AUCKLAND 2023: Speaking with Bill Kleyman on the Impact of AI on the Data Center

Bill Kleyman boasts impressive career credentials, including spells at Switch, MTM Technologies, and currently at Neu.ro Inc., a company at the forefront of AI, ML, and Generative AI development. His commitment to the data center industry is reinforced by his work with the Infrastructure Masons, AFCOM, and the Nomad Futurists. His level of attainment may, in part, be explained by his commitment from an early age to working in data centers: “I’m a rarity in this industry because I wasn’t adopted into it. I wasn’t pulled from other industries or anything like that. I came here, quite natively actually”. 

Bill will be basing his keynote at the W.Media Convention in Auckland on 2nd November on his considerable experience in digital technologies and data centers, and he will identify some significant upcoming challenges in the marriage of the two. 

He sees the revolution AI is causing as fundamentally changing how we interact with data. While AI is not a new technology, he states that what seems to have ‘come out of nowhere’ is the intelligence of the responses: “We can actually have conversations with and generate real viable content. That’s what really took this entire industry by storm: the idea that you can literally ask questions on your data. AI is not doing the thinking for us. They are helping us interact with data more intelligently and more effectively. For the first time, they’re helping us ask questions about our data.”

So, what will be the impact on data centers? Again, the answer is simple, even if the implications are not: “Every data center will become an AI data center. It all depends on how quickly you can get there. The demand for Neu.ro’s ‘white label’ AI platform that gets deployed into data centers is not driven by demand from data centers themselves, but from their customers, coming and saying: ‘We don’t want to be in the cloud. We’re being charged so much money. We want to hold data locally and privately; we don’t want information to be used to train some large open AI foundational model’. But when we [Neu.ro] go to our colocation partners, they say, “I have nothing for you – I don’t have the capacity, or I’m not ready for that type of density”. The situation is referenced by one of Bill’s mentors, Peter Gross, quoted by Bill that “The ‘datacenter industry loves innovation as long as it’s ten years old” and you’re going to hear me say that is really the challenge – we don’t have ten years, we don’t even have ten months.”

He sees this emerging source of demand presents challenges for the data center: “They have to figure out a way to enter this market, but it’s challenging. Running at 9 or 10 KW will not cut it for many modern AI applications. You have to rethink your architecture.To be viable in this market, you must be 30, 40, 50 kW or more, and you’ll have to innovate. This is what I’m going to talk about”.

Bill references the challenge in the title of his talk: “Nonstop Innovation,” stating that “I could have titled it, you know, ‘putting the tires on the car as it’s going on the highway.’ And so, all of a sudden, this past year has been about a critical business question: How do we accelerate the adoption and utilization of data center systems to support this extraordinary trend? We’ve never seen this before – again, where the market is not entirely being driven by data centers but by both the data and the end user”.

Survival may be based on the speed at which data centers adapt to new demands. Bill continues: “And the big question is existential: do they move to you? Or do they move to one of your competitors? The challenge for many of these data center partners is that they’ll put in a bunch of GPUs. We are the overarching platform, the white label service that sits on top of those data center GPUs, and what we’ve done is we’ve gone to the data center world, and we said, be good at what you’re good at. So capacity, space, and power, don’t change a damn thing. But put those GPUs in as long as my software is already on top”. 

His presentation will focus also on the role of markets such as Australia and New Zealand. “For model training, big GPU farms don’t need to be anywhere near the data source. You’d need to get the large data set to your facility, but after that, it’ll spend weeks and months being trained and fine-tuned. Why would you train the system in Northern Virginia, where you’re paying crazy amounts of dollars per kilowatt, or California or Texas or anywhere in these very dense markets? So what I absolutely believe is that these secondary markets are going to start to see data center development”. 

To describe what he hopes the audience takes away from this presentation, Bill references the statement that “Vision without execution is just hallucination. What I want people to understand is, number one, that AI is not a bad thing –  especially when you begin to understand the amazing use cases. And two, find a gold nugget in what I’m talking about. Take it away with you and act on it. Execute on what you saw today, even if it’s just one thing, because then it’s no longer a hallucination. You’re acting on these trends”. 

Bill indicates that at the foundation of what he will speak about is “a deeper appreciation and understanding of just how far our commercial industry has come even in just the past few months and what that means to everyone in that room.” 

To learn more and to register, please visit https://w.media/events/new-zealand-cloud-datacenter-convention-2023-2/

Publish on W.Media
Author Info:
Picture of Nick Parfitt
Nick Parfitt
Share This Article
Related Posts
Other Popular Posts
Southeast Asia News
South Asia News
Southeast Asia News