Dell Technologies adds the PowerEdge XE8812 to its AI Factory with NVIDIA. The server, based on the NVIDIA Vera Rubin NVL4 architecture, is designed for intensive HPC and AI workloads and supports up to 144 GPUs per rack.
The growing demand for AI infrastructure is forcing organizations to invest in scalable solutions. Global AI investments will rise by 44 percent year-over-year in 2026. 87 percent of organizations see innovation and 75 percent see AI as a strategic priority.
Powerful specifications for demanding workloads
The Dell PowerEdge XE8812 was developed for AI training, inferencing, and molecular simulations. Thanks to a fanless design, direct liquid cooling, and extra memory, larger models and simulations run entirely in-memory. This minimizes latency caused by data movement.
The server offers a compute density of up to 144 GPUs per rack, more than 300 kW of power, and fully liquid-cooled CPUs and GPUs. With 50 percent more memory per socket and additional GPU memory, larger AI and HPC workloads can run more efficiently.
Open architecture and rapid deployment
The PowerEdge XE8812 is based on the open ORv3 standard and is managed via iDRAC, Dell Integrated Rack Controller, and OpenManage Enterprise. Real-time telemetry and automated leak detection help identify risks early. Dell PowerRack offers pre-integrated and validated rack systems. In combination with Dell ProDeploy white-glove services, the racks can be deployed and ready for live workloads in just over six hours.
More than 5,000 customers worldwide are already using the Dell AI Factory. For example, Dell, NVIDIA, and NERSC in the United States are working on the Doudna supercomputer, while in France, InstaDeep is scaling up the Kyber cluster. The Dell PowerEdge XE8812 will be available worldwide in early 2027.