Palantir and Nvidia launch sovereign AI stack for US government security

The collaboration addresses significant concerns regarding data sovereignty, particularly for defense sectors that cannot rely on closed commercial models due to risks of data leakage to third-party providers. By utilizing open models, agencies can fine-tune AI on proprietary data without exposing model weights or operational insights. The architecture supports on-premises, edge, and cloud deployments, featuring explicit authorization protocols, customer isolation, and full auditability to meet strict national security standards.

The hardware foundation relies on Nvidia Blackwell Ultra systems with 8 GPUs per node and Spectrum-X networking. A key addition to this expanded partnership is the integration of supply chain optimization capabilities using Nvidia’s cuOpt engine. Nvidia has already deployed this stack internally to manage its global supplier network, which involves handling the complexity of 1.3 million parts per AI rack, demonstrating the system’s practical utility in high-stakes logistics environments.