Palantir and Nvidia Create Sovereign AI Stack for On‑Prem Government Use

Built to run large language models on isolated government systems, the deal tests whether Palantir can convert its government backlog into steady revenue.

Overview

Palantir announced a strategic partnership with Nvidia to integrate Nvidia’s Nemotron models and Blackwell Ultra GPU support into Palantir’s AIP, Foundry, Ontology and Apollo stack for on‑premises and air‑gapped U.S. government deployments.
The companies describe the combined offering as an “intelligent engine” that lets agencies train and run LLMs without sending sensitive data to closed commercial model weights.
The announcement on Monday produced a modest stock lift of about 4–4.6%, but Palantir’s shares remain well below their 50‑ and 200‑day moving averages and are down sharply year‑to‑date.
Neither company named agency customers or disclosed financial terms, and investors will watch Palantir’s August quarterly results for signs the firm can turn such sovereign deals into recurring, margin‑preserving revenue.
The partnership follows Palantir’s strong Q1 performance and large backlog and highlights both a strategy of productizing secure, third‑party model integrations and the execution risks tied to reliance on outside hardware and international contract headwinds.