant player in our industry," an unusually absolute framing for a company whose R&D budget runs two orders of magnitude below the frontier labs it now positions itself against.

What does the spending gap mean?

For R&D managers, the contrast is the story. Frontier labs sink billions into foundation model training; Palantir's proposition is that post-training open-weight models on proprietary operational data captures enterprise value without that capital outlay. The model, in effect, arrives pre-funded by other companies' research budgets. Whether post-training alone can deliver the reliability enterprises demand remains unproven in the call's own terms — Palantir offered no measured benchmarks, only demand claims.

How did Karp answer the skeptics?

D.A. Davidson analyst Gil Luria, citing Anthropic's S-1 filing, noted that about half of Anthropic's customers work with the lab directly. He asked how Palantir persuades CEOs who are not yet convinced. Karp declined the premise entirely: "We really cannot satiate demand. So this is a market that is growing so rapidly. We don't need the people who aren't convinced."

Karp and Chief Technology Officer Shyam Sankar used the call to critique the broader AI market, questioning the value of rising token consumption and frontier labs' approach to enterprise productivity. Karp said customers were tiring of "token raw usage, lightweight companies reselling tokens, software companies that are reselling tokens and it counts as revenue." He added: "People are just tired of it."

He also argued that early AI investors have a structural blind spot. "It's hard to see the reality when you're getting rich on something that's not actually working the way you think it is." In industry, he said, buyers realize they need Palantir once AI "gets real."

How did the market respond?

Investors were unmoved by the rhetoric. Palantir shares fell 3.13% to $178.34 during Monday's regular session, ahead of the earnings release, then dropped another 2.64% to $173.64 in after-hours trading as of approximately 7:26 p.m. Eastern, according to Google Finance. The sell-off was sector-wide: Microsoft fell 2.83%, Meta 1.34%, NVIDIA 1.26% and Broadcom 1.18%.

The call followed a brutal late-October stretch for technology stocks. Alphabet shares fell 12.6% on October 29, the company's worst day on record. Intel plunged 28.6% the following day after a disappointing profit forecast. Microsoft opened a voluntary exit program reportedly covering about 60,000 employees, roughly 27% of its workforce. By October 26, the neocloud companies CoreWeave, Nebius and IREN had each lost more than 30% for the month amid concerns about their debt-funded expansion.

Karp opened his remarks by calling it "by far the most watched earnings call in the world" — a claim he offered without attribution, and one the market's after-hours verdict did not visibly endorse.

Palantir's bet, as laid out on the call, is that post-training open-weight models on customer data will outcompete both token resellers and frontier labs in enterprise deployments. The company did not disclose R&D figures of its own, leaving the 43-to-1 ratio — drawn from Anthropic's S-1 — as the only hard spending number on the table. Investors and R&D portfolio planners alike will now watch whether that leverage strategy holds as the AI spending correction deepens.