Overview
- Palantir and NVIDIA publicly launched the intelligent engine on June 29, 2026 to let U.S. government agencies and critical infrastructure operators run NVIDIA Nemotron open models inside sovereign, air-gapped, and classified environments.
- The offering integrates NVIDIA Nemotron and NVIDIA AI Enterprise runtimes with Palantir’s AIP, Ontology, Foundry and Apollo to provide model training, deployment and operational workflows on customer infrastructure.
- Built governance features include explicit data authorization, architecturally enforced tenant isolation, secure perimeter controls, data portability, a documented right to erasure, and full audit trails for data and model artifacts.
- The engine collects telemetry and trace data to post-train and align models so they continually improve for mission tasks, while customers can change model weights and retain full ownership of resulting models.
- Palantir and NVIDIA disclosed no named procurement contracts or deployment timelines, and agencies that adopt the stack will face upfront GPU, staff and lifecycle costs that shift responsibility for security, upkeep and evaluation onto operators.