Dell and NVIDIA deliver integrated AI rack-scale infrastructure

Dell Technologies announced integrated rack-scale AI infrastructure co-engineered with NVIDIA.

• Main announcement: Dell unveiled integrated rack-scale systems built on NVIDIA Rubin GPUs and NVIDIA Vera CPUs, including the PowerEdge XE9812 (72-way GPU-accelerated rack-scale server), XE9882L/XE9880L/XE9885L with 8-way HGX Rubin NVL8 GPU acceleration, the factory-integrated IR9000 rack, and PowerSwitch SN6000 Series Ethernet switches (1.6 TbE, up to 409.6 Tb/s switching capacity, up to 2,048 breakout connections). Dell servers also earned NVIDIA Confidential AI certification for encrypted host-memory and secure CPU–GPU transfers.
• Background and details: The announcement emphasizes co-engineered compute and networking (NVLink, Spectrum-6 ASIC, co-packaged optics, liquid cooling), NOS choices including Cumulus Networks and Dell SONiC, integration with NVIDIA Quantum-X800 InfiniBand switching and NVQLink for hybrid quantum-classical workflows; partner Quantum Machines validated quantum control with the R7615. Implementation details include factory-integrated rack systems, liquid-cooled switch options, and validated reference architectures and deployment services.