Dell and Palantir are teaming up on an on-premises AI operating system. Built on the Dell AI Factory, it marks the first time Palantir's core Foundry platform and Ontology data engine have been moved entirely into a customer's local data center.
The new architecture swaps out some key components. In the initial joint reference architecture, Dell's ObjectScale and PowerFlex storage directly replace Ceph, which Palantir previously relied on for persistence. On the hardware side, Dell PowerEdge GPU servers with Nvidia HGX B-series accelerators handle model training and inference, while CPU nodes run Palantir's control plane.
This setup is aimed squarely at customers in defense, public safety, and core banking — industries that can't move sensitive data to the cloud due to compliance restrictions. The system is built on Palantir's Rubix and Apollo zero-trust runtime, using multi-tenant container isolation and continuous audit logs to keep private data and AI agent interactions under tight control, even in air-gapped environments.