While some may confuse this as NVIDIA competing with cloud service providers (CSPs), the DGX Cloud AI is offered in close partnership with CSPs, who love getting revenue without the hassles. And enterprises love access to state-of-the-art hardware and software in their favorite cloud.
What is “DGX Cloud” and why does NVIDIA offer it?
NVIDIA DGX Cloud is a cloud-based AI supercomputing service that gives enterprises access to the infrastructure and software needed to train and deploy advanced AI or generative AI models. DGX Cloud provides dedicated clusters of NVIDIA DGX Cloud AI supercomputing systems paired with the NVIDIA AI Enterprise software suite. Accessible via a simple web browser, the service gives any enterprise entry into AI supercomputing – using the best NVIDIA hardware and software. Like any cloud, the goal is to remove the complexity of acquiring, deploying, and managing on-premises infrastructure.
NVIDIA developed DGX Cloud specifically because customers wanted the on-premise-like power of DGX systems with quality, performance, and software integrations on a cloud service provider (CSP) of choice. (Yeah, its gold plated.) With DGX Cloud, NVIDIA just made it easy.
DGX Cloud uses, surprise, DGX Servers. These platforms are NVIDIA’s very best technology and are indeed plated in gold. NVIDIA
Most, if not all, cloud service providers offer NVIDIA GPUs as a bare-metal service. The distinction here is that most use “HGX” servers, which fit into the clouds’ infrastructure and can be customized by the customers’ Original Design Manufacturer (ODM). NVIDIA realized that to streamline access to their very best technology at scale, they could install it in the cloud with integrated NVIDIA AI Enterprise software and sell access to enterprises for virtually any cluster size. As enterprises create or customize large language models (LLMs) DGX Cloud provides access to pre-configured clusters and integrated software.
The NVIDIA DGX H100 Server NVIDIA
DGX Cloud AI supercomputing services are offered and supported by NVIDIA and hosted on Google Cloud, Microsoft Azure and Oracle Cloud. To be clear, the servers reside in the CSP’s data centers, not in any NVIDIA data center. This means that customers can choose the cloud provider that best meets their needs, and they can be confident that their DGX Cloud servers will be located in a secure and reliable environment.
Conclusions
NVIDIA has always managed to provide what CSPs and their customers want, with whatever networking the CSP already has, while pushing the envelope to increase performance and speed customers’ time to market. That’s what DGX Cloud is all about. I’m pretty sure that if AWS wanted to host DGX Cloud servers, NVIDIA would be more than happy to sell it. But don’t misinterpret AWS not offering DGX Cloud (yet?) as the two companies partnership is rock solid. In fact, NVIDIA and AWS have made significant AI partnership announcements over the past several months.
The key takeaway here – Cloud service providers love NVIDIA’s AI platform and love DGX Cloud even more because it offloads all the software integration, configuration and testing work to NVIDIA.