Dell expands enterprise AI capabilities with new infrastructure

Dell Technologies has announced its latest suite of AI infrastructure solutions built upon NVIDIA's Vera Rubin platform. This development aims to introduce PowerEdge servers equipped with the Vera Rubin NVL72 architecture, delivering 3.6 exaflops of processing power.

The PowerSwitch networking solutions will integrate Spectrum-6 technology, offering 102.4 terabits per second switching capacity. This enhancement, coupled with BlueField-4 integration, aims for 800 gigabits per second performance, supporting a range of NVIDIA Nemotron 3 AI models.

Dell enhances its AI server line by embedding NVIDIA's Rubin platform, a move highlighted during Dell Technologies World in 2025. This collaboration aims to advance both agentic AI and AI reasoning capabilities.

Upcoming PowerEdge servers will feature the NVIDIA Vera Rubin NVL72 platform, delivering AI performance in conjunction with 75 terabytes of fast memory. They will also incorporate resiliency capabilities within Dell's rack-scale infrastructure.

Expanding HGX Server Capabilities

The expansion of Dell's AI server portfolio continues with support for NVIDIA HGX Rubin NVL8 configurations. These systems aim for approximately 400 petaflops of AI performance, featuring 2.3 terabytes of HBM4 memory and 176 terabytes per second memory bandwidth.

This is designed to offer deployment flexibility for diverse data centre requirements, supporting high-performance AI infrastructure without requiring major facility changes.

Networking & Data Processing Enhancements

Dell's PowerSwitch portfolio sees growth with the inclusion of NVIDIA Spectrum-6 Ethernet technology, extending support to 512 ports of 800G connectivity with co-packaged optics. This targets a 5x power efficiency improvements and enhances reliability by 10-fold.

Complementing these networking advancements are innovations within the Dell AI Data Platform with NVIDIA BlueField-4 data and storage processor integration, offering capabilities in KV Cache storage tiering and AI-native data processing.

Supporting Intelligent AI Models

The integration of the NVIDIA Nemotron family, spanning from 30B to 500B parameters, marks an advancement in AI intelligence and efficiency. These models, with their hybrid mixture-of-experts architecture, seek to empower organisations to deploy sophisticated reasoning workloads. 

As AI infrastructure demands surge, driven by cloud and enterprise environments, Dell's offerings, in collaboration with NVIDIA, aim to address these challenges by delivering scalable architecture solutions that evolve with organisational needs.

These platforms aim to support enterprise AI infrastructure, providing technology designed to help businesses deploy and scale AI applications effectively.