Palantir and NVIDIA on July 1 unveiled a Sovereign AI initiative built to run NVIDIA Nemotron open-weight models on Blackwell Ultra GPUs inside Palantir's Sovereign AI Operating System reference architecture — a stack designed for U.S. government agencies and critical-infrastructure operators that need to keep sensitive data on-premises, on-edge or fully air-gapped.
Ownership, Not Just Access
Under the deal, Palantir customers can deploy, customize and post-train Nemotron models on their own data while retaining full ownership of the resulting model weights — a sharp break from the API-only postures of OpenAI and Anthropic. NVIDIA supplies the acceleration layer and open models; Palantir supplies the ontology, Foundry data fabric, AIP orchestration and Apollo continuous delivery. Together they form what the partners call the Sovereign AI OS Reference Architecture.
Wall Street Reprices Palantir
Palantir shares closed July 1 at $125.73, up 7.8% on unusually heavy 57M-share volume — a roughly $22B one-day market-cap gain. Q1 revenue hit $1.6B, up 85% year-over-year with U.S. commercial up 133% and U.S. government up 84%. Management raised full-year guidance to $7.65-7.66B, implying ~71% growth, and lifted U.S. commercial guidance above $3.22B. Adjusted free cash flow reached $925M, a 57% FCF margin.
Positioning Ahead Of A $177B Market
The stack extends existing federal wins including the Space Force TCAM, the DoD's Titan program and the JADC2 data lattice. Precedence Research forecasts the sovereign AI infrastructure market at $177B by 2035 at a 28% CAGR. Combined with adjacent moves — BlackRock and MGX's Aligned Data Centers push, the Japan-NVIDIA Noetra AI factory, and the Microsoft-Mistral Azure sovereign cloud — sovereign compute is quickly becoming a distinct product category rather than a hyperscaler afterthought.
Reporting based on coverage from 24/7 Wall St., Yahoo Finance and Palantir Investor Relations.