# NVIDIA DGX

> Source: https://aiwiki.ai/wiki/nvidia_dgx
> Updated: 2026-06-24
> Categories: AI Hardware
> From AI Wiki (https://aiwiki.ai), a free encyclopedia of artificial intelligence. Quote with attribution.

**NVIDIA DGX** is [nvidia](/wiki/nvidia)'s line of integrated artificial-intelligence supercomputers, purpose-built systems that package the company's highest-end data-center GPUs, CPUs, high-speed [nvlink](/wiki/nvlink) and [nvswitch](/wiki/nvswitch) interconnects, storage, and a preconfigured software stack into a single appliance that runs deep-learning training and inference out of the box. Introduced with the DGX-1 in April 2016 and marketed as the "AI supercomputer in a box," the DGX brand has grown from a single 8-GPU server into a portfolio spanning rack-scale machines, scale-out reference architectures (DGX SuperPOD), a cloud service (DGX Cloud), and deskside personal systems (DGX Spark and DGX Station). The original DGX-1 became a landmark of the modern AI era when NVIDIA chief executive Jensen Huang hand-delivered the first production unit to [openai](/wiki/openai) on 15 August 2016. [1][2][3]

## What is NVIDIA DGX?

The defining idea behind DGX is the "AI supercomputer in a box": rather than asking customers to assemble GPUs, networking, and software themselves, NVIDIA sells a turnkey system that is validated and supported as a unit. DGX systems are built around NVIDIA's highest-end GPUs of each generation and use the company's proprietary [nvlink](/wiki/nvlink) interconnect, and later the [nvswitch](/wiki/nvswitch) (NVLink Switch) fabric, to let the GPUs share memory and exchange data at bandwidths far higher than standard PCIe. They ship with NVIDIA's enterprise AI software, including the optimized CUDA libraries and container stack, and are sold to enterprises, research labs, and cloud providers. The 8-GPU NVLink-connected node that the DGX-1 established has since become the standard building block of AI data centers, replicated by NVIDIA's server partners and the major clouds. [1][2]

NVIDIA describes the original system's purpose in stark terms. "The DGX-1 is easy to deploy and was created for one purpose: to unlock the powers of superhuman capabilities and apply them to problems that were once unsolvable," Huang said at the 2016 launch. [3]

## How does DGX differ from HGX?

DGX should be distinguished from NVIDIA's related HGX platform. HGX is a GPU baseboard that NVIDIA sells to server makers, who build their own systems around it; DGX is NVIDIA's own fully assembled, NVIDIA-branded product that uses the same underlying GPU boards. Many DGX generations share their baseboard design with the corresponding HGX module. Because DGX is the system NVIDIA builds first for each new GPU architecture, it doubles as the company's reference architecture and validation platform: the configuration that server partners and cloud providers later replicate at scale. [2]

## What are the DGX generations?

The original DGX-1 was announced in April 2016 at NVIDIA's GPU Technology Conference and built around eight Pascal-generation [nvidia_p100](/wiki/nvidia_p100) GPUs, delivering up to 170 teraflops of half-precision (FP16) performance, which NVIDIA said matched "the throughput of 250 x86 servers" in a single box, for roughly 129,000 US dollars. General availability in the United States began in June 2016. On 15 August 2016, Huang personally hand-delivered the first production unit to [openai](/wiki/openai) at its San Francisco office, with co-founder Elon Musk present; the gift has since become an emblem of the early deep-learning era. In 2017, at GTC, the DGX-1 was refreshed with eight Volta-generation [nvidia_v100](/wiki/nvidia_v100) GPUs, raising performance to roughly 960 FP16 teraflops at a list price near 149,000 US dollars; buyers of the Pascal version were offered a free upgrade to V100 boards. [1][3][4][5]

NVIDIA followed on 27 March 2018 with the DGX-2, which Huang introduced as "the world's largest GPU." It combined sixteen 32 GB V100 GPUs across two baseboards, joined by the first-generation [nvswitch](/wiki/nvswitch) fabric so that all sixteen GPUs could communicate as a single memory pool at 2.4 TB/s of bisection bandwidth. The DGX-2 reached about 2 petaflops of deep-learning performance and listed at 399,000 US dollars. [1][6]

The third generation, the DGX A100, launched on 14 May 2020 with eight Ampere-architecture [nvidia_a100](/wiki/nvidia_a100) GPUs, 320 GB of total GPU memory (later 640 GB with 80 GB cards), and around 5 petaflops of AI performance, at a starting price of about 199,000 US dollars. The fourth generation, the DGX H100, was announced in March 2022 and used eight Hopper-architecture [nvidia_h100](/wiki/nvidia_h100) GPUs with 640 GB of HBM3 memory, delivering 32 petaflops at the new FP8 precision, roughly six times the prior generation. It also introduced ConnectX-7 400 Gb/s InfiniBand networking and a pair of BlueField-3 data-processing units. [1][7][8]

With the Blackwell architecture, NVIDIA shifted its top-end DGX toward rack-scale designs. The air-cooled DGX B200, shown in 2024, pairs eight [nvidia_b200](/wiki/nvidia_b200) GPUs with an x86 host to provide up to 72 petaflops of training and 144 petaflops of inference performance. The flagship became the [gb200](/wiki/gb200) NVL72, a liquid-cooled rack that links 36 Grace CPUs and 72 Blackwell GPUs through a single NVLink Switch domain. It exposes about 13.5 TB of unified GPU memory and up to roughly 1.4 FP4 exaflops, with the NVLink Switch System providing 130 TB/s of GPU-to-GPU bandwidth inside the rack. NVIDIA markets the DGX GB200 as a building block in which each unit is a full NVL72 rack, and customers connect many such racks to scale out. In 2025 the line advanced to the [nvidia_dgx_b300](/wiki/nvidia_dgx_b300) (DGX GB300), based on the Grace Blackwell Ultra (GB300) superchip; the GB300 NVL72 rack again integrates 72 Blackwell Ultra GPUs and 36 Grace CPUs, with up to 1,440 FP4 petaflops of dense Tensor Core compute and about 20 TB of HBM3e GPU memory. [2][9][10][11]

| System | Year | GPUs | AI performance | Notable details |
|---|---|---|---|---|
| DGX-1 (Pascal) | 2016 | 8x P100 | ~170 TF (FP16) | First unit donated to OpenAI; ~129,000 USD |
| DGX-1 (Volta) | 2017 | 8x V100 | ~960 TF (FP16) | V100 refresh; ~149,000 USD |
| DGX-2 | 2018 | 16x V100 | ~2 PF | First NVSwitch; 399,000 USD |
| DGX A100 | 2020 | 8x A100 | ~5 PF | 320 GB GPU memory; ~199,000 USD |
| DGX H100 | 2022 | 8x H100 | 32 PF (FP8) | 640 GB HBM3; ConnectX-7, BlueField-3 |
| DGX B200 | 2024 | 8x B200 | 72 PF train / 144 PF infer | Air-cooled Blackwell node |
| DGX GB200 NVL72 | 2024 to 2025 | 72x B200 + 36 Grace | ~1.4 EF (FP4) | Liquid-cooled rack; 13.5 TB unified memory |
| DGX GB300 NVL72 | 2025 | 72x Blackwell Ultra + 36 Grace | ~1,440 PF (FP4 dense) | Grace Blackwell Ultra superchip; ~20 TB HBM3e |

## What is the DGX SuperPOD?

The DGX SuperPOD is NVIDIA's scale-out reference architecture for connecting many DGX systems into a single supercomputer. Rather than a single product, it is a validated blueprint covering compute nodes, InfiniBand and Ethernet networking, management nodes, storage, power, and cooling, sold as a turnkey solution. The design is modular: nodes are grouped into "scalable units," for example 32 DGX H100 nodes per scalable unit, and a SuperPOD can range from a handful of nodes to several thousand. [12][13]

NVIDIA's internal Eos supercomputer is the reference example of a DGX SuperPOD. Revealed in detail in 2024, Eos is built from 576 DGX H100 systems, totaling 4,608 H100 GPUs interconnected with Quantum-2 400 Gb/s InfiniBand, and delivers about 18.4 exaflops of FP8 AI performance. Measured on the double-precision LINPACK benchmark, Eos recorded an Rmax of roughly 121 petaflops, which placed it among the top ten systems on the TOP500 list at the time. NVIDIA uses Eos to develop and benchmark its own models and software. [12][14]

## What is DGX Cloud?

DGX Cloud is a service that rents DGX infrastructure rather than selling the hardware outright. Jensen Huang announced it at GTC in March 2023, positioning it so enterprises could reach an AI supercomputer "from a browser." Crucially, NVIDIA does not run its own data centers for the service; instead it places DGX-configured capacity inside partner clouds and prices the service itself. Oracle Cloud Infrastructure was the first host, with Microsoft Azure and Google Cloud following. At launch, instances were priced at about 36,999 US dollars per month for an eight-GPU node, a premium relative to comparable on-demand GPU instances from the major clouds. [15][16]

NVIDIA later adjusted its cloud strategy. By 2025 the company had stepped back from positioning DGX Cloud as a direct rival to Amazon Web Services and Microsoft Azure, redirecting much of the capacity toward its own internal research and model development. In May 2025 it introduced DGX Cloud Lepton, a marketplace that aggregates GPU capacity from multiple cloud and infrastructure partners and connects developers to it, with NVIDIA acting as an intermediary rather than an owner-operator. [16][17]

## What are the personal DGX systems?

Alongside the data-center line, NVIDIA has long offered smaller deskside DGX systems. The original DGX Station, introduced in 2017, was a workstation-form-factor machine with four V100 GPUs, and a DGX Station A100 followed with four A100 GPUs. In the Blackwell generation NVIDIA revived and expanded this category. [1]

At CES in January 2025 the company previewed a compact personal AI computer under the codename Project DIGITS. In March 2025, at its spring GTC event, NVIDIA renamed it [dgx_spark](/wiki/dgx_spark) and detailed the design. DGX Spark is built around the GB10 Grace Blackwell superchip, co-developed with [mediatek](/wiki/mediatek), pairing a 20-core Arm CPU with a Blackwell GPU and 128 GB of coherent CPU-GPU LPDDR5X memory, and includes ConnectX-7 200 Gb/s networking. NVIDIA rates it at up to about 1 petaflop of AI performance at low precision and positions it as a desktop machine for prototyping and running models locally. DGX Spark began shipping in October 2025 with a starting price of 3,999 US dollars, sold both directly and through partners. [18][19]

NVIDIA also announced a more powerful deskside system, the DGX Station, built on the GB300 Grace Blackwell Ultra superchip. It combines a 72-core Grace CPU with a Blackwell Ultra GPU over a 900 GB/s NVLink-C2C link, offers 784 GB of coherent memory (252 GB of HBM3e plus 496 GB of LPDDR5X), and is rated at up to roughly 20 petaflops of AI performance, with ConnectX-8 networking for scaling. NVIDIA presented it as a deskside machine capable of working with models up to about a trillion parameters, available to order through partners including Asus, Dell, Gigabyte, HP, MSI, and Supermicro and shipping during 2025. [20][21]

## Why does DGX matter?

DGX systems have played an outsized role in the modern AI boom relative to their unit volume. The hand-delivered DGX-1 helped seed early research at OpenAI; NVIDIA later called the 2016 moment one that "ignited the modern AI revolution." The line established the template, integrated GPU, interconnect, and software, that NVIDIA, its server partners, and cloud providers replicated at enormous scale, with the 8-GPU NVLink node becoming the default unit of AI compute. The SuperPOD reference architecture turned that template into a repeatable way to stand up frontier-scale training clusters, and DGX hardware doubles as the showcase and validation platform for each new NVIDIA GPU architecture. By extending the brand downward to DGX Spark and the Blackwell-era DGX Station, NVIDIA has sought to put the same software environment used in its largest data-center systems onto individual developers' desks. [1][2][18]

## References

1. Nvidia DGX. Wikipedia. https://en.wikipedia.org/wiki/Nvidia_DGX
2. Decoding Nvidia's Blackwell Products: how do B200s, GB200s, HGX/DGX systems, and NVL supercomputers differ? Modal. https://modal.com/blog/nvidia-blackwell
3. NVIDIA Launches World's First Deep Learning Supercomputer. NVIDIA Newsroom, 2016. https://nvidianews.nvidia.com/news/nvidia-launches-world-s-first-deep-learning-supercomputer
4. NVIDIA Delivers DGX-1 'Supercomputer in a Box' to OpenAI. TOP500, 2016. https://www.top500.org/news/nvidia-delivers-dgx-1-supercomputer-in-a-box-to-openai/
5. NVIDIA Advances AI Computing Revolution with New Volta-Based DGX Systems. NVIDIA Newsroom, 2017. https://nvidianews.nvidia.com/news/nvidia-advances-ai-computing-revolution-with-new-volta-based-dgx-systems
6. NVIDIA Announces the DGX-2 System. TechPowerUp, 2018. https://www.techpowerup.com/242761/nvidia-announces-the-dgx-2-system-16x-tesla-v100-gpus-30-tb-nvme-memory-for-usd-400k
7. NVIDIA Ships World's Most Advanced AI System, NVIDIA DGX A100. NVIDIA Newsroom, 2020. https://nvidianews.nvidia.com/news/nvidia-ships-worlds-most-advanced-ai-system-nvidia-dgx-a100-to-fight-covid-19-third-generation-dgx-packs-record-5-petaflops-of-ai-performance
8. NVIDIA Announces DGX H100 Systems. NVIDIA Newsroom, 2022. https://nvidianews.nvidia.com/news/nvidia-announces-dgx-h100-systems-worlds-most-advanced-enterprise-ai-infrastructure
9. GB200 NVL72. NVIDIA. https://www.nvidia.com/en-us/data-center/gb200-nvl72/
10. GB300 NVL72. NVIDIA. https://www.nvidia.com/en-us/data-center/gb300-nvl72/
11. DGX GB300: AI Factory Infrastructure for Enterprises. NVIDIA. https://www.nvidia.com/en-us/data-center/dgx-gb300/
12. NVIDIA DGX SuperPOD, A Turnkey AI Supercomputer. NVIDIA. https://www.nvidia.com/en-us/data-center/dgx-superpod/
13. DGX SuperPOD Architecture (Reference Architecture, DGX H100). NVIDIA Docs. https://docs.nvidia.com/dgx-superpod/reference-architecture-scalable-infrastructure-h100/latest/dgx-superpod-architecture.html
14. NVIDIA Reveals Eos Supercomputer: 4,600 H100 GPUs for 18 AI Exaflops. Inside HPC, 2024. https://insidehpc.com/2024/02/nvidia-reveals-eos-supercomputer-4600-h100-gpus-for-18-ai-exaflops/
15. NVIDIA Launches DGX Cloud, Giving Every Enterprise Instant Access to AI Supercomputer From a Browser. NVIDIA Newsroom, 2023. https://nvidianews.nvidia.com/news/nvidia-launches-dgx-cloud-giving-every-enterprise-instant-access-to-ai-supercomputer-from-a-browser
16. Nvidia steps back from DGX Cloud, stops trying to compete with AWS and Azure. Tom's Hardware, 2025. https://www.tomshardware.com/tech-industry/nvidia-steps-back-from-dgx-cloud
17. DGX Cloud Lepton. NVIDIA. https://www.nvidia.com/en-us/data-center/dgx-cloud/
18. NVIDIA DGX Spark Arrives for World's AI Developers. NVIDIA Newsroom, 2025. https://nvidianews.nvidia.com/news/nvidia-dgx-spark-arrives-for-worlds-ai-developers
19. Personal AI Supercomputer Powered by Blackwell, NVIDIA DGX Spark. NVIDIA. https://www.nvidia.com/en-us/products/workstations/dgx-spark/
20. NVIDIA DGX Station: The Ultimate Desktop AI Supercomputer. NVIDIA. https://www.nvidia.com/en-us/products/workstations/dgx-station/
21. Nvidia unveils DGX Station workstation PCs with GB300 Blackwell Ultra inside. Tom's Hardware, 2025. https://www.tomshardware.com/tech-industry/artificial-intelligence/nvidia-unveils-dgx-station-workstation-pcs-gb300-blackwell-ultra-inside