“As time goes on, you’re going to see the generative AI really drag not just compute systems, but the traditional core servers as well as traditional storage, because everything is going to be optimized for AI,” Dell ISG president Arthur Lewis told Morgan Stanley Technology, Media & Telecom Conference this week.

Dell ISG President Arthur Lewis said cost, performance, and security policies keep 83 percent of the world’s data on-prem, giving it a gravity that plays into Dell’s position as the market leader in servers and storage once that data has AI applied to it.

“What we see is a lot of customers are testing out workloads in the cloud just to prove out concepts. But when it comes to putting AI in production, they’re coming to us to help them think through an on-prem solution,” Lewis said at the Morgan Stanley Technology, Media & Telecom Conference on Tuesday. “Given the breadth of our portfolio, our services capability, and our other durable advantages, we feel like we’re very well positioned to capture the opportunity.”

Dell is sitting on an ISG total addressable market (TAM) of $265 billion inside of a market growing at 7-percent. Dell expects its own ISG revenue to grow into the mid-teens this year, driven partially by the sale of AI servers and storage. According to market research firm IDC, the generative AI TAM is $152 billion by 2027, growing at a 20 percent compound annual growth rate. Even that growth Dell feels is a “lagging indicator” based on demand it sees.

[RELATED: Dell COO Jeff Clarke Says 2024 ‘Didn’t Go As Planned’ But Expects AI-Fueled Rebound]

“Obviously, growing faster than the average. And obviously, a lagging indicator, right? Because customer demand continues to grow, and we would expect that demand to grow. And therefore, AI is the significant tailwind that we called back in October.”

Dell won $1.5 billion in AI server revenue last year and has a $2.9 billion backlog of AI servers at the end of the fourth quarter. He said this early AI business is taking a couple forms.

“There’s sort of two games that are being played right now. I think in the Tier 2 CSP community – and these are the CoreWeave, the Denvr Dataworks of the world that are building out large training and inferencing infrastructure to support GPU-as-a-service to their end customers – we’re in the early innings of the baseball game. But when it comes to the enterprise, we’re in the car pulling up into the stadium and the teams are still trying to figure out the rules of the game.”

To do that, however, Dell is being leaned on by customers with talks that evolve naturally from policy, to infrastructure, data storage and data security.

“We’re having conversations with enterprise customers that start way before the infrastructure conversation. We’re talking about, ‘Hey, Dell, what use cases are you guys seeing?’ ‘Hey, Dell, how do you think about model selection?’ ‘Hey, Dell, how do you think about data preparation?’ ‘How do you think about fine-tuning?’ ‘How do you think about RAG?’ And then you get into an architecture conversation and then you get into an infrastructure conversation,” Lewis said.

As the expectations of the technology have gone up, the conversation around adoption is driven by board-level and C-level executives, he said. This leads to talk about how to incorporate AI into other areas and more hardware conversations.

“Hey, I should really be thinking about modernizing my operations of the business in order to make generative AI that much better. I need to be thinking about streamlining my processes,” Lewis said. “I need to be thinking about automating my processes. And then I need to be thinking about generative AI.” I think this is going to drive a massive change in our industry. It’s going to drive a massive change in infrastructure. This goes well beyond AI-optimized servers that we’re talking about today in the context of Tier 2 cloud service providers.”

GenAI Being Put To Work In Every Industry

Lewis said GenAI is being used by every industry including financial services, manufacturing, retail and healthcare. The early use cases are focused on four areas: software development, sales chatbots, customer service and content creation.

“Software development comes up a whole lot. ‘Hey, Dell, how are you thinking about software development? Where in the software development process do you think generative AI plays the best role?’” he said.

Lewis said customers are also looking to use AI to save the costs associated with content creation.

“Something as mundane as content creation,” he said. “You’d be surprised, and I certainly was surprised, how much money we at Dell spent on this and how much money customers say they spent on this, and the amount of savings that they could realize through deploying generative AI.”

Enterprises are experimenting with other use cases using cloud computing, however keeping the compute close to the data is more cost effective.

“If you think about the majority of the world’s data today is cold, sitting in backup and archive, we think the equation of that’s going to flip, where more of the data is hot/warm and is constantly circulating, feeding the AI engine, which is going to cause a significant retooling of data centers around the world,” Lewis said. “So, again, this is, like, an incredibly exciting time to be in technology, and I think we’re extremely well-positioned to capture the opportunity.”

Why Dell Has Found Early Success In Generative AI

Lewis said Dell has fundamentals that it has gotten right, beginning with its PowerEdge XE 9680 Rack Server, which launched with Nvidia’s flagship AI chip.

“We spent a lot of time on the architecture of this product. We had some pretty aggressive design points. Design point number one, it had to be the densest platform in the industry,” he said. “Design point number two, it had to offer silicon diversity for all of the various GPU suppliers. Design point number three is that it offered diversity for networking and connectivity. And design point number four, it had to offer complete flexibility for thermals, moving from air cooling to liquid-assisted to direct-to-chip cooling,” he said.

Dell’s other AI-capable servers such as the 9640, the 750xa, the 760xa, are paired with the storage portfolio like Dell’s PowerScale F210 and PowerScale F710, which can meet the increased density, performance and I/O specifications needed for generative AI.

“What’s also important here is that we have the ability to co-engineer this system. Because generative AI is a system. It comprises compute, networking, and storage. And the ability to engineer that under one roof is incredibly important,” Lewis said, “because if you’re trying to create this system through a myriad of partnerships, that level of collaboration that is required assigns a very high beta to that project.”

These in-house engineered projects are turned into Dell-validated designs, blueprints for creating complex systems for a variety of GenAI applications.

“So, if a customer comes and says, ‘Hey, I’m looking at a sales chatbot’ or ‘I’m looking for a customer service’ or ‘I’m looking for something content creation,’ we have the ability to talk to them about a validated solution that comprises compute, networking, and storage. That’s the foundation of our differentiation,” he said. “On top of that, we have the ability to build out the ecosystem. We have the only partnership in the industry with Hugging Face and we have the only partnership with Meta and Llama 2, and that’s going to become incredibly important, as indexing data is incredibly important when you think about the generative AI system.”

He said Dell is bringing all of its resources to bear to enable GenAI adoption at the enterprise.

“We also have the ability to provide financing through Dell Financial Services, a one-stop shop for product and financing,” he said. “And then, lastly, but also extremely importantly, we have the industry’s largest sales team – not only largest, largest and most knowledgeable sales team – and we have world-class supply chain that is wielding the power of Dell’s scale. You combine those five things together, and we very much like the hand that we’re playing.”