DGX Cloud gets reworked, will mostly be for internal research
Nvidia has pared back its ambitions for DGX Cloud, once an effort to compete with its biggest customers.
The Information reports that the platform, which is a cloud service built on top of other cloud services, has struggled to find users, all while upsetting Nvidia’s hyperscale bankrollers.
The service will instead be primarily used by Nvidia’s own researchers.
Nvidia has committed to spend at least $13 billion to rent back its own GPUs from hyperscalers such as AWS and Google, and then sought to rent them out to companies like ServiceNow. In the past, it said that this cloud business could grow to $150bn in revenue.
However, Nvidia has now stopped offering the platform to new customers after limited success, thanks to higher prices than directly going to cloud providers. It has also stopped stating that its cloud spending commitments are partly for DGX Cloud in earnings reports.
“DGX Cloud is a great success as a cloud for Nvidia’s own AI research and development and a sandbox in which we work with cloud partners to optimize their compute stacks and developers on their CUDA AI stacks,” DGX Cloud head Alexis Black Bjorlin said to The Information, disputing the report.
“DGX Cloud is fully utilized and oversubscribed, and we are expanding its scale.”
When Bjorlin spoke to DCD last year, she said: “A lot of our customers want to spend their software and their AI/ML expertise developing applications, not necessarily managing the underlying infrastructure – which you may think is a cloud service provider element – but it’s actually the AI infra stack. I think that’s where DGX Cloud offers something that’s uniquely differentiated, we’re meeting the customers wherever they are in their journey.”
This May, the company launched Nvidia DGX Cloud Lepton, a compute marketplace that connects GPUs from providers including CoreWeave, Crusoe, Firmus, Foxconn, GMI Cloud, Lambda, Nebius, Nscale, SoftBank Corp., and Yotta Data Services.
It has also seen pushback from those companies, who see it as a way for Nvidia to get direct relationships with their customers. After a slow start, it is unclear how successful Lepton will be.
https://www.datacenterdynamics.com/en/news/nvidia-pulls-back-on-its-own-cloud-platform-report/

