At Data Center World 2023, Data Center Knowledge’s Steven Hill connected with Greg Stover, global director of high-tech development at Vertiv, to learn how the company is contributing to the modular prefabrication trend.
Their discussion revolves around the topic of modularity and its application in data centers. The discuss the use of prefabrication and modular offerings in both large cloud-scale data centers and enterprise contexts.
This is Part 2 of their discussion. Click here for Part 1.
The transcript follows below. Edits have been made for clarity and length.
Steven Hill: So, the next question is modularity. Is this prefabrication something that is only for large, cloud-scale data centers, or is this something that can be used in an enterprise context?
Greg Stover: Great question. The answer is yes, yes, yes. You see large cloud guys creating hubs. For them right now, the hubs are colos. Those are their spokes. But some of them are looking at modular solutions. Colo guys are developing rather deployable modular solutions to create their hub-and-spoke environment.
But for enterprises, it's a great solution. Some of them gave up half of their IT infrastructure because they moved everything to the cloud or maybe they're half cloud, half colo. And now, they have new needs and they're repatriating some stuff out of the cloud. And because it's latency-driven often, where do I put it?
Another thing that we saw recently that's very interesting is that the demand is so great, and all the big boys are taking down all the available capacity. Some enterprises are having a difficult time having colos even care about their business at a small retail level.
Steven: Because there's so much competition.
Greg: Yeah, if you're a large colo operator and you can sell one guy 50 megawatts, why go get 50 customers, right? You look at the growth of Microsoft, Oracle, Google, Meta, and all these guys. They're taking down tons of capacity. And I've got friends that I’ve known for 20 years that are telling me, ‘I used to be getting wined and dined and courted every day. Nobody's asking for my business because it's only 250 kilowatts or 300 kilowatts. Maybe it's a megawatt.’ That's a real problem because IT is now a competitive advantage.
For most of my career, IT was accounting, finance, and billing – important things, but an expense. Today, IT is a competitive differentiator. So, if there are new technologies, if there are new applications that help your business be a better user experience, you can't wait. You have to deploy it and modular is a great solution for that.
Steven: So, it's not just a niche, though, because again, you look at it and modularity has been in the housing industry for as long as I've been alive.
Greg: It started as a niche in housing. Now it’s a massive best practice.
Steven: Precisely, and the same thing you think is the direction that they're moving in data centers?
Greg: Absolutely, and we're going to continue to build giant campuses, especially with generative AI. So, you look at generative AI and how that works – large language models, hugely compute intensive, hugely storage intensive. Those 500,000-square-feet, million-square-feet data centers, we're going to keep building those. But you think about inferencing, and I've heard this a lot right here at the show. Probably four or five people that I consider subject-matter experts, learning is going to get done in the cloud. Inferencing is going to be done at the edge because inferencing is very latency attached.
Steve: And latency seems to be the big issue for any kind of task like that.
Greg: You think about 5G. What's it doing in solving for latency? So, you think about what will drive the need for modular data centers. And why do I have to have compute in all these different places? We have IoT, millions of data points. Do I want to ship all that back to the cloud? No, I want to normalize it at my site.
You're paying for your traffic back. It is just going to get thrown away. So, if I can normalize data at the source – this goes to the cloud, and this gets stored. This needs to be acted on, so I act on it now. This doesn't need to be acted on, so I can dump it. Makes a ton of sense that it's going to drive the need for more compute everywhere.
I mean, I've got a couple of what I call the disruptors. They view a micro data center on every corner.
Steve: I could see where modularity would be a key factor for the edge. This is something that I always found interesting when we started looking at IoT and trying to define where the edge is. That's a long conversation we could have at any point. But ultimately, it seems to be the place where actions first need to be taken, where the response will affect something at that point. And latency comes into play there.
So ultimately, this is the one thing that's odd about it: We've been spending decades building large, secure, highly redundant data centers, and now we just want to grab those things and put them out on the street. So, that's where the question is. How can we build the same level of security, the same level of redundancy, and the same level of protection on smaller or micro data centers or medium-sized edge?
Greg: Well, think about cloud zones. People used to say, ‘I need to have a redundant data center, redundant DR site.’ And some of the CSPs, they can fail a whole region and the workload just moves to the other region. And why would that not be true for edge sites? The software and the technology are there. If I've got 50 edge sites, am I ‘1 + 49’ now? Theoretically, I have that. And all the things that are critical in a mission-critical space we can do in a modular space.
In fact, in some cases, especially versus legacy buildings, we can make it more efficient because we can match the pieces up. Think of how much enterprise compute ends up in spaces that weren't purpose-built. It's all over the place. And now we've got AI coming. We've got generative AI coming. We've got IoT and 5G growing.
So, we're all going to have to deal with it. Data is only good if you can make it actionable.
And here's another thing with the colos, and all of us are experiencing this: Inflation is real. We all have to pass it through. But supply and demand are real, as well. And that also affects pricing models. So, what used to cost X, now is X plus 30%. That's very real.
I can bring in a mission-critical facility that has best-of-breed power, best-of-breed cooling, best-of-breed management, and systems that are matched. They're assembled and tested at the factory. So, less installation out at the site. And candidly, with the labor shortage, all the good operators are getting good talent. But who are the other guys?
Steve: The competition is everywhere.
Greg: Right, and we can bring scale. Maybe we're building you 10 modules and that will show up in the costing model. And we do all that work, and we test it so that shortens the commissioning at the other end. We ship it out of the door, and we know it works.
Steve: Absolutely. But again, it's already designed – ideally designed – even though the fact that it can be put together as a modular construction.
Greg: Right, it’s Legos. And we can do as much or as little as they want. And we can modify it, and if there needs to change, we can adjust it. So, right now we're at a point where we've got, I don't know, 10 or 15 reference designs, different customers, different things. So, we're trying to make it as simple as possible for people that don't want to be in the infrastructure business but need to have it to be able to deploy it successfully and be comfortable with it so they can sleep at night.
Steve: That brings up an interesting point. What are the drivers in terms of being able to justify it? What's our ROI?
Greg: Everybody's getting better at what we're doing because efficiency is No 1. Sustainability has to be part of the equation. We play well into the sustainability piece. We play well into the efficiency piece because, again, we can do it at scale. We can do it over and over. We can take the labor out of it. The last number I saw was it is arguably 30% more cost-effective. It's certainly faster. I mean, the constraint for everybody still is the supply chain and hardware.
But take the hardware supply chain out and go back to normal days. If somebody was going to try to build a legacy stick-built site, the best we're doing it in is six months. The average might be 12 months and the normal was 18 months. But now people want it now. And the cloud guys can do it now, but there's a cost that goes with that. Are my applications easily exportable to the cloud? We're learning what apps belong in the cloud and what ones don't belong in the cloud.
Steve: Right, and that's how it's going to be an ebb and flow.
Greg: Yeah. Well, it's just learning, right? I mean, everybody, in the beginning, said, ‘I'm going to put everything in the colo,’ and the cloud came out. ‘I'm going to put everything in the cloud,’ and the hybrid world is not going away. It's going to be cloud. It's going to be colo. It's going to be enterprise. It's going to be edge, DR sites. It's going to be a whole ecosystem.
At the end of the day, though, it all has to work. And you have to be able to take all those redundancies, all of those security issues, and whether it's a 500,000-square-foot field, a 4x4, or a 40-foot container, you have to have those things. So, I think there are great ways for the modular prefabricated space to meet those requirements.
We need it faster, we need it cheaper, and we need it more efficiently. Usually, in the past, you couldn't get all those things. It's like, which do you want? Because you can't have all three, but modular allows that.
Steve: It should be a lot closer to being a reality.
Greg: Yeah, you can get that. ‘I need it fast, I need it reliable, and I don't want to think about it.’ And then you think about the new technologies that are coming, and that's very disruptive to an enterprise user. Our colo customers and our cloud customers are very adept at the latest and greatest technology. But if you have a legacy site that was built for one-and-a-half to three-kilowatt racks, and your CIO says, ‘Hey, this generative AI thing is saving companies and we're going to do it, right?’ And BDS and all the other companies are making this awesome disruptive technology. And they need to have it in their building. That building was not built to accommodate a 30-, 50-, 100-kilowatt rack.
Maybe the better solution is to build a high-density pod. And I'm going to drop it in the parking lot and let that be my problem, right?
Steve: Right. And right from the very beginning that it'll be able to handle the equipment.
Greg: Yeah. Well, even in some legacy colos, you know, how they're managing high-density racks now. They're spreading the load. I can't put a 10 in a row.
Steve: No, put one here, put one there. Wherever you've got the capacity.
Greg: Not efficient. Not optimizing the space. I could put 10 racks in a container, 50 kilowatts a piece. There are 500 kilowatts and drop them in your parking lot. And we have remote management tools and monitoring tools and all those things that customers expect from us so that they can sleep at night.
Steve: Yes. Well, this has been a fascinating conversation. Greg, it's been a treat to talk to you.
Greg: I've enjoyed it very much.
Steve: Excellent! Enjoy the rest of the show and thank you all for tuning in today! And we'll see you at later shows. Have a great day.
Greg: Call your local Vertiv of office so we can help.
Steve: There you go.