NVIDIA and Palantir detailed the pilot deployment of AI for supply chains: Nemotron models and the cuOpt tool in Foundry
Palantir is integrating Nemotron 3 Ultra from NVIDIA into the Foundry platform and combining it with the cuOpt tool to optimize supply chains. It is being piloted in the supply chain of NVIDIA itself (1.3 million components per Vera Rubin rack).
Palantir has published details of its collaboration with NVIDIA on AI-powered supply chain management, announced earlier as part of their sovereign AI partnership. Palantir is integrating open Nemotron models, specifically Nemotron 3 Ultra, into the Foundry platform, where companies can fine-tune them on their own operational data. The cuOpt tool from NVIDIA handles scenario planning and optimization—it is intended to identify bottlenecks earlier, speed up material allocation and evaluate alternatives. According to NVIDIA and Palantir, decisions and their outcomes are fed back into the models, allowing the system to improve gradually. Palantir plans to share further details at AIPCon 11.
The pilot deployment targets the supply chain of NVIDIA itself, which includes millions of parts, thousands of suppliers and a global network of partners—according to NVIDIA, a single Vera Rubin rack contains 1.3 million components. Companies in manufacturing, pharmaceuticals and energy are expected to be able to run the entire stack on their own hardware or in the cloud through the Sovereign AI OS reference architecture; infrastructure partners include Dell, Cisco, Rackspace and Nebius. NVIDIA and Palantir originally announced their collaboration to bring sovereign AI to critical supply chains, stating that the first deployment would be within the operations of NVIDIA itself.
According to Artificial Analysis, Nemotron 3 Ultra was the strongest open US model at the time of its June launch; since then, according to the same source, it has been surpassed by Inkling from Thinking Machines Lab. The Nemotron model family is intended to serve as an alternative to Chinese models with open weights. Palantir CEO Alex Karp has long advocated the idea that companies should run their own models instead of handing over knowledge to external providers; Mistral founder Arthur Mensch recently made a similar argument.
Why it matters
The published details reveal a concrete architecture for a sovereign AI approach to supply chain management—a combination of open models that a company can fine-tune and run itself, paired with an optimization tool for scenario planning. For companies in manufacturing, pharmaceuticals and energy, this represents a concrete offering for deploying AI on sensitive operational data without relying on an external model provider, and the test case (the supply chain of NVIDIA itself) will show whether the approach works at a realistic scale of millions of components.
What was added since the original report
Verified updates
-
Palantir is integrating open Nemotron models (specifically Nemotron 3 Ultra) into the Foundry platform for fine-tuning; The Nvidia cuOpt tool handles optimization and scenario planning in supply chains; The pilot deployment targets the supply chain of Nvidia itself, with 1.3 million components in a single Vera Rubin rack; The system includes a feedback loop—decisions and their outcomes are fed back into the models for continuous improvement; The solution is available for manufacturing, pharmaceuticals and energy through the Sovereign AI OS reference architecture
- Palantir is integrating open Nemotron models (specifically Nemotron 3 Ultra) into the Foundry platform for fine-tuning
- The Nvidia cuOpt tool handles optimization and scenario planning in supply chains
- The pilot deployment targets the supply chain of Nvidia itself, with 1.3 million components in a single Vera Rubin rack
- The system includes a feedback loop—decisions and their outcomes are fed back into the models for continuous improvement
- The solution is available for manufacturing, pharmaceuticals and energy through the Sovereign AI OS reference architecture
Relevant practical impact
What this means
For a business
Companies in manufacturing, pharmaceuticals and energy will be able to deploy a combination of Nemotron models and the cuOpt tool in the Foundry platform on their own hardware or in the cloud, which, according to NVIDIA and Palantir, will enable earlier detection of bottlenecks and faster allocation of materials in the supply chain without sharing operational data with third parties.
ProcessesCheck the original
Event sources
independently confirmed · 2 publishers, 1 independent. We count feeds from the same owner only once.