비즈니스와 금융

팔란티어, OpenAI 클라우드 모델 거부하며 93% 매출 성장 보고

Victor Maslow

The number that stopped Wall Street mid-session was $1.935 billion — nearly double what Palantir collected in the same quarter a year earlier, and $123 million above what analysts had forecast. What followed was a 29.5% surge in the stock and, according to chief executive Alex Karp, proof of something the company has argued for a decade: that governments would eventually refuse to cede control of their AI to the companies building it.

“Our customers have declined to become vassal states of the language labs,” Karp told investors after the earnings release. The language labs he named were OpenAI and Anthropic — the two companies best positioned to dominate government AI procurement through cloud-based subscriptions. The alternative Palantir offers is different in kind: AI deployed inside a government’s own classified infrastructure, running models without sending data to an outside server. The quarter showed that the distinction is now a purchasing decision, not a philosophical one.

Government revenue grew 90% year-over-year to $809 million, but the more striking figure was on the commercial side. Enterprise revenue jumped 149% to $764 million, far outpacing the government segment and extending Palantir beyond its defense origins into private-sector industries where data sovereignty is increasingly a regulatory and competitive requirement, not just a procurement preference.

The Maven Smart System, Palantir’s AI platform operating within US classified defense networks, reached a milestone the company had been building toward for years: the first officially designated program of record launched on the platform in the quarter, with more than 25,000 military and civilian builders now developing applications within it. In a detail that illustrates both Karp’s rhetoric and its limits, Palantir has integrated Anthropic’s Claude models into Maven — running Claude inside Palantir’s classified environment, not on Anthropic’s servers. The arrangement lets the company attack the AI labs publicly while quietly using their models under controlled conditions.

The skepticism on Wall Street has not disappeared. Palantir’s 93% growth rate is partly a function of favorable comparisons against a period when its commercial AI pivot was still early, and those comparisons will tighten as base effects normalize. The full-year guidance of $8.154 billion — implying roughly 80% annual growth — relies on both government and commercial customers sustaining their current pace of expansion. Activist short-sellers had bet against that scenario; the quarter triggered an estimated $3 billion in short-covering, forcing positions to close at a loss rather than at a vindicated thesis.

For those working across government and enterprise technology, the AI sovereignty model Palantir is selling is not neutral. A defense-grade AI stack that requires no public cloud infrastructure also requires fewer of the middleware and integration roles that support cloud-based AI deployments. Palantir’s headcount grows more slowly than its revenue — a pattern the company has explicitly engineered — and the model it is selling to governments mirrors how it intends to run itself: smaller staff, greater automation.

Palantir raised its full-year 2026 revenue guidance to $8.154 billion, representing roughly 80% growth. Third-quarter results are expected in late October 2026.

태그: , , , , ,

토론

댓글 0개가 있습니다.