The partnership delivers a turnkey AI datacenter stack, from hardware procurement to application deployment, targeting customers with strict data sovereignty requirements
Palantir Technologies and Nvidia have announced a joint reference architecture designed to give enterprises and governments a fully integrated, on-premise AI operating system, combining Nvidia’s latest accelerated computing hardware with Palantir’s software suite in a single deployable stack.
The product, called the Palantir AI OS Reference Architecture (AIOS-RA) is built on Nvidia’s Enterprise Reference Architectures and tested to run Palantir’s full software lineup, including AIP, Foundry, Apollo, Rubix and AIP Hub. The architecture is purpose-built to run on Nvidia Blackwell Ultra systems, equipped with eight Blackwell Ultra GPUs and Spectrum-X Ethernet networking for AI training and inference workloads.
The partnership targets a specific and growing segment of the enterprise AI market, namely customers who cannot route sensitive data through public cloud infrastructure. The sovereign AI architecture is particularly critical for customers with existing GPU infrastructure, latency-sensitive workflows, data sovereignty requirements, and high geographic distribution, according to the announcement.
The structure gives deploying organizations total control over their data, AI models, and applications, a requirement that has become increasingly prominent among defense agencies, financial institutions, and foreign governments wary of cloud-based AI services operated by U.S. hyperscalers.
Akshay Krishnaswamy, Palantir’s chief architect, framed the offering as a natural extension of the company’s work in sensitive government environments. He noted that every deployment Palantir has managed has had to operate under strict constraints, and that the Nvidia partnership allows customers to build on existing hardware investments while accessing a fully integrated, production-ready system.
Justin Boitano, Nvidia’s vice president of enterprise AI platforms, described the collaboration as a response to the growing complexity of AI infrastructure, arguing that latency-sensitive and data-sovereign deployments require a full-stack approach built from silicon to software.
The announcement came during Palantir’s AIPCon 9 event in Miami, where the company also unveiled a separate multi-year partnership expansion with GE Aerospace focused on military aircraft readiness, and a new collaboration with Ondas and World View to develop a multi-domain intelligence platform. The clustering of announcements signals an accelerating push by Palantir to deepen its footprint across defense, aerospace, and critical infrastructure at a moment when AI spending in those sectors is growing rapidly.