NVIDIA DGX Cloud

What it does

NVIDIA describes DGX Cloud as its own cloud environment for building and operating AI at scale. It is used to develop open frontier and foundational models, validate new system architectures and run production AI workloads. Its operational lessons are packaged as open software components under NVIDIA DSX OS. The page presents it as infrastructure rather than an app: customers bring training workloads, and the managed clusters provide the compute.

Through AWS, Google Cloud, Microsoft Azure and Oracle Cloud Infrastructure, NVIDIA offers DGX Cloud as a fully managed AI training platform with co-engineered accelerated computing clusters, flexible term lengths and access to NVIDIA experts. The page also points to models trained on it, including Nemotron, Alpamayo, Isaac GR00T and Cosmos, plus performance benchmarking and the Exemplar Cloud validation programme for partners. What sets it apart is that NVIDIA runs the same environment for its own model development.

It suits ML research teams, enterprises training large models, and cloud partners building AI factories. Access is not self-serve: buyers request a private offer on a cloud marketplace or contact NVIDIA, and a contract with a supported cloud provider appears to be involved. No published rates, hardware configurations or minimum commitments were confirmed, so budget and capacity planning require direct quotes.

What it can do

• Managed multi-node AI training clusters co-engineered with each cloud provider
• Delivery through AWS, Google Cloud, Microsoft Azure and Oracle Cloud Infrastructure
• Access to NVIDIA experts with flexible term lengths
• NVIDIA DSX OS: open, modular software for operating AI factories
• Performance benchmarking and the Exemplar Cloud validation programme

Works with

• Amazon Web Services
• Google Cloud
• Microsoft Azure
• Oracle Cloud Infrastructure

Example workflows

Train a large model on a managed cluster

Inputs: training code, datasets and a target cluster size. Steps: request a private offer on a supported cloud marketplace, agree a term, provision the managed cluster, launch multi-node training with NVIDIA expert support. Output: a trained model checkpoint.

Validate a partner AI cloud

Inputs: a partner's GPU infrastructure and workloads. Steps: apply NVIDIA performance benchmarking, compare results against validated hardware and software stacks, adopt DSX OS components for operations. Output: benchmark results and an operating baseline for the AI cloud.

Before you choose

• Pricing is quote-based through private offers, so costs cannot be compared up front
• It is infrastructure for training at scale, not a ready-to-use chat or app product
• Availability and configurations depend on the chosen cloud partner

Listed under

• Coding

Get started

Use the marketplace private-offer links or the contact form

1. Choose a supported cloud marketplace (AWS, Google Cloud, Azure or OCI)
2. Request a private offer or contact NVIDIA via the form
3. Agree term length and capacity with the provider and NVIDIA

The page links to private-offer requests on AWS, Google Cloud and Oracle marketplaces and a request on Azure. No rates, minimum terms or hardware costs are published. Free tier unconfirmed.

NVIDIA DGX Cloud doesn’t publish prices. You get a quote by contacting sales. See their pricing page ↗

Sources for this overview

Product pages, documentation, and pricing checked 28 Oct 2026. Each source supports part of the overview; availability and prices can change.

• NVIDIA’s AI factory in the cloud. ↗www.nvidia.com
• NVIDIA’s AI factory in the cloud. ↗www.nvidia.com
• NVIDIA DGX Cloud Solution Brief ↗resources.nvidia.com

Company website

nvidia.com: product, pricing and policy pages

Web search

Confirms the official site

Social accounts

Accounts the company links, and its posts on X

Human review

Checked by a person

knowledge_base

Last checked 28 Oct 2026. Every field keeps its source and date; unknown means we couldn’t confirm it.