Dell Technologies is expanding its Dell AI Factory range with NVIDIA to include the PowerEdge XE8812 server, designed for the most demanding HPC and artificial intelligence workloads. The new platform is set to launch globally early next year and demonstrates the direction in which AI infrastructure is moving: from individual accelerated servers to factory-integrated systems at rack-scale.
The most significant change is the use of the NVIDIA Vera Rubin NVL4 architecture. Dell states that the PowerEdge XE8812 will enable configurations to scale up to 144 GPUs per rack within the OCP-based PowerRack 9100. The system is set to utilise direct liquid cooling and support environments with a power consumption exceeding 300 kW per rack. These are specifications that, until recently, were mainly associated with specialised supercomputing centres, but are now increasingly becoming the benchmark for large-scale AI deployments.
The reason is simple. AI models, scientific simulations and industrial workloads are beginning to utilise similar resources: a large number of GPUs, fast memory, high-performance networking and predictable power management. Dell emphasises that the new generation is set to offer greater memory capacity and more cores than previous systems based on the GB200 NVL4. In practice, this means the ability to run larger models and simulations in memory, without the costly delays resulting from data transfer between infrastructure layers.
The company is also seeking to lower the barrier to adoption. PowerRack is to be delivered as an integrated and validated system, which shortens deployment time and reduces the risks typically associated with manually assembling large AI clusters. This is important because the market is increasingly less concerned with whether to invest in AI, and is instead focusing on how quickly one can move from pilot to production.
Dell points out that over 5,000 customers are already using the Dell AI Factory. Examples include the Doudna supercomputer for the US Department of Energy, InstaDeep’s Kyber cluster in France, genomic research at the Wellcome Sanger Institute in the UK, and the MAVERIC system at Monash University in Australia.
The conclusion is clear: Dell and NVIDIA want to capture a larger share of a market in which AI infrastructure is becoming a strategic asset. It is no longer just about faster servers, but about ready-to-use computing environments for science, industry and nations building their own AI capabilities.
