your123 - stock.adobe.com

Dell enters generative AI market with Nvidia partnership

The initiative provides enterprises with both Dell servers and Nvidia's H100 GPUs and software. It could appeal to Dell users that want to create generative models on premises.

Dell Technologies has stepped into the generative AI market.

The computer hardware and software provider launched a joint project with AI vendor Nvidia to make generative AI capabilities available for on-premises Dell customers for their business needs.

Dell revealed the partnership, dubbed Project Helix, on May 23 at the Dell Technologies World 2023 conference in Las Vegas.

Customers will have access to Dell servers such as PowerEdge XE9680 and PowerEdge R760xa for training and AI inferencing as well as Nvidia H100 Tensor Core GPUs and Nvidia networking. The project also includes Nvidia AI Enterprise software, which supports Nvidia's large language model framework, NeMo and NeMo Guardrails for building responsible generative AI chatbots.

AI in a box

"What Dell wants to do is basically give you AI in a box," said Jack Gold, an analyst at J. Gold Associates.

Compared with cloud providers like Google or Microsoft, Dell focuses on enterprises needing their own infrastructure.

"Dell is throwing in the hardware piece as well and saying, 'We'll sell you the whole box with the software,'" Gold said.

While other vendors, such as independent SambaNova Systems, also offer complete generative AI packages (with both the hardware and software), Dell's differentiating factor is its momentum with its Apex service portfolio, Gold said.

However, more providers are expected to offer services like these soon, he continued. "But if you're a Dell house, if you're buying from Dell, it's attractive to buy from Dell because now they're giving you a capability that you didn't have before."

Some drawbacks

The initial expense is one drawback for enterprises looking into an on-premises generative AI infrastructure. However, the cost could still be lower than running LLMs on the cloud.

Dell is throwing in the hardware piece as well and saying, 'We'll sell you the whole box with the software.'
Jack GoldAnalyst, J. Gold Associates, LLC

"It can be more affordable in the long term, depending on what you're doing, because you're not paying for service providers' profit margins," said Karl Freund, an analyst at Cambrian AI. "So while it costs more upfront to do that, in any kind of significant-scale operation, on premises is going to be cheaper."

Another drawback might be security, Gold said. Enterprises will have to trust that what Dell and Nvidia provide will be secure and not feel at risk of losing data in the case of a breach.

Enterprises may also have to rely on Dell's IT team to ensure the infrastructure runs smoothly, Gold added.

However, that can also be an advantage to enterprises that need more talent, he said. Enterprises that need more IT employees to manage their infrastructure will appreciate having Dell's team help them with that task.

For Dell, the main challenge will be proving ways its infrastructure stands out, even though it is one of the biggest original equipment manufacturers.

"Each OEM will find a way to do so, but it may not be obvious to IT users," said Daniel Newman, an analyst at Futurum Research. "It will be on Dell and others to prove its differentiation."

The Project Helix system will be available through traditional Dell channels and with Apex flexible consumption options starting in July, according to the vendor.

The Dell conference started May 22 and is being held in person and virtually until May 25.

Esther Ajao is a news writer covering artificial intelligence software and systems.

Tech News This Week 06-02-2023

Next Steps

Business efficiency a place to start with generative AI

Dig Deeper on AI business strategies