Palantir and NVIDIA have launched a strategic initiative to run custom Nemotron open models directly within secure, air-gapped perimeters for national security, defence, and critical infrastructure operators.
Palantir Technologies and NVIDIA have launched a strategic initiative to deliver an intelligent engine that runs NVIDIA Nemotron open models within sovereign environments. The partnership specifically targets U.S. government agencies, national security, defense, and critical infrastructure operators, allowing them to fully own their AI systems rather than calling third-party, closed-source models via an API.
By deploying frontier-quality Nemotron open models directly within classified, air-gapped environments, the system completely isolates data from public networks. This eliminates the risk of sensitive operational insights migrating into the weights of public commercial models.
The joint framework natively integrates NVIDIA’s accelerated compute infrastructure and the NVIDIA AI Enterprise software suite (including NVIDIA NIM microservices) with Palantir’s primary critical infrastructure applications: AIP, Foundry, Ontology, and Apollo. To continuously optimize performance, the platform tracks user telemetry and operational traces to create a secure “data flywheel.” This feedback loop enables agencies to train and realign models on their own infrastructure while retaining total ownership of the model weights.
The enterprise application focuses on three distinct phases:
-
Deployment Engineering: Delivering base and customized Nemotron models safely inside classified, air-gapped perimeters.
-
Context Engineering: Enabling agencies to optimize prompts, context structures, and overall agentic workflow behaviors in production.
-
Model Engineering: Empowering government operators to directly update model weights based on localized data, evaluations, and mission outcomes.
