Palantir deploys NVIDIA Nemotron in air-gapped US govt AI stack

Palantir announced a new intelligent engine that uses NVIDIA Nemotron open models to serve U.S. government agencies in air-gapped (fully isolated) environments. The offering combines NVIDIA's open-source foundation models with Palantir's Sovereign AI Operating System—built on AIP, Ontology, Foundry, and Apollo—to enable government agencies to run customized models on their own infrastructure without external network access.

Agencies can train models on their own data, retain full ownership of model weights and operational knowledge, and achieve full auditability—key requirements for national security and data sovereignty. The Nemotron stack runs on NVIDIA accelerated computing and enterprise-grade deployments can tap NVIDIA AI Enterprise software suite. Open models allow agencies to inspect, modify, and deploy AI in sensitive environments where closed commercial models may breach data security or privacy laws.

The partnership reflects broader U.S. strategy to democratize AI access while maintaining control. With ~3 million civilian employees, the U.S. government operates like one of the world's largest enterprises, and open models with transparency help address trust concerns. The combination of Nemotron's customizability and Palantir's infrastructure software creates a data flywheel: as agencies deploy models in production, they can continually improve them using new data within their own secure environments.

For architects: this validates open-model infrastructure as production-grade for mission-critical, regulated workloads—inference on sensitive data where transparency and control outweigh closed-API convenience. Expect more govt-scale deployments to follow, locking in on-premise inference economics for both Palantir and NVIDIA in the public sector.