Turnkey, ready-to-deploy enterprise AI infrastructure built on the Vera Rubin platform and managed with NVIDIA Mission Control.
NVIDIA DGX Vera Rubin NVL72 is the turnkey, NVIDIA-built implementation of the Vera Rubin NVL72 platform: 72 Rubin GPUs and 36 Vera CPUs delivered as a complete, ready-to-deploy rack for training and long-context inference at scale.
It is the building block of NVIDIA DGX SuperPOD and ships with NVIDIA Mission Control, NVIDIA AI Enterprise and DGX OS, backed by three years of enterprise business-standard support for hardware and software.
A complete DGX rack with networking, management and software, designed to remove deployment guesswork.
Automates configuration, cluster operations, workload management and facility integration.
Scales from a single rack to a full DGX SuperPOD with a proven reference architecture.
Vera CPUs and Rubin GPUs let enterprises scale training and long-context inference within existing energy budgets.
Nine L1 NVLink Switches connect all 72 GPUs in one high-bandwidth domain.
Enterprise business-standard support for hardware and software is included.
| GPUs | 72× NVIDIA Rubin |
|---|---|
| CPUs | 36× NVIDIA Vera |
| GPU memory | 20.7 TB HBM4 |
| GPU memory bandwidth | 1,400 TB/s |
| NVFP4 inference | 3,600 PFLOPS |
|---|---|
| NVFP4 training | 2,520 PFLOPS (dense) |
| FP8/FP6 training | 1,260 PFLOPS (dense) |
| Compute fabric | 144+ single-port NVIDIA ConnectX-9 VPI, 800 Gb/s InfiniBand and Ethernet |
|---|---|
| DPUs | 18× dual-port NVIDIA BlueField-4 VPI, 400 Gb/s InfiniBand and Ethernet |
| NVLink | 9× L1 NVIDIA NVLink Switches |
| Management | Host baseboard management controller (RJ45) |
| Software | NVIDIA Mission Control, NVIDIA AI Enterprise, NVIDIA DGX OS |
|---|---|
| Support | Three years enterprise business-standard (hardware and software) |
Next-generation rack-scale platform for agentic AI and reasoning, unifying 72 Rubin GPUs and 36 Vera CPUs in a single NVLink domain.
Blackwell Ultra rack-scale system built for AI reasoning, training and data processing at scale.
Liquid-cooled 72-GPU NVLink domain that acts as a single massive GPU for trillion-parameter LLM inference and training.
Tell us your workload and timeline for the NVIDIA DGX Vera Rubin NVL72. We reply within one business day with a configuration, pricing and lead time.