Hitachi Vantara Expands Hitachi iQ AI Infrastructure With Nvidia HGX Platform

By adding Nvidia’s HGX H100 and H200 GPUs to the Hitachi iQ AI-ready infrastructure, Hitachi Vantara looks to take it to a wider range of enterprises and eventually midrange customers looking for a complete technology stack for AI workloads.

Data storage and infrastructure technology developer Hitachi Vantara Tuesday unveiled a new version of its Hitachi iQ AI solution stack that features Nvidia’s HGX GPUs.

Hitachi iQ, unveiled earlier this year, brings Hitachi Vantara’s AI-ready infrastructure technologies together with Nvidia’s GPU technologies, said Tanya Loughlin, Hitachi Vantara’s director for AI and unstructured solutions product marketing.

The Hitachi iQ brand encompasses AI-ready infrastructure stacks and services the company brings to market, Loughlin told CRN. The first was in July with the launch of Hitachi iQ with Nvidia DGX GPUs and Nvidia BasePod certification she said.

On Tuesday, Hitachi Vantara is officially launching Hitachi iQ with Nvidia’s HGX GPU, Loughlin said.

Currently, with the DGX-version, the Hitachi iQ AI systems are integrated in the field. However, the HGX versions are integrated by Hitachi Vantara before shipping to channel partners and customers for final configuration, she said.

“What’s exciting about this is we are reselling all of it,” she said. “Nvidia’s DGX can only be sold by DGX-certified partners. With HGX, we're packaging it and reselling the whole solution. Our infrastructure, our storage, the Nvidia components, the HGX compute layer, as well as Nvidia's AI enterprise, which is basically a framework to build AI applications on. We’ll be able to package that up, and that goes right out of our distribution center directly to customers.”

The DGX was one of the original high-end GPU servers that Nvidia built, with eight H100 or H200 GPUs, and were originally shipped by Nvidia to the company’s ODM partners, said Gary Hemminger, senior director for AI solutions product management at Hitachi Vantara.

“The HGX H100s and H200 we’re selling are basically Supermicro versions of the DGX,” Hemminger told CRN. “It has the same parts in it. About 90 percent of the value of the system is still the GPUs, but it’s basically the same system. In fact, we ran through the BasePod testing with DGX. Then once we got BasePod certification, we took the DGX out and replaced it with the HGX and validated that they get basically the same performance. Actually, the HGX is slightly higher, but basically the same capacity, same GPUs, same network interfaces, same performance.”

The Hitachi iQ integrated system includes an Nvidia HGX H100 system based on a Supermicro SuperServer; Nvidia software including Nvidia Base Command Manager, Nvidia AI Enterprise and Nvidia GPU Direct Storage; Hitachi Content Software for File with Hitachi Vantara storage nodes; Nvidia networking components; and Hitachi Vantara HA810 G3 servers for management.

The go-to-market for the DGX and HGX versions is different, Hemminger said.

“The go-to-market with the DGX is basically meet-in-the-channel and work with DGX-authorized resellers to differentiate our storage solution and our other solutions so that so they plug those in,” he said. “With the HGX, our go-to-market is we can sell everything.”

Loughlin said this gives partners who aren’t DGX-certified a fully integrated stack solution to resell.

“We are having numerous conversations with DGX-certified partners to then have them sell. … So there’s a lot of different go-to-market avenues that we’re working on with partners,” she said.

Hitachi Vantara expects the Hitachi iQ’s primary workloads to be mainly large language model creation, fine-tuning, and RAG (retrieval-augmented generation), Hemminger said.

“Maybe not so much inferencing,” he said. “You’ll probably see that more in the mid-range to entry products. In terms of industry verticals and use cases, right now we’re focused mostly on where we’re strong, vertical-wise, which is banking and financial, health care, telecommunications and manufacturing.”

Hemminger said that Hitachi Vantara has seen no supply issues related to recent reports that Nvidia is shifting some orders away from Supermicro.

“In terms of supply, that’s not been an issue at this point,” he said. “In fact, the major issue has been a change in U.S. export law. There is the acceptable country list, the embargoed list, and the restricted list. For the restricted list, you basically had to get special permission. ‘Restricted.’ The approved list was about 50 countries worldwide. Now that’s shrunk to 25. So more countries got put on the restricted list, and so the burden is on us, Nvidia, and Supermicro now because we have to get exceptions every time for the countries on those restricted list. There’s really no change in supply, just that what countries we can sell to without having to deal with anything is.”

The expansion of Hitachi iQ is really an evolution of what Hitachi has always done, said Dave Cerniglia, president of Consiliant Technologies, an Irvine, Calif.-based solution provider and longtime Hitachi Vantara channel partner.

“Hitachi Vantara’s message has stayed fairly consistent,” Cerniglia told CRN. “It’s all about the data and what do you do with that data. You have these enterprises and mid-size companies out there collecting massive amounts of data. And what is AI? AI is just another way of how you leverage that data, and how do you leverage those workflows to make an organization more profitable.”

Cerniglia said Hitachi Vantara continues to unveil technologies that are increasingly robust and scalable.

“Hitachi focuses on the scalability, the performance, that they can deliver to their customers with these massive amounts of data,” he said. “So to me, that’s the AI story. It’s really, how do you turn a company’s data into intelligence? How do you turn that data into being an asset? So I’m excited. I think it’s another way for us to be able to offer customers more for what they’re looking for.”