Dell Technologies is the first OEM to deliver a fully validated, direct liquid-cooled networking rack, and it is already running in production.1
The Dell PowerRack for networking is now live at Nscale as part of an NVIDIA Vera Rubin NVL72 scalable unit, marking a shift in how AI infrastructure gets built and deployed. At rack scale, performance is no longer just a compute question. Networking, power, cooling, validation, and deployment have to work as one engineered scalable unit, and that system now exists.
PowerRack for networking consolidates switches, cabling, power distribution, PowerCooled RCDU and management software into a single factory-integrated architecture, reducing the complexity that has historically slowed data center deployments and eroded uptime.
From networking components to a production-ready rack
AI factories place new demands on the network. As GPU systems scale, the network must provide the bandwidth, latency, resiliency and operational consistency required to keep distributed workloads moving efficiently. The challenge is not simply selecting switches and assembling them in a rack. It is integrating the network into the complete AI infrastructure design.
Dell PowerRack for networking is built around that integrated model. Each Dell IR9148 rack uses eight (8) PowerSwitch SN6600-LD switches along with the shuffle units and scale-out fabric leaf and spine configurations. Built on the NVIDIA Spectrum-6 ASIC and fully liquid-cooled, the SN6600-LD delivers 102.4 Tb/s of switching capacity per ASIC, totaling more than 819 Tb/s of switching capacity per rack. Rather than asking customers to source, configure and validate each element independently, Dell engineers the networking rack as a complete system — with direct liquid cooling, power distribution and rack-level integration designed into the solution from the start. L11 PowerRack for Networking is fully flexible in meeting different L12 datacenter requirements.
The result is a fully validated networking rack that can be connected to water, power and fabric cabling, and immediately deployed as part of a larger, scalable AI unit. This approach helps reduce integration complexity and gives customers a more predictable path from delivery to production.
Direct liquid-cooled networking for the AI factory
As AI infrastructure density increases, traditional approaches to thermal management face growing constraints. Direct liquid cooling helps address the heat generated by high-performance networking equipment while supporting the power and space requirements of modern data centers.
With Dell PowerRack for networking, direct liquid cooling is integrated into the rack design and validated as part of the complete system. That means cooling is not treated as an afterthought or a separate integration exercise. It is part of the engineered architecture, alongside networking, power and deployment readiness.
This rack-level approach can help customers simplify planning, reduce deployment risk and establish a repeatable foundation for scaling AI infrastructure.
Built for NVIDIA Vera Rubin NVL72
The Dell PowerRack for networking is in full production at Nscale for a NVIDIA Vera Rubin NVL72. This production deployment demonstrates how networking can be delivered as an integrated building block within a next-generation AI architecture.
For AI infrastructure operators, the value of a scalable unit is consistency. Each unit can be engineered, validated and deployed according to a repeatable design, helping organizations scale without re-solving the same integration challenges for every rack.
Dell’s role extends beyond supplying networking hardware. We bring together the rack, cooling and power considerations required to deliver a validated system that is ready to become part of the customer’s production environment.
Execution is the differentiator
The transition to AI factories is not only a technology challenge. It is an execution challenge. Customers need infrastructure that can move from design to deployment with fewer handoffs, clearer validation and greater operational predictability.
Dell PowerRack for networking is designed to help address that need:
-
Integrated by design: Networking, direct liquid cooling and power distribution are brought together at the rack level.
-
Fully validated: The system is engineered and tested as an integrated networking rack rather than as disconnected components.
-
Production-proven: Full production at Nscale demonstrates the solution in an NVIDIA Vera Rubin NVL72 scalable-unit deployment.
-
Built to scale: A repeatable rack design supports the expansion of AI infrastructure with greater consistency.
For Nscale, the execution was a game changer.
Making networking ready for what comes next
The AI infrastructure market is entering an era in which the rack — not the individual component — is the unit of deployment. That shift makes system integration, thermal design and operational readiness central to the customer experience.
Dell PowerRack for networking gives customers a fully validated, direct liquid-cooled networking rack designed for that era. As part of a NVIDIA Vera Rubin NVL72 system at Nscale, it shows how Dell is helping make next-generation AI infrastructure more integrated, more repeatable and more ready for production.
For organizations building AI factories, the objective is clear: move from complex infrastructure integration to a scalable system that is engineered to work together from the start. PowerRack is a unified family of factory-integrated rack solutions spanning compute, networking, and storage, giving organizations a consistent, validated foundation across every layer of the AI infrastructure.
Dell reported this
Source: www.dell.com
Source link
