NVIDIA DGX Station

The ultimate deskside AI supercomputer, powered by NVIDIA Grace Blackwell.

Order Now

Contact a partner to order your system.

Order Now

Overview
Features
Workloads
Specs
Partners

Overview
Features
Workloads
Specs
Partners

Overview

The Ultimate Deskside AI Supercomputer

NVIDIA DGX Station™ is the ultimate deskside AI supercomputer for building and running AI. Powered by the NVIDIA GB300 Grace Blackwell Ultra Desktop Superchip, it features a massive 748 GB of coherent memory and up to 20 petaFLOPS of AI compute performance, enabling seamless development and execution of massive AI workloads, supporting models up to 1 trillion parameters. Preconfigured with an NVIDIA AI software stack, developers, researchers, and data scientists can rapidly develop, fine-tune, and inference with large AI models and long-running AI agents locally and seamlessly deploy to the data center or cloud.

Bringing the Spark to Europe: Local Intelligence Powers Next Wave of Agents

At GTC Berlin, NVIDIA is highlighting secure, local systems across RTX Spark, DGX Spark and DGX Station systems to accelerate coding, engineering, robotics and AI factory development

Read the Blog

All Stations Go: Developers Around the World Power Up DGX Station

Learn about how researchers and builders are using the systems to push the frontier of AI.

Read Blog

Features

Technological Breakthroughs

NVIDIA GB300 Grace Blackwell Ultra Desktop Superchip

NVIDIA DGX Station features the NVIDIA Blackwell Ultra GPU connected to a high-performance NVIDIA Grace CPU using the NVIDIA NVLink®-C2C interconnect to deliver best-in-class system communication and performance.

NVIDIA Blackwell Generation Tensor Cores

NVIDIA DGX Station features NVIDIA Blackwell Generation Tensor Cores, enabling 4-bit floating point (NVFP4) AI. FP4 increases the performance and size of next-generation models that memory can support while maintaining high accuracy.

NVIDIA ConnectX-8 SuperNIC

The NVIDIA ConnectX®-8 SuperNIC™ supercharges hyperscale AI computing workloads. With up to 800 gigabits per second (Gb/s), ConnectX-8 SuperNIC delivers extremely fast, efficient network connectivity, and supports linking up to two DGX Stations to scale model capacity and performance even further.

Preconfigured AI Software Stack

DGX Station features an optimized, preconfigured operating system, Ubuntu with NVIDIA AI Developer Tools. This environment comes pre-installed with system-specific configurations and essential CUDA-X™ libraries, providing a seamless, full-stack foundation for accelerating AI development, data science, and local inference workloads.

NVIDIA NVLink-C2C Interconnect

NVIDIA NVLink-C2C extends the industry-leading NVLink technology to a chip-to-chip interconnect between the GPU and CPU, enabling high-bandwidth coherent data transfers between processors and accelerators.

Large Coherent Memory for AI

AI models continue to grow in scale and complexity. Running these models within NVIDIA DGX Station’s large coherent memory allows for massive-scale models and agents to be developed and run efficiently within one memory pool, thanks to the C2C superchip interconnect that bypasses the bottlenecks of traditional CPU and GPU systems.

IT Manageability

DGX Station is built for enterprise deployment. Out-of-band telemetry via the BMC, Redfish API support, and NVIDIA Data Center GPU Manager give IT teams visibility, monitoring, and remote management capabilities. Hardware root of trust and enterprise secure boot help keep every system in your fleet protected and compliant throughout its lifecycle.

NVIDIA RTX PRO GPU

NVIDIA DGX Station can be configured with up to one additional NVIDIA RTX PRO™ Blackwell Generation GPU in addition to the GB300 Superchip. This combination gives developers the ability to pair data center-grade AI compute capability with powerful, ray-traced visualization and simulation.

NVIDIA AI Enterprise

Enterprises can fast-track AI deployment on NVIDIA DGX Station with an NVIDIA AI Enterprise software license, which provides a suite of software tools, libraries, and frameworks with enterprise support, simplifying the development, deployment, and scaling of production AI applications in the enterprise.

*NVIDIA NIM support planned in June, and production branch support planned in October.

The memory alone lets us train multibillion-parameter LLMs we simply couldn’t before. On-demand, deskside compute access to the latest data-center-class AI machines means we’re not waiting around to try out our ideas. It’s made a real difference in how fast we can iterate.

—Kilian Weinberger, Professor of Computer Science at Cornell University

DGX Station’s fast compute and massive shared memory enable homology search at a new level of scale. NVIDIA GB300’s architecture allows us to load our entire suite of protein structure databases—including the AFDB50 with over 60 million structures—directly into memory for near-instant searches.

— Milot Mirdita, Assistant Professor at Sungkyunkwan University School of Medicine

Using the DGX Station enabled our R&D team to evaluate combining AI weather prediction models with AI downscaling methods to produce high-resolution weather forecasts for electric power grid operations—supporting our research while meeting the needs of our electric utility and ISO/RTO collaborators.

— David Larson, Senior Team Lead, Transmission Operations & Planning at EPRI

Compared with our previous environment of four NVIDIA A100 80GB GPUs, we were able to train models up to three times larger while eliminating parallelization overhead. We ran between two and eight training jobs concurrently and evaluated 40 configurations in total—significantly improving our experiment cycle time.

— Rajendra Boppana, Researcher at University of Texas at San Antonio

Thanks to its large VRAM, we trained multiple models simultaneously—fine-tuning RF-DETR to detect drones using our own datasets. Our large model achieved an AP0.50:0.95 of 66.1. We then deployed it on the Jetson Orin NX, achieving real-time inference at 30 FPS.

— Wennie Tabib, Systems Scientist, Robotics Institute at Carnegie Mellon University

quote 1
quote 2
quote 3
quote 4
quote 5

Workloads

The Ultimate AI and Data Science Desktop

Agentic AI

NVIDIA NemoClaw is an open source stack that adds privacy and security controls to OpenClaw. With one command, anyone can run always-on, self-evolving agents anywhere. Combining NemoClaw with NVIDIA DGX Station, developers have best-in-class capabilities for developing autonomous, long-running agents locally, reducing additional token costs.

Learn More About NemoClaw

Artistic image depicting NVIDIA OpenShell

AI Development

With an extensive library of NVIDIA CUDA-optimized libraries to accelerate deep learning and machine learning training, combined with NVIDIA DGX Station’s massive memory and superchip throughput, NVIDIA’s accelerated computing platform provides the ultimate AI development platform across industries, from developing predictive maintenance, to medical imaging analysis, to natural language processing applications.

Explore AI Solutions

AI Inference

AI models are rapidly expanding in size, complexity, and diversity—pushing the boundaries of what’s possible. NVIDIA DGX Station accelerates inferencing for running large AI models locally on the system, delivering incredible speed for large language model (LLM) token generation, data analysis, content creation, AI agents, and more.

Explore AI Inference Solutions

Data Science

With NVIDIA AI software, including RAPIDS™ open-source software libraries, GPUs substantially reduce infrastructure costs and provide superior performance for end-to-end data science workflows. The large coherent memory pool provided by the NVIDIA DGX Station superchip allows for massive data lakes to be ingested directly into memory, removing data science bottlenecks and accelerating throughput.

Explore Data Science Solutions

Physical AI

DGX Station features a powerful GB300 superchip and can be combined with an additional NVIDIA RTX PRO 6000 Blackwell Workstation GPU to deliver best-in-class compute, simulation, and visualization on the desktop. This combination is ideal for building and deploying Physical AI—which requires powerful local training, inference, and simulation to train, test, and continually optimize autonomous systems and visual AI agents.

Explore Solutions for Physical AI

Personal Cloud

NVIDIA DGX Station can serve as an individual desktop for one user to run advanced AI models using local data, or as a centralized compute node for multiple team members to fine-tune and run IP-specific models on-demand. DGX Station supports NVIDIA Multi-Instance GPU (MIG) technology to partition into as many as seven instances for local development with multiple users, each fully isolated with its own high-bandwidth memory, cache, and compute cores – and easily scaled out to larger MIG deployments in the cloud and data center. This gives administrators the ability to support every workload, from the smallest to the largest, with guaranteed quality of service (QoS) and extends the reach of accelerated computing resources to every user.

Learn About MIG Technology

NVIDIA DGX Station for Windows

NVIDIA DGX Station for Windows puts a trillion-parameter AI supercomputer on every enterprise desk to build, run, and connect always-on, frontier AI agents to enterprise applications and workflows.

Learn More

Specifications

Technical Specifications*

NVIDIA GPU

1x NVIDIA Blackwell Ultra

NVIDIA CPU

1x Grace 72-Core Neoverse V2

GPU Memory

252 GB HBM3e ECC | 7.1 TB/s

CPU Memory

496 GB LPDDR5X ECC | 396 GB/s

NVLink-C2C

900 GB/s

Networking | Peak Bandwidth

NVIDIA ConnectX-8 SuperNIC | Up to 800 Gb/s | Ethernet

Supported OS

Ubuntu with NVIDIA AI Developer Tools

MIG

FP4 Tensor Core

20 | 15³ PFLOPS

FP8 / FP6 Tensor Core

10 PFLOPS

Int8 Tensor Core

330 TOPS

FP16/BF16 Tensor Core

5 PFLOPS

TF32 Tensor Core

2.5 PFLOPS

FP32

80 TFLOPS

FP64 / FP64 Tensor Core

1.3 TFLOPS

Total System Power

1,600 W | 20A circuit required

Decoders

7 NVDEC
7 nvJPEG

Supported RTX PRO GPU⁴

NVIDIA RTX PRO 6000 Workstation Edition
RTX PRO 6000 Blackwell Max-Q Workstation Edition
RTX PRO 4000 Blackwell SFF Edition
RTX PRO 2000 Blackwell

Storage

4x M.2 Gen 5 Slots

Ethernet Ports

2x QSFP112 (400 Gbs per port)
1x RJ45 10 GbE
1x RJ45 1 GbE (BMC for system management)

PCIe Slots

1x PCIe Gen 5 x16
2x PCIe Gen 5 x16 (x8 electrical)

Additional Ports and Connectors⁴

Front IO: 2x USB Type C (USB3.1), 2x USB Type A (USB3.1), Audio
Rear IO: 4x USB Type A, USB Micro-B (BMC) mDP5 (BMC), Audio

Peak rates are based on GPU boost clock.
All Tensor Core specifications are with sparsity unless otherwise noted.
Without sparsity.
Support may vary by system. Check with your OEM to confirm specific support and configurations.
Mini DisplayPort is intended only for system management.

View Datasheet

NVIDIA DGX Station Playbooks

Leverage our curated playbooks to get started on various AI projects on DGX Station. Playbooks contain step-by-step recipes to demonstrate the art of the possible.

Browse Playbooks