Uvation MarketplaceMarketplace

NVIDIA DGX SuperPOD with DGX Vera Rubin NVL72 - Spectrum-X Ethernet

Deploy large-scale AI infrastructure with NVIDIA DGX SuperPOD, combining Vera Rubin NVL72 racks, Spectrum-X Ethernet, 72 Rubin GPUs, 36 Vera CPUs, 75 TB of fast memory per rack, and liquid cooling.

72 x
NVIDIA Rubin GPUs per rack
36 x
NVIDIA Vera CPUs per rack
75 TB
Total fast memory per rack

Rack-Scale Compute for
Large AI Workloads

Combine high-density GPU compute, fast memory, and Ethernet networking in a scalable AI infrastructure platform designed for demanding training and inference workloads.

High-Density AI Compute

Integrates 72 Rubin GPUs and 36 Vera CPUs in each DGX rack to support compute-intensive AI workloads.

High-Capacity Fast Memory

Provides 75 TB of fast memory per rack for memory-intensive model development and execution.

Scalable Ethernet Architecture

Connects multiple DGX Vera Rubin NVL72 racks through a Spectrum-X Ethernet-based architecture for larger AI deployments.

Unified Rack-Scale AI Infrastructure

Bring GPU compute, CPU processing, local NVMe storage, networking, and liquid cooling together in a coordinated infrastructure platform.

NVIDIA DGX SuperPOD with DGX Vera Rubin NVL72 - Spectrum-X Ethernet architecture
72
Rubin GPUs per rack
36
Vera CPUs per rack
75 TB
fast memory per rack
3
racks per scalable unit

Workloads

Support AI development and deployment with rack-scale compute, high-capacity memory, and Ethernet connectivity for distributed workloads.

Large Scale
AI Training

216 Rubin GPUs and 216 TB/s NVLink 6 deliver 2,520 PFLOPS NVFP4 per rack.

Foundation Model
Development

75 TB fast memory (20.7 TB HBM4) powers multi-trillion parameter models.

Real-Time AI Reasoning and
Agentic Inference

3,600 PFLOPS NVFP4 and dual 800 Gb/s Spectrum-6 Ethernet deliver ultra-low latency for autonomous agentic workflows and interactive reasoning at scale.

Engineered to dominate

Built for Demanding AI Compute

Combine high GPU density, substantial fast memory, and scalable rack architecture to support intensive AI workloads across a coordinated infrastructure deployment.

Exploded view of the NVIDIA DGX SuperPOD with DGX Vera Rubin NVL72 - Spectrum-X Ethernet
72
Rubin GPUs per rack
36
Vera CPUs per rack
75 TB
fast memory per rack
3
racks per scalable unit
Exploded view of the NVIDIA DGX SuperPOD with DGX Vera Rubin NVL72 - Spectrum-X Ethernet
72
Rubin GPUs per rack
36
Vera CPUs per rack
75 TB
fast memory per rack
3
racks per scalable unit

The rack, itemized.

SUBSYSTEM

SPECIFICATION / CONFIGURATION DETAILS

Processor

36 × NVIDIA Vera CPUs per rack

Operating System

DGX OS; Ubuntu supported per DGX rack

Chipset

N/A

Accelerators / GPU

72× Rubin GPUs per rack; 3 racks per scalable unit

Memory

75 TB fast memory per DGX rack

Primary Storage

1× 2 TB E1.S NVMe OS drive per compute tray

Secondary Storage

4× 7.68 TB E1.S NVMe local drives per compute tray; certified partner shared storage

Security

Third-generation NVIDIA Confidential Computing (Vera Rubin NVL72 platform)

Chassis

Multi-rack, liquid-cooled deployment

Power Supply

4× power shelves per DGX rack; up to 110 kW per shelf

Ports

ConnectX-9 compute and BlueField-4 storage/management connections

Slots / PCIe Architecture

Per DGX rack: 9 L1 NVLink switches; scale-out fabric per variant

Bezel

N/A

Form Factor

3 DGX Vera Rubin NVL72 racks per scalable unit

Network Card

NVIDIA ConnectX-9 adapters and NVIDIA BlueField-4 DPUs

Warranty

NVIDIA Enterprise Support; Contract-Dependent

NVIDIA DGX SuperPOD with DGX Vera Rubin NVL72 - Spectrum-X Ethernet power and networking

Power Fabric

4× power shelves (24× 18.3 kW PSUs)

Deliver up to 440 kW per rack to feed the 100% direct liquid-cooled NVL72 system.

144× ConnectX-9 at 2× 800 Gb/s

Dual 800 Gb/s Spectrum-6 Ethernet ports per Rubin GPU for inter-rack cluster networking.

18× dual-port BlueField-4 DPUs

2× 400 Gb/s for storage and in-band fabric, alongside dedicated 1GbE OOB management.

Nuclear-Ready, Plug and Play AI Factories

Bypass grid constraints with self-contained, high-density inference ready infrastructure directly adjacent to clean, base load nuclear power.

Global Deployment Engine

Any Infra, Anywhere in the world.

Global Shipping and Deployment Capability treats advanced infrastructure as mobile, modular freight.

Packaged complex computing nodes, cooling manifolds, and power distribution units into certified, standard form factors.

360° Service Ecosystem

Unifying Legacy with Next-Gen Compute.

Vertically integrated service ecosystem that serves as an operational bridge between legacy environments and next-generation, AI-native infrastructure.

Finance and Leasing

Capital-as-a-Service for High-Density Compute

Convert upfront capital expenditures (CapEx) into predictable operating expenses (OpEx).

Uvation allows enterprises and sovereign entities to deploy multi-megawatt AI factories and advanced GPU clusters without liquidity constraint.