(858) 613-1818Quote list 0
Coming soon

8 node HGX Rubin NVL8 AI pod, coming soon

Eight HGX Rubin NVL8 servers with 64 Rubin GPUs on an 800 Gb/s fabric, with control plane and flash storage, designed now on the B300 pod template.

GPU: NVIDIA HGX Rubin NVL8

Illustration of one rack with eight GPU nodes and a top of rack switch
GPU nodes
8 x 8 GPU HGX Rubin NVL8 servers, 64 x Rubin, 288 GB HBM4 each; 2 TB per node as published for HGX Vera Rubin NVL8
Fabric
64 x 800 Gb/s ports, non blocking, Quantum-X800 or Spectrum-X class; adapter generation per NVIDIA when published
CPU and control
x86 or Vera hosts per NVIDIA's published options, 2 x CPU nodes, 6 x control plane nodes
Storage
64 TB usable all flash minimum, 1 TB per GPU, 2 x NVMe storage nodes
Racks and power
Not yet published; plan on at least the HGX B300 pod figures of 3 to 4 racks and about 125 kW
Delivery
Expression of interest now, firm quotation when NVIDIA and system partners release configurations

Overview

The 8 node HGX Rubin NVL8 pod is our standard pod design carried forward to NVIDIA's next eight GPU baseboard. Eight HGX Rubin NVL8 servers with 64 Rubin GPUs sit on a non blocking 800 Gb/s fabric with the CPU, control plane and storage nodes a production cluster needs. The node figures NVIDIA has published so far, 2 TB of HBM4 at 130 TB/s and 400 PFLOPS of sparse NVFP4 per node, are as published for HGX Vera Rubin NVL8; the x86 column is not yet published.

Chassis sizes, per node power, cooling options and reseller delivery dates have not been published, so the pod is listed as coming soon and we are not taking firm orders. What we can do now is register your interest, reserve a position in our allocation queue, design the fabric, storage and control plane, and review your facility against the HGX B300 pod figures, since a Rubin pod will need at least that much power and cooling.

If the project cannot wait, the 8 node HGX B300 AI pod is deliverable today on the same fabric design and runs the same CUDA software, so a cluster can start on Blackwell Ultra and add Rubin nodes as they ship.

Use cases

  • Planning 2026 and 2027 training clusters
  • Roadmap alignment for AI factories
  • Reasoning model inference at scale

Specifications

Compute
GPU nodes8 x 8 GPU HGX Rubin NVL8 AI server; chassis size and cooling not yet published
GPUs64 x NVIDIA Rubin, 288 GB HBM4 each
GPU memory2 TB and 130 TB/s per node as published for HGX Vera Rubin NVL8; x86 HGX Rubin NVL8 figures not yet published
NVFP4 performance400 PFLOPS sparse inference per node as published for HGX Vera Rubin NVL8; x86 variant not yet published
CPU nodes2 x 1U dual socket EPYC 9005 or Intel Xeon 6 compute servers, meeting 6 threads and 24 GB per GPU
Control plane6 x 1U single socket EPYC 9005 control plane servers
Fabric
East west64 x 800 Gb/s, InfiniBand XDR on Quantum-X800 class switches or 800 GbE on Spectrum-X class switches, non blocking
North south16 x 400 Gb/s ports into a 400 GbE leaf pair; BlueField-3 or BlueField-4 class, per NVIDIA when published
Management1 GbE out of band network for every BMC and switch
Storage
Capacity64 TB usable minimum, 1 TB per GPU, expandable
Nodes2 x 2U 24 bay all NVMe storage servers with 2 x 400 Gb/s each
SoftwareParallel file system or NVMe over Fabrics, specified with the order
Facility and availability
RacksNot yet published; plan on 3 to 4 racks as for the HGX B300 pod
Planning powerNot yet published; plan on at least 125 kW as for the HGX B300 pod
CoolingAir cooled and liquid cooled chassis expected from system partners, not yet published
Interim platform8 node HGX B300 AI pod on the same fabric design, deliverable now

Figures describe the system class. Exact configuration, weights and power draw are confirmed on the quotation.

What's included

  • Expression of interest and queue position
  • Fabric, storage and control plane design ahead of delivery
  • Site readiness review against HGX B300 pod requirements
  • Specification updates as NVIDIA publishes them
  • Firm quotation when configurations are released

Compliance

Compare it with the next class up or down

The nearest classes we quote are the 8 GPU HGX Rubin NVL8 AI server and the 8 node HGX B300 AI pod, 64 GPUs. If neither is quite right, tell us the workload and we specify the system between them.

Talk to an engineerRequest a quotation