Dell Technologies unveiled the PowerEdge XE8812, the latest server in its AI Factory lineup built with NVIDIA Vera Rubin NVL4 architecture. The fanless, direct‑liquid‑cooled chassis can pack up to 144 GPUs per rack, delivering a dramatic jump in compute density and memory capacity for the most demanding HPC and AI tasks.

The platform expands core counts from 144 to 176 and adds 50% more memory per socket, letting institutions run massive models entirely in‑memory without staging or swapping. Coupled with NVIDIA CUDA‑X libraries, the XE8812 aims to eliminate latency bottlenecks that plague modern AI simulations and scientific workloads.

Dell leverages the open ORv3 rack standard, offering 300 kW power support and 100% direct liquid cooling for CPUs and GPUs. Integrated iDRAC, Rack Controller and Open Manage Enterprise provide real‑time telemetry and automated leak detection, simplifying deployment and reducing risk for large‑scale installations.

Early next year the XE8812 will ship globally, already slated for DOE’s Doudna supercomputer and the Wellcome Sanger Institute’s genome‑decoding pipeline. By consolidating compute, memory and cooling into a single dense rack, Dell gives researchers a turnkey path to run exascale AI and HPC workloads without sprawling infrastructure.