Uvation MarketplaceMarketplace

NVIDIA Vera Rubin NVL72 OEM-Integrated Rack-Scale AI Platform

An OEM-integrated, 100% liquid-cooled rack-scale platform built on NVIDIA MGX NVL72, uniting 72 Rubin GPUs and 36 Vera CPUs.

72
Rubin GPUs
3,168
Olympus CPU cores
75 TB
Total fast memory per rack

Rack-Scale Computing for
Agentic AI

Vera Rubin NVL72 brings 72 Rubin GPUs and 36 Vera CPUs into one MGX NVL72 rack, integrated by OEM partners for training, reasoning and inference.

Built on Third-Generation MGX NVL72

OEM-integrated systems build on the third-generation NVIDIA MGX NVL72 rack design, with 100% liquid cooling across the rack.

3,168 Olympus Cores Across 36 Vera CPUs

Thirty-six Vera CPUs provide 3,168 custom NVIDIA Olympus cores and up to 54 TB of LPDDR5X memory alongside the GPUs.

Confidential Computing in Every Domain

Third-generation confidential computing extends across CPU, GPU and NVLink domains to help protect sensitive AI workloads.

216 TB/s of Sixth-Gen
NVLink Bandwidth

Sixth-generation NVLink connects all 72 Rubin GPUs across the rack,
with 20.7 TB of HBM4 at 1,400 TB/s.

NVIDIA Vera Rubin NVL72 OEM-Integrated Rack-Scale AI Platform rack architecture
72×
Rubin GPUs
216 TB/s
NVLink bandwidth
20.7 TB
HBM4 GPU memory
1,400 TB/s
HBM4 bandwidth

Workloads

Built for large-scale AI training, reasoning and inference, including demanding models and long-context workloads.

Long-Context Reasoning Models

20.7 TB of HBM4 and 75 TB of total fast memory give long-context reasoning models room to run within one rack.

High-Volume NVFP4 Inference

Up to 3,600 PFLOPS of NVFP4 inference across 72 Rubin GPUs serves demanding models at rack scale.

Multi-Rack AI Training

Quantum-X800 InfiniBand or Spectrum-X Ethernet scales training across racks, backed by 32.4 TB/s of scale-out bandwidth.

Engineered to dominate

OEM Flexibility, Rack-Scale Performance

Vera Rubin NVL72 pairs Rubin GPU compute and Vera CPUs with sixth-generation NVLink and full liquid cooling, integrated by OEM partners.

Exploded view of the NVIDIA Vera Rubin NVL72 OEM-Integrated Rack-Scale AI Platform
3,600 PFLOPS
NVFP4 inference
75 TB
Total fast memory
54 TB
LPDDR5X CPU memory
32.4 TB/s
Scale-out bandwidth
Exploded view of the NVIDIA Vera Rubin NVL72 OEM-Integrated Rack-Scale AI Platform
3,600 PFLOPS
NVFP4 inference
75 TB
Total fast memory
54 TB
LPDDR5X CPU memory
32.4 TB/s
Scale-out bandwidth

The rack, itemized

SUBSYSTEM

SPECIFICATION / CONFIGURATION DETAILS

Processor

36 × NVIDIA Vera 88-core CPUs

Operating System

N/A

Chipset

NVIDIA Vera Rubin Superchips (2 Rubin GPUs + 1 Vera CPU each)

Accelerators / GPU

72× NVIDIA Rubin GPUs; sixth-generation NVLink

Memory

75 TB total fast-access memory: 20.7 TB HBM4 and 54 TB LPDDR5X

Primary Storage

N/A

Secondary Storage

N/A

Security

Third-generation confidential computing across CPU, GPU and NVLink domains

Chassis

Third-generation MGX NVL72 rack design; 100% liquid-cooled

Power Supply

N/A

Ports

N/A

Slots / PCIe Architecture

9 × L1 NVIDIA NVLink switches; sixth-generation NVLink

Bezel

N/A

Form Factor

OEM-integrated rack-scale platform

Network Card

ConnectX-9 SuperNICs and BlueField-4 DPUs; final integration is OEM-specific

Warranty

N/A

NVIDIA Vera Rubin NVL72 OEM-Integrated Rack-Scale AI Platform power and networking

Scale-Out Networking
Beyond the Rack

32.4 TB/s scale-out bandwidth

Aggregate bidirectional scale-out networking reaches 32.4 TB/s, linking the rack to larger AI deployments.

ConnectX-9 SuperNICs and BlueField-4 DPUs

ConnectX-9 SuperNICs and BlueField-4 DPUs provide connectivity and infrastructure services;
final integration is OEM-specific.

Quantum-X800 or Spectrum-X fabrics

Scale out across racks with NVIDIA Quantum-X800 InfiniBand

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.