NVIDIA DGX Station · GB300 Grace Blackwell Ultra

A data-centre AI supercomputer that stands on the floor.

Up to 20 petaFLOPS of FP4 compute and 748 GB of coherent memory in a deskside tower — enough to run trillion-parameter models and always-on AI agents inside your own building, on Windows.

NVIDIA DGX Station·GIGABYTE· ASUS·MSI
252 GBHBM3e on the GPU
496 GBLPDDR5X system memory
900 GB/sNVLink-C2C interconnect
1,600 W20 A circuit required

Why DGX Station

Data-centre performance without leaving the Windows estate

Most AI programmes stall because the compute lives somewhere the team cannot touch. DGX Station puts a GB300-class system on the floor beside the team and keeps it inside the infrastructure IT already manages.

Runs Windows natively

No new group policies, no OS migration. The station deploys inside existing Windows infrastructure, with the NVIDIA AI software stack and WSL support for Linux toolchains.

One coherent memory pool

252 GB of HBM3e and 496 GB of LPDDR5X are unified over NVLink-C2C at 900 GB/s, so a trillion-parameter model fits in a single pool instead of being sharded across rented nodes.

Scale out, not sideways

A ConnectX-8 SuperNIC at up to 800 Gb/s links multiple stations together — or connects them to a GB300 cluster in the cloud or data centre, same software stack.

Three OEM builds, one reference design

NVIDIA reference, or an OEM build

The GB300 Grace Blackwell Ultra platform is available as NVIDIA's own DGX Station and through authorised OEM partners. The silicon is identical; the chassis, service terms and fleet compatibility are not.

Reference build

NVIDIA DGX Station

The reference system from NVIDIA — GB300 Grace Blackwell Ultra, 748 GB coherent memory, NVIDIA AI software stack and NVIDIA support. Available alongside the OEM builds below.

Enquire
GIGABYTE W775-V10-L01 DGX Station
GIGABYTE

W775-V10-L01

  • GB300 Grace Blackwell Ultra, 20 PFLOPS FP4
  • 748 GB coherent memory (252 GB HBM3e + 496 GB LPDDR5X)
  • Closed-loop liquid cooling, single 1600 W PSU
  • 2 x 400 Gb/s QSFP via ConnectX-8, 10 GbE LAN
View product
ASUS ExpertCenter Pro ET900N G3 DGX Station
ASUS

ExpertCenter Pro ET900N G3

  • Built on NVIDIA DGX Station architecture
  • GB300 Grace Blackwell Ultra, 20 PFLOPS FP4, 748 GB
  • Ubuntu certified, NVIDIA AI software stack
  • Enterprise workstation chassis and ASUS service
View product
MSI XpertStation WS300 DGX Station
MSI

XpertStation WS300

  • Built on NVIDIA DGX Station architecture
  • GB300 Grace Blackwell Ultra, 748 GB coherent memory
  • Dual 400 GbE QSFP via ConnectX-8 SuperNIC
  • Linux and Windows support, 1600 W Titanium PSU
View product
HP ZGX Fury AI Station
HP

ZGX Fury AI Station

  • GB300 Grace Blackwell Ultra, 20 PFLOPS FP4
  • 748 GB coherent memory
  • ConnectX-8 SuperNIC
  • HP Z workstation build and service
View product
Dell Pro Max with GB300
Dell

Pro Max with GB300

  • GB300 Grace Blackwell Ultra, 20 PFLOPS FP4
  • 748 GB coherent memory
  • ConnectX-8 SuperNIC
  • Dell Pro Max enterprise service
View product

Technical specifications

GB300 Grace Blackwell Ultra platform

Platform specifications as published by NVIDIA for DGX Station. Final configuration depends on the OEM system you order.

NVIDIA GPU1x NVIDIA Blackwell Ultra
NVIDIA CPU1x Grace 72-core Neoverse V2
GPU memory252 GB HBM3e ECC · 7.1 TB/s
CPU memory496 GB LPDDR5X ECC · 396 GB/s
NVLink-C2C900 GB/s
NetworkingNVIDIA ConnectX-8 SuperNIC · up to 800 Gb/s Ethernet
Supported OSMicrosoft Windows (DGX Station for Windows) · DGX OS
FP4 Tensor Core20 PFLOPS (15 PFLOPS without sparsity)
FP8 / FP6 Tensor Core10 PFLOPS
INT8 Tensor Core330 TOPS
FP16 / BF16 Tensor Core5 PFLOPS
TF32 Tensor Core2.5 PFLOPS
FP3280 TFLOPS
FP64 / FP64 Tensor Core1.3 TFLOPS
Total system power1,600 W · 20 A circuit required
Storage4x M.2 Gen 5 slots
Ethernet ports2x QSFP112 (400 Gb/s each) · 1x RJ45 10GbE · 1x RJ45 1GbE (BMC)
PCIe slots1x PCIe Gen 5 x16 · 2x PCIe Gen 5 x16 (x8 electrical)
Optional RTX PRO GPURTX PRO 6000 Blackwell Workstation / Max-Q · RTX PRO 4000 Blackwell SFF · RTX PRO 2000 Blackwell

Peak rates are based on GPU boost clock. Tensor Core figures are with sparsity unless noted. RTX PRO support varies by OEM system — confirm the exact SKU before ordering.

What it runs

Frontier agents, not benchmarks

AI agents on Windows

Run always-on agents connected to enterprise apps, with token generation happening locally instead of on a metered API.

Large-model inference

Serve models up to a trillion parameters inside your own network — prompts and data never leave the building.

Fine-tuning

748 GB of coherent memory means full fine-tunes that would otherwise need a multi-GPU cluster.

Data science

Ingest and process large data lakes in memory, end to end, without paging to disk.

Physical AI

Pair GB300 with an RTX PRO 6000 for ray-traced simulation alongside AI training and inference.

Linux toolchains via WSL

Run Linux AI frameworks on Windows through WSL, without a second machine or a second image.

Smaller workload? Start with NVIDIA DGX Spark.

The DGX Spark is the GB10 deskside system — 1 PFLOPS FP4, 128 GB unified memory and models up to 200B parameters, at roughly 240 W. We stock the NVIDIA DGX Spark in 1TB and 4TB, plus GIGABYTE, ASUS and MSI builds of the same platform. It is the right first box for pilots and edge deployments; DGX Station is the step up when a single model needs the full 748 GB pool.

See DGX Spark

Homepage placement (preview)

How it would sit on the front page

A flagship strip directly under the hero, above the category cards — one platform, four numbers, one action.

Sizing a DGX Station for your workload?

Tell us the model size, the concurrency and the room it will sit in — including whether a 20 A circuit is available. We will confirm the OEM system, memory configuration and power requirement before you commit.

Request a configuration
💬✉☏