Palantir Brings NVIDIA Nemotron Open Models to US Government Agencies

Palantir has introduced a new intelligent engine built on NVIDIA Nemotron open models to serve US government agencies, NVIDIA said in a June 29 blog post.

Per the post, the launch positions open models for deployment in closed, secure environments — the air-gapped and classified settings typical of government work. NVIDIA framed the launch as a case for open-source innovation in American AI.

The material reviewed contained no contract details, no named agencies and no technical specifics about the engine itself — its architecture, its accreditation status and where it sits in Palantir's product line were all undisclosed.

The announcement is confirmed as NVIDIA's account; Palantir's own description of the engine was not part of the material reviewed. Which agencies will actually run Nemotron-based workloads, and under what contracts, remains unverified.

Background

Palantir Technologies, founded in 2003, built its business on data-integration software for intelligence, defence and civilian agencies before expanding into commercial markets. Its platforms — Gotham for government and defence work, Foundry for enterprise data operations, and an AI platform layered across both — are already embedded in US federal workflows, giving the company a distribution channel into precisely the secure environments the NVIDIA post describes. Nemotron is NVIDIA's family of openly released models, which the chipmaker positions as a foundation others can fine-tune and deploy without the licensing and data-residency constraints of closed commercial models.

The pairing addresses a structural problem in government AI adoption: classified and air-gapped networks cannot call out to cloud-hosted commercial models, so agencies handling sensitive data need models that run entirely inside their own accredited infrastructure. Open-weight models are one answer, and both companies have commercial reasons to champion them — NVIDIA sells the hardware such deployments run on, and Palantir sells the software layer that operationalises them. The framing of open models as a pillar of American AI competitiveness also aligns with a broader Washington policy debate over whether open or closed systems better serve national security.

What comes next

Verification would come from Palantir's own product documentation, federal accreditation records or contract announcements naming agencies and vehicles — none of which were part of the material reviewed. Watch for FedRAMP or Defense Department authorisation listings, task orders that reference Nemotron-based capability, and whether Palantir describes the engine in its own investor and product communications.