Breaking

Palantir's Karp Says Government Clients Moved to NVIDIA's Nemotron

July 5, 2026 at 12:31 EDT

Foundation Models
Open Source
Sales & Business

Palantir CEO Alex Karp says some U.S. government customers have switched from proprietary AI models such as those from OpenAI and Anthropic to NVIDIA's open-source Nemotron models. The aim is to keep sensitive work inside Palantir's trusted application layer rather than relying directly on external commercial models.

June 29, 2026 · Palantir

Some U.S. Government Users Swap Proprietary AI for Open-Weight Models

Palantir CEO Alex Karp says certain government customers have switched from OpenAI and Anthropic's closed models to NVIDIA's open-weight Nemotron — keeping sensitive data inside a trusted, air-gapped application layer.

The Stack

Palantir AIP · Foundry · Ontology · Apollo + NVIDIA CUDA-X · NIM · Nemotron

Target

U.S. agencies & critical infrastructure, in air-gapped and classified environments

Proprietary vs. Open-weight

Why sovereignty-focused customers are shifting

Proprietary — OpenAI / Anthropic

Weights: not published
Deployment: mainly via API
Data: transmitted externally
Billing: per-token

Open-weight — NVIDIA Nemotron

Weights: published, customer-controlled
Deployment: local, air-gapped capable
Data: contained in environment
Billing: controlled by customer

Karp's Argument

Per-token billing is a “wealth tax” — enterprises want control of their data, models and alpha.

Deploy: vLLM · SGLang · llama.cpp · Ollama · NVIDIA NIM No proprietary insights leaking into a model's weights

The Case For

IP-leakage worries and rising token costs are pushing a real, practical shift toward open-weight models — and many customers already run them, easing expansion.

The Skeptic's View

Customer counts, switch numbers, benchmarks and pricing stay undisclosed. With little public evidence, some read the claims as heavily marketing-driven.

Continue reading

The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.