Nvidia and Palantir are building an AI system for supply-chain analysis, beginning with Nvidia’s own network of millions of parts, thousands of suppliers and global manufacturing partners. A single Vera Rubin rack contains 1.3 million components, illustrating the scale of the planning problem.
Palantir is adding Nvidia’s open Nemotron models to its Foundry platform so customers can adapt them using operational data. Nvidia’s cuOpt software will handle optimization and scenario planning. The combined system is designed to identify bottlenecks, allocate materials and compare alternatives, with the results of decisions feeding back into later model updates.
Companies in manufacturing, pharmaceuticals and energy will be able to run the stack on their own infrastructure or in the cloud through a reference architecture called Sovereign AI OS. Dell, Cisco, Rackspace and Nebius are listed as infrastructure partners.
The first test inside Nvidia gives the companies a complex operational environment, but it also means the initial performance claims will come from vendors evaluating their own deployment. Customers will need evidence that recommendations improve real outcomes, remain explainable and do not amplify errors in supplier or inventory data.