NVIDIA
HPE — GB200 NVL72 72-GPU Liquid-Cooled Rack
NVIDIA GB200 NVL72 rack-scale GPU system — 72 Blackwell GPUs and 36 Grace CPUs in a single liquid-cooled NVL72 rack integrated by HPE.
NVIDIA GB200 NVL72 rack-scale GPU platform, delivered as a single factory-integrated NVL72 rack from HPE (Cray). Seventy-two GB200 Grace Blackwell GPUs connect over fifth-generation NVLink into one coherent 72-GPU accelerator — roughly 13.5 TB of HBM3e per rack for the largest AI training and inference workloads, direct-liquid-cooled by an in-rack CDU. Factory-integrated and NVLink-domain-complete. Configuration and availability confirmed at quote.
Specifications
| GPU architecture | NVIDIA GB200 (Grace Blackwell) |
|---|---|
| GPUs per rack | 72× GB200 (18 compute nodes, 4 GPUs each) |
| GPU memory | 192 GB HBM3e per GPU (≈13.5 TB per rack) |
| CPU | 36× NVIDIA Grace |
| NVLink fabric | 5th-gen NVLink · 1.8 TB/s per GPU · 130 TB/s all-to-all in-rack |
| Scale-out networking | ConnectX-7 · 400 Gb/s per GPU · Quantum-2 InfiniBand / Spectrum-X |
| Form factor | Single 19" NVL72 rack, fully integrated |
| Cooling | Direct-liquid-cooled · in-rack CDU |
| Rack power | ≈120 kW per rack (liquid-cooled) |
| Rack weight | ≈1.36 t (≈3,000 lb) per rack |
| Condition | New |
| Warranty | 3-yr HPE Pointnext Tech Care + liquid-cooling preventive maintenance |
Specifications shown are headline figures. Confirm full configuration details when you request a quote.
Last updated
Request a quote for NVIDIA GB200 NVL72 (HPE)
Tell us your requirements and we'll get back to you with availability and pricing.
or reach us directly at