Palantir Technologies and Nvidia have formed a partnership focused on sovereign AI, SDX Central reported on March 13.

The partnership was announced ahead of Nvidia's annual GTC developer event.

Under the partnership, Palantir will introduce AIOS-RA (AI OS Reference Architecture) based on the Nvidia Enterprise Reference Architecture.

AIOS-RA is a platform aimed at customers that must meet data sovereignty requirements, have existing GPU infrastructure, run latency-sensitive workloads or operate in geographically distributed environments.

The hardware consists of 8 Nvidia Blackwell Ultra GPUs and Spectrum-X Ethernet networking for AI inference and training. The software stack includes Nvidia AI Enterprise, CUDA-X libraries, Nemotron open models and Magnum IO. Palantir will provide the Rubix zero-trust Kubernetes platform, the Apollo autonomous deployment and lifecycle management service, and its enterprise AI platform AIP.

For Nvidia, the partnership is a move to expand its sovereign AI portfolio. Nvidia has already signed similar contracts with Orange Business, India's NextGen and the British government. Separately, Palantir is working with Accenture to support Sovereign AI, a British infrastructure company that provides a sovereign AI foundation to Europe's private companies and government sector.

Justin Boitano (저스틴 보이타노), vice president of enterprise AI platforms at Nvidia, said, "Environments that demand low latency and data sovereignty require a full-stack architecture that spans silicon, systems and software." Akshay Krishnaswamy (악샤이 크리슈나스와미), Palantir's chief architect, said, "From our first deployment for the U.S. government to today, our software has been designed so customers can maintain control in the most complex and sensitive environments."