Uvation MarketplaceMarketplace

Supermicro GB300 NVL72 SuperCluster

The Supermicro GB300 NVL72 SuperCluster is a rack-scale AI system combining 72 NVIDIA B300 Tensor Core GPUs, 36 Grace CPUs, and NVLink Switch trays in a liquid-cooled architecture

GPUs
72 NVIDIA B300 Tensor Core
GPUs
CPUs
36 NVIDIA Grace
CPUs
Power Rating
132 kW
Power Rating

Rack-Scale AI Computing with
Grace Blackwell Ultra

Designed for large-scale AI computing, the SuperCluster integrates GPU acceleration, high-bandwidth memory, and NVLink connectivity within a rack-scale system.

NVIDIA Grace CPU Architecture

Integrates 36 NVIDIA 72-core Grace ARM Neoverse V2 CPUs for compute-intensive workloads.

High-Capacity GPU Memory

Provides approximately 21 TB of HBM3e GPU memory and 17 TB of LPDDR5X memory.

Integrated Rack Architecture

Combines 18 liquid-cooled compute trays with 9 NVLink Switch trays.

High-Density GPU Acceleration

GB300 Grace Blackwell Ultra Superchips bring GPU compute and high-capacity memory together in a rack-scale platform

Exploded view of the Supermicro GB300 NVL72 rack with compute trays and NVLink Switch trays
72 NVIDIA B300 Tensor
Core GPUs
288 GB
per GPU
17 TB LPDDR5X
Approximately
21 TB HBM3e
Approximately

Workloads

The SuperCluster's GPU resources, CPU architecture, and rack-scale interconnect support demanding AI computing workloads.

Large-Scale AI Computing

720 petaFLOPS of FP8 compute and approximately 21 TB of HBM3e help teams train trillion-parameter foundation models.

GPU-Accelerated Processing

1,080 petaFLOPS of dense FP4 compute and 72 Blackwell Ultra GPUs unified at 1.8 TB/s NVLink interconnect bandwidth speed up high-throughput parallel processing and real-time inference.

High-Performance Computing

36 NVIDIA Grace CPUs with 2,592 cores and over 17 TB of LPDDR5X memory feed complex simulation, genomics and scientific research.

The rack, itemized.

SUBSYSTEM

SPECIFICATION / CONFIGURATION DETAILS

Processor

36× NVIDIA 72-core Grace ARM Neoverse V2 CPUs

Operating System

N/A

Chipset

GB300 Grace Blackwell Ultra Superchips

Accelerators / GPU

72× B300 Tensor Core GPUs; 288 GB each

Memory

Approximately 21 TB HBM3e plus 17 TB LPDDR5X

Primary Storage

up to 8 × E1.S PCIe 5.0 drives

Secondary Storage

N/A

Security

Dedicated 1G/10G Out-of-Band IPMI Management Network

Chassis

48U Liquid-Cooled Rack Enclosure with 18 × 1U Compute Trays and 9 × NVLink Switch Trays

Power Supply

8 × 1U 33kW Power Shelves with 6 × 5.5kW PSUs per Shelf

Ports

Per compute tray: 4 × NVIDIA NVLink switch ports

Slots / PCIe Architecture

Up to 8 × E1.S PCIe 5.0 Drive Bays per Compute Tray; 144 Total

Bezel

N/A

Form Factor

48U Rack-Scale, 19-inch Rack

Network Card

4 × NVIDIA ConnectX-8 800Gb/s SuperNICs per compute tray; up to 2 × BlueField-3 DPUs

Warranty

N/A

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.