NVIDIA and Palantir announce a partnership
Partnership Provisional 83% confidence first seen
Nvidia and Palantir announced they are applying their “sovereign AI” partnership inside Nvidia by fine-tuning a supply-chain AI model for Nvidia’s own operations. The deployment uses Palantir’s Foundry/AIP and Ontology and Nvidia’s cuOpt software to fine-tune the 30B-parameter “Nemotron 3.5 Lightning,” which Nvidia says achieved 86.7% accuracy on a supply-allocation task versus 55.5% for a 550B Nemotron model (~18x larger). The companies said the result serves as a proving ground for expanding this sovereign, customer-deployable approach to other industries.
Decision brief
- What changed
- Nvidia and Palantir said they have deployed their existing “sovereign AI” partnership inside Nvidia by fine-tuning Nvidia’s 30B-parameter Nemotron 3.5 Lightning for Nvidia’s own supply-chain operations using Palantir Foundry/AIP/Ontology and Nvidia cuOpt. Nvidia said the fine-tuned model reached 86.7% accuracy on a supply-allocation task, compared with 55.5% for a 550B-parameter Nemotron model.
- Why it matters
- This matters because Nvidia is positioning a smaller, task-specialized model as materially more effective than a much larger general model for a real operational workflow, which supports investment cases for domain-specific AI over sheer model scale. For leaders, the announcement is also a reference deployment: Palantir and Nvidia are using the approach in Nvidia’s own operations and framing it as a customer-deployable pattern for regulated or data-sensitive environments that want localized control.
- Evidence
- The claim comes from Nvidia and Palantir’s announcement as reported by The New Stack. The coverage consistently states the deployment is being used for Nvidia’s supply chain and cites Nvidia’s reported benchmark of 86.7% versus 55.5%, but it relies on a single article and company-provided results rather than independent validation.
- What remains uncertain
- It is unclear how the reported accuracy was measured, whether the benchmark reflects production KPIs such as fill rate, inventory turns, or planner productivity, and how well the result generalizes beyond the specific supply-allocation task. The coverage does not verify deployment scope, implementation cost, integration effort, or whether other customers have reproduced similar gains with the same sovereign AI architecture.
- Monitor next
- Watch for an independent customer case study or operational metric disclosure showing production impact from this Nvidia-Palantir supply-chain model beyond benchmark accuracy.
Analytical support, not advice — assumptions and open questions stated above.