NVIDIA DGX™ systems are purpose-built AI supercomputers designed to accelerate innovation and deliver enterprise-grade performance for the most demanding AI and data science workloads. By integrating the latest NVIDIA GPUs, high-bandwidth interconnects and a complete AI software stack, DGX systems provide organisations with a turnkey infrastructure solution for training large language models, powering generative AI, advancing scientific research and enabling data-driven business transformation. Trusted by leading enterprises, research institutions and governments worldwide, DGX systems set the industry standard for performance, scalability and reliability in AI infrastructure.
At their core, DGX systems combine the latest NVIDIA GPUs with high-bandwidth NVLink™ and NVSwitch™ interconnects, enterprise CPUs, ultra-fast storage and the full NVIDIA AI software stack, optimised and ready to deploy.
From the ground-breaking NVIDIA DGX A100™, which delivered petaflops of AI power in a compact 6U rack unit, to the next-generation NVIDIA DGX H200™, boasting exascale-class performance and networking, each DGX stands as a pillar of AI infrastructure.
But DGX isn’t just hardware. Every system includes NVIDIA Base Command™, NVIDIA AI Enterprise and direct access to a global network of DGXperts, teams of AI specialists who support you every step of the way.
This vertical integration of hardware, software and human know-how makes DGX the backbone of AI innovation across industries.
Whether you're advancing biotech, financial analytics, autonomous systems or creative media workflows, NVIDIA DGX systems enable you to scale your AI with confidence and speed.
Frequently Asked Questions
What exactly are NVIDIA DGX systems?
DGX (Deep GPU Xceleration) systems are NVIDIA’s purpose-built AI supercomputers that combine GPUs (typically 8 per system), NVLink/NVSwitch interconnects, enterprise CPUs, high-speed storage and software—providing a plug-and-play environment for deep learning, AI training and inference.
How has DGX evolved over time?
- DGX-1™ introduced GPU-accelerated AI to the enterprise in April 2016 with up to eight Pascal or Volta GPUs.
- DGX-2™ elevated performance further with up to 16 GPUs and NVSwitch interconnects for massive parallel compute.
- DGX A100™ consolidated all AI workloads—training, inference, analytics—into a 6U system delivering up to 5 petaFLOPS.
- DGX H100™ raised the bar again with 8 Hopper GPUs, 640 GB HBM3, 30 TB NVMe, 2X networking and full AI software stack, ideal for generative AI and LLMs.
- DGX GH200™ and GB200™ – scalable platforms targeting tera-model training and modular supercomputing.
Which use cases thrive on DGX systems?
DGX enables breakthrough innovation across sectors:
- Biotech & Healthcare – accelerated drug discovery, diagnostic imaging, predictive modelling. Institutions like Johns Hopkins and the Tokyo-1 supercomputer rely on DGX H100.
- Media & Creative – real-time ray tracing, cinematic rendering, content generation using generative AI.
- Finance & Telco – fraud detection, risk analytics, telecommunication insights.
- Research & Academia – universities worldwide use DGX for large-scale LLM training and scientific simulations.
What software ecosystem comes with DGX?
Every DGX system includes:
- NVIDIA Base Command™ – manages compute, storage and network infrastructure.
- NVIDIA AI Enterprise™ – streamlines AI development with optimised frameworks and models.
How do organisations scale DGX systems for broader deployment?
- DGX PODs – reference architectures that simplify cluster formation with your chosen storage vendors.
- DGX SuperPODs – turnkey AI factories combining multiple DGX nodes with storage and networking for hyper-scale deployment. The model behind high-performing systems like Selene.
Is DGX accessible for smaller teams or early-stage use?
Yes. NVIDIA offers starter kits and entry pathways to make DGX accessible without steep complexity. DGX Station or upcoming compact systems offer powerful AI capabilities in smaller footprints.