The Slingshot Effect: Today’s Mission Critical Market
The post-pandemic data center boom isn’t slowing down, so how are hyperscalers, builders, and manufacturers keeping up? Join STO Mission Critical Project Executive, Jim DiNoia, as he discusses the influx of mission critical work with SVP of STO Mission Critical, Martin O’Neill, and Vice President of Data Center Engineering at Sentinel Data Centers, Nick Marzorati.
HOST
James DiNoia
Project Executive, STO Mission CriticalView Bio
GUEST
Nick Marzorati
Vice President, Data Center Engineering, Sentinel Data CentersView Bio
GUEST
Martin O'Neill
Senior Vice President, STO Mission CriticalView Bio
Introduction (00:05):
Welcome to Building Conversations, a construction podcast powered by the STO Building Group. On today’s episode, STO Mission Critical Project Executive Jim DiNoia discusses the Mission Critical construction landscape in the years following the pandemic with SVP of STO Mission Critical, Martin O’Neill, and Vice President of Data Center Engineering at Sentinel Data Centers, Nick Marzorati.
James DiNoia (00:35):
Hi everyone. My name is Jim DiNoia. Welcome to Building Conversations. I am a project executive with STO Mission Critical. STO Mission Critical is a brand of the STO Building Group that specializes in data center construction. I am here with my colleague Martin O’Neill, SVP of STO Mission Critical, and our distinguished guest, Nick Marzorati, VP of Engineering with Sentinel Data Centers. Today we’re going to just talk about a subject that we are calling the Slingshot Reaction of the Covid-era. Let’s put it that way, for the last few years where it’s gone, where we are now, and where we may think it’s going. Today, we’ll bring up some things that we hope that will inspire or be intriguing to the audience and from an engineering, construction, and owner’s perspective. So let’s start with that and we’ll have Nick and Martin introduce themselves.
Nick Marzorati (01:35):
Nick Marzorati, VP of Data Center Engineering at Sentinel. I come from a consulting MEP background. I graduated Stevens Tech in 2011. I took a job working consulting engineering for the last ten or so years and moved over to the Sentinel role a little over a year ago. So, my focus at the consulting firm was predominantly data center mission critical work. So that’s kind of how I landed in the role at Sentinel Data Centers.
Martin O’Neill (02:03):
Martin O’Neill, Senior Vice President of STO Mission Critical. Been working in the industry building data centers for the last 25 years now. I’ve had the pleasure of working with Sentinel for, geez, I think it’s almost 20 years now, building data centers all over the U.S.
James DiNoia (02:20):
Thanks guys. I think what we’ll do is hit our first question, and I think Martin, you’re going to start with this one, and then Nick, you can chime in. The first thing is, I think it’s been talked about the effects of COVID, but since it arrived three years ago, we felt changes in the impacts to the landscape and the Mission Critical industry. There’s been a consistent rise to the effect of projects, conception, and execution. So where would you take that, Martin and just talk us through the different things that have happened and what we’ve experienced.
Martin O’Neill (02:51):
Yeah, sure. I think you’ve said it too already, Jim. Covid affected all verticals in the construction industry, not just Mission Critical, right?
Martin O’Neill (02:58):
We became a very social industry to a very virtual industry, both on active projects and prospective projects. What we saw specifically from a mission critical standpoint was just the volume and the size of the projects that were being put out on the street. Historically, data centers, you know, they are large scale, but there was a lot of smaller, medium-sized data centers from an eight-megawatt IT, critical IT load to the point where now most of the RFPs that we’re answering are for 400,000 square feet, day one, multiple story, and probably 36 to 40 megawatts day one of critical IT. And that hasn’t stopped, it hasn’t slowed down. In fact, it’s accelerated. We’re seeing that from more hyperscalers, but we’re also seeing it now on a vast amount from the developer end model. That was a big change for us, and I’ll say kudos to all of our teams on all of our platforms for being able to deal with that, responding to RFPs and executing work virtually during Covid, but then post-Covid transitioning back to it, I think one other thing that we saw as a big change during Covid and post-Covid was the size and the scale of the data centers drove the energy providers, the utilities, to max out in certain areas.
Martin O’Neill (04:15):
So for us, from a geographical standpoint, we were looking at new regions that we had never worked in before because, you know, everyone’s seeing what happened down in North Virginia, right, with the utility and the transmission and the stress on the transmission lines, and the fact that we can’t get that power to the sites anymore in a quick fashion. So people have been looking at different locations. And then also I think Nick, you’ll agree the cost of land, I mean, it’s gone through the roof. So geographical reach for us has changed dramatically. But again, you know, it’s part of what we’re doing part of the industry.
Nick Marzorati (04:51):
Yeah, I mean, just kind of piggybacking off of the slingshot – when Covid started, right? It was really scary for everybody. There was the kind of everybody work from home, you almost thought there was going to be kind of a lull, a slowdown, and I think we’re calling it the slingshot because we weren’t really, we weren’t really expecting there to be an influx of all these data center requests, RFPs that are coming through.
Nick Marzorati (05:14):
And so, my hunch is that really it was kind of data center developers and customers and some of the larger guys, they really wanted to hedge their bets and push each other to kind of get into that next tier. You almost saw this exponential growth through the data designer through the Covid era. But part of the issue was that it was creating a lot of labor shortages because the factories were kind of pulling people out of them. And as a result, you kind of have these two compounding things. You’ve got an intense amount of RFP requests coming in, but then the factories and fabricators, everybody kind of slowing down a bit to kind of gear up for preparing for the pandemic. And so you kind of ended up in a very interesting time.
Nick Marzorati (05:57):
And then post-Covid, it almost feels a bit like we haven’t caught up. Production hasn’t caught up with the requests. There was almost this one- or two-year backlog, and now you see that with equipment lead times, right? And you kind of see that with almost everything all the way up from the supply chain all the way up and it kind of pushed stress and not even just the build side of the clinical stress on the design side as well. Because, you know, at the time I was at an MEP consulting firm during Covid, and I can say we were, it was really, really tough to keep our head above water. The requests were just coming in. And so, I think that that’s still happening now, right? And so, I think it’s interesting to kind of see how this current state of everybody being overloaded, factors being overloaded, lead times being two years out, how everybody starts to adapt to that. And then the kind of traditional project approaches. They all start to change and form around that. But really interesting and exciting to be a part of it and see the growth just exponentially just expand.
James DiNoia (06:57):
Yeah. I’m starting to wonder if the slingshot has actually released yet, or if it’s still ramping up over the next five years.
Nick Marzorati (07:04):
Yeah.
James DiNoia (07:05):
But all of us in the data center business have seen exponential growth in the industry. I think we all agree that the scale has a macro impact to the supply chains as we were talking about with your fabricators and stuff. How has it changed the facilities, how they’re designed and what are the major challenges, risks, and strategies from the construction standpoint?
Martin O’Neill (07:26):
Well, I think Nick touched on a few of those. Jim, historically, data centers have always had long lead times when it comes to what I call big iron, right? The generators, the UPS, the switch gears, the chillers, all the big kit. And when I say long lead times, I mean like 50 weeks, not 92 weeks, which, or a year plus right now is normal or approaching two years. But one thing that we did notice, and again, it’s not just unique to the data center industry, it’s unique to the whole construction industry, but it affects us, too. The cost increase in the lead time challenges now started to affect structural studs, structural steel, precast roofing, stuff that we never had a problem getting became a problem for us. So that was something that we had to start building into our schedules way in advance. Logistically, how do we build the building with limited materials on hand, right? We’ve had some clients that have done a really good job of pre-purchasing a lot of the, whether it’s the exterior skin of the building, right? The precast, the steel, all that good stuff. They’ve been able to get that ahead of the RFP ahead of the CM and then assign it to the CM’s we have it there.
Martin O’Neill (08:35):
I think another big challenge is it has been cost. We’ve got a major issue with schedule, but we’ve also got a major issue with cost. So when we budget out a project day one, and then six months later when the design’s finalized, that day one may pencil out for a deal and then six months later that may not pencil out. So, the costs have gone crazy. I mean, structural steel, I think structural steel went up 45%, right? Not quite 50%, but it almost doubled. Roofing doubled. We’re seeing some stabilization in some of those materials, but still on the mechanical side, the electrical side, it’s pretty scary being able to obtain those parts and things that we need to execute the work.
Nick Marzorati (09:24):
Yeah. On the developer side, from a schedule perspective, it’s just getting really challenging to plan. So now lead times for you know, back in the day you’d always pre-purchase, like you said, all the gens and transformers and stuff. And so, it almost seems now that the pre-purchase scheduled things have gotten such long lead that we’re way off from the traditional project schedule, meaning from customer engagement through landing the deal and in parallel lining up the equipment showing up. So, what we’re seeing it kind of doing is putting a pinch on the design side because you want to remove risk on the build sides, but you kind of have to take the risk and procure some of the equipment before the deal has come to fruition. And so, the way that you manage that, at least from my perspective on the design side, is you kind of start to adapt a more of a prototypical approach where you kind of have equipment that could be potentially teed up for the new project that’s coming, but maybe it’s kind of going to work for the project that you’ve kind of already got. In the rear-view mirror if you will. So that kind of prototypical approach, it’s weird. It’s like all these big supply-and-demand kind of tug-of-war situation that’s happening. And then I think that changes how the buildings are designed, right? So, they’re no longer super niched to a specific client. You kind of start to see a more widescale standardization, such that your equipment procurement process can line up with the actual project schedule that has to come about. And then the cost volatility, that’s another scary one. I mean, it makes planning and budgeting and all the conversations that you have with your customer much more challenging because you’re educating your customers on the costs. We’re talking two, three times the cost of what it used to be, some of the things. And so, your customers are basically, a lot of them balk at it, right?
Nick Marzorati (11:05):
They can’t believe it. And contractor bids coming in on just the mechanical and electrical and stuff like that. And we’re still kind of in this rapid growth period where we haven’t really got those supply chains worked out. So, a lot of the balls are in the vendor’s court, right? They hold all the keys for the lead times in the schedule. And so, we’re left trying to cater to that via modifying the designs a bit.
Martin O’Neill (11:28):
And I think something that you touched on too, Nick, which is really important for us, is that it’s the interest in bidding work right now because with the strain on, on the trades on the vendors, not only Mission Critical healthcare booming as well, what we found is, you know, we’ll go out to certain regions and really get, you know, we’ll get interest, but not as much interest as we would get in the past, right? And so, I think the best success we’ve had is where we’ve been able to partner up with a client and they’ve been fully transparent with us on the pipeline. So, they’ll give us plenty of advanced notice on work in different regions, size, scale, all that good stuff. And we’re able to forecast that and be transparent with them and say: Hey, you know, this is something that we can go do for you, right? But then also socialize that with the right subcontractors and vendors so that they get advanced notice to say, this job is coming versus just getting an RFP that hits the street and you’ve got four weeks to respond.
Nick Marzorati (12:26):
Yeah, for sure. It kind of feels like everything is moving towards a design-build. And I know on the opportunities we work on with Structure Tone, we’re engaging you guys almost the minute that we get the RFP, right? We’re engaging in constructability discussions and lead time discussions and what’s the latest on equipment and pricing and manufacturing. So you almost have to in this climate.
James DiNoia (12:46):
We found that all across the board, the developers want to get the contractors on board as soon as possible, or at least get their feedback as soon as possible to keep going, to get things ahead of schedule. Nick, let’s keep you on a roll here to talk about trends, technologies, and we see things stand out more than others for changes.
Nick Marzorati (13:06):
Yeah. I think it’s a really expansive question, so starting from when I kind of got into the mission critical industry, you know, call it a decade, a little bit more than a decade ago, we were building one to two to three megawatt halls, probably at the largest.
Nick Marzorati (13:21):
And so that would be probably for buildings that were more or less going to end up being 10 to 20, I’d say at the tops megawatts total. So now, you know, it seems like gone are the days of those smaller halls. It’s like if you’re go through going through the process of standing up the shell, standing up and, you know, coordinating with all the utilities, it seems like everybody’s pushing into this five to seven to nine or even 10-megawatt hall sizes. And I think a lot of customers are kind of sometimes even going above that. So, that kind of changes the way that the buildings are looking over the last five to seven years. And so, it’s almost a result, like I said, this is why it’s called the Covid slingshot, if you will. But in the fall and the wake of the, the pandemic, the mass expansion, everybody wants to build the biggest and the most scalable, the largest, and they want to really get the most out of the lemon, if you will. So, you see a lot of larger halls, but with that comes people wanting to optimize the building shell themselves, right? And so, you’re really going towards kind of increasing the density as well. So not only are the halls getting larger, meaning they can support more IT capacity, but they’re also increasing in density. And so, density 10 years ago is probably in 100 to 200 watts a square foot. And now you’ve got a lot of customers that have requirements that are pushing you into the 350 to 400 and some other customers are in some other projects are really trying to squeeze it down.
Nick Marzorati (14:44):
We can kind of get up to 500 megawatts a square foot, which is, you know, I think doable. But I think it starts to push the envelope on some of the technologies. And I think another trend you’ll probably start to see, and you know, the advent of getting those data halls and so dense and so large really is, it was kind of a perimeter cooling problem to solve for a lot. And I think perimeter cooling is a strategy that we’ve seen for forever, right? So typical downflow crawl units to a raised floor, but it almost seems now there’s no more raised floors, right? So, I don’t know about you guys, but not a lot of raised floors are being built, so it’s hard to achieve the higher densities and the higher capacities with traditional raised floors.
Nick Marzorati (15:20):
So, you’re starting to see a lot of fan wall utilization. So, I think that the introduction of the fan wall style cooling really helped get data center halls to the capacities that you’re starting to see today. So, I could keep going if you’d like. It kind of depends.
James DiNoia (15:37):
Why don’t we have Martin chime in and then I’ll come back with another question on that for you.
Martin O’Neill (15:40):
Yeah, I think one of the big trends we certainly are seeing in addition to design, but it’s all about logistics, right? So, we’re seeing a lot of developers asking CS to come on board early to get pad ready, right? So instead of putting the shell up right away, it’s get it pad ready so that, you know, we rub the side out, we get it level, we get it ready, and then that way the developer’s able to go and respond to the RFP very quickly versus, I mean, on that front end, you probably have four to six months of work, right?
Martin O’Neill (16:07):
Depending on the size and scale of the site. So that’s definitely a trend that we’re seeing in the industry that we didn’t see before and prefab, right? I mean, that’s not new. That’s been around for a long time. People have been pushing it, but we’re doing a lot more with the limitations on labor. We always used to look at it and say, if you can reduce the labor in the field, you increase the safety as well, right? Because you’re putting into a, into a more controlled environment, but now we have to because there are, there are limitations on labor and getting that labor to the available site. So, we are seeing a lot more prefab going out there. So skid mounts, whether it’s an electrical room, mechanical skid, that’s just becoming the norm now for us.
Nick Marzorati (16:46):
Yeah, speed to market or speed to build or speed to deployment is almost, it feels like it’s the only way to meet the demand. It can only really happen through technological advances. And so, I think another big one that we’re starting to see a lot of in water constrained environments, you’re starting to see a lot more air-cooled chillers that are primarily cooling the data centers. So, I think a few years ago the technology was still pretty inefficient, and people were struggling to get an air-cooled chiller to work with the latest energy codes and those kind of compliance paths. But, now the introduction of the Danfoss Turbocor compressor within air cooled chillers is kind of changing a little bit of the landscape of how data centers are coming about. And I think it used to be how some people used to do it in the past, you know, traditionally Sentinel, we’ve always loved to build water cooled plants and if we have an abundance of water we’ll always build water cooled plants soon we’d find them to be the most effective best CapEx and OpEx game for your buck.
Nick Marzorati (17:42):
So, but the advent of those new chiller technologies is really helping speed to market developers.
James DiNoia (17:47):
So you can bring in extra power as much as you want. The complication now going forward is the heat rejection, right? That’s really your main problem going forward.
Nick Marzorati (17:58):
Yeah. So I guess if you’re asking if we have unlimited power and no water, now the problem to solve is, yeah. So I think then you kind of get into some nice relationships between like a multi-story. That’s another thing we didn’t really talk about a lot, but a lot of these data centers that are going now are multistory, right? It didn’t used to be like that. Well, there would be those that were popping up back in the day, but now it almost seems like the standard, and we’ll get back to that point in a sec, but the reason why they’re going multistory is because they’re trying to get the best bang for the land.
Nick Marzorati (18:27):
Right? But what ends up happening is you’ve got to kind of find a spot for the heat rejection. And so, as you go vertical, you’re losing area to reject that heat. So yeah, every project’s a little bit unique still, but you have to find that balance between how tall can your building go, how dense can you get versus what you can fit on the roof practically, right? So that it’s not a super niched solution into something that pushes you away from the speed to market deployment. So a lot of variables at play to try to solve that, that problem, which is all part of the exciting delivery.
James DiNoia (18:59):
So that might need its own podcast, who knows?
Nick Marzorati (19:01):
Potentially, yeah. Potentially.
Martin O’Neill (19:03):
True.
Nick Marzorati (19:04):
Next week.
James DiNoia (19:05):
Yep. There you go. So let’s go on to our fourth question here. Have you or have you seen others apply or working towards more sustainable solutions?
Martin O’Neill (19:13):
I’ll do a little bit at front and then I’ll hand it over to the technician over here who can go into real detail. But certainly from a sustainability standpoint, one of the big things we’ve seen certainly as a challenge, right, is power generation on-site, is having diesel generator backup generators on-site in the quantity that they are now, again, back not even 5, 6, 7 years ago, right? It was unheard of to have 60, 70 generators supporting one, one facility, not only impacting the environment or the local neighborhoods as well, right? From a noise and a visual standpoint. So, what we’ve seen is a lot of people now looking at eliminating diesel generators and UPS in totality, right? And going to a fuel cell type of approach, whether it’s a Bloom Energy or someone similar to that.
Martin O’Neill (20:05):
So to me that’s a big sustainability thing. I mean, what’s the Achilles heel of a data center, right? It’s lack of power and water.
Nick Marzorati (20:13):
Yeah, on the sustainability. I mean, it’s definitely a consideration for every new build that goes in. I guess it just feels like with the size and the densities of the halls being, you know, the building’s kind of going up, there’s not yet a renewable energy source that can practically meets the demand of a data center. And so rather than considering potentially offloading some of the utility power to some sort of a renewable source, which is really tough to do with the 40 to 60 to 80 megawatt, you know, campus from Sentinel’s end, we look to achieve the lowest possible PUE. So, that for us comes from a couple different, couple different ways.
Nick Marzorati (20:52):
And so, it’s the advent of evaporative cooling and we use a lot of evaporative cooling. But we typically couple that with, and much of the industry is getting there as well, with much warmer temperatures in the halls. So, the warmer that you can deliver air to the racks, obviously the less mechanical cooling which you have to do. And so depending on certain geographies, you could potentially deliver almost the outside air entirely for the entire year. Obviously, there’s limitations to some of that and there’s issues that have to get worked out. But, from the overall sustainability play, we’ve considered the Bloom Energy and I think it’s a great technology if the gas is kind of available. And I think that’s, for us, the sustainability plays reducing our annualized PUE to absolute bare minimum as close to one as possible.
Nick Marzorati (21:34):
And so, we’ve got inactive, we’ve got a couple active facilities that have our water-cooled buildings that are boasting annualized PUE less than 1.1 on an annual basis. Which is, in my mind, this as sustainable as it gets, right? So, outside of generating your own utility power, right?
Martin O’Neill (21:52):
Yeah. And I mean there’s been a lot of discussion over the last 12 to 24 months about nuclear power, right? So, how does that play into this industry if it ever will? But standalone plants that will support those campus builds because, again, campus builds weren’t the norm back in the day. Now everyone that’s building or developing is putting up more than one shell. It’s normally three shells, four shells at a minimum, and it could be as much as nine or more total.
Martin O’Neill (22:20):
Right? There was probably a handful or less of folks in the industry that were doing those campus builds. Now it’s a norm.
Nick Marzorati (22:28):
Yeah. I mean the nuclear power you’re getting into this is kind of the macro planning for the expansion of data centers, starting from the early 2000s to now, and so, how are local utility companies scaling up to meet that? And we see certain areas are highly constrained, right? Your Northern Virginia utilities are extremely constrained. You start to see a lot of those pockets of constrained areas. And so it’s definitely a challenge. I do, nuclear, you know, we might be a ways away depending on how certain developers get there. I’m sure utility companies are starting to plan for that, this massive expansion and considering building that into their overall pipeline. New plants, new generation plants as their grids start to tap out.
James DiNoia (23:12):
That’s for sure. Last topic for this podcast, I think would be the future. How do we predict the future?
Martin O’Neill (23:18):
Can you give me some lottery numbers, Jim?
James DiNoia (23:23):
What kind of advancements do you see happening because of the chip design. You know, you see the NVIDIA stuff that’s come out every, seems like every week now there’s something, a new or an advancement or their stock is just exploding, but what effect does that have on Sentinel?
Nick Marzorati (23:39):
Yeah, I mean the future planning is a good nerd out topic, right? The AI certainly seems to be the hottest thing right now. At least from my perspective, we’re trying to understand what that means for a lot of our customers. And so, we have some theories about this, basically bolstering their current global production capacity increases. But this AI bug might be basically the new parallel path that a lot of these customers take. And so we’re kind of at an inflection point where you might just start to see all the big players start to double everywhere, right? Just doubling and doubling and continue to exponentially double how do you plan for that, right? While also trying to meet the speed to market and the supply demands, you know, those big swaths of supply and demand, issues. So I think that’s the hardest question to ask and answer is really how do you plan for that? For us it’s all about just acquiring land in relatively economic areas where we can, that aren’t as constrained from a utility power perspective. But also, we’re working on kind of prototyping and understanding pushing the limits on densification and things like that. But it seems right now the most practical method for heat rejection is still water. There are certainly some customers that are pushing the bounds, but for us it seems like water as the main medium for heat rejection is the move. So, I think the closer you can get that water directly to the racks, the best we will be in the future for planning. And so, we’ll still be likely employing a water-based cooling rejection method for a while as a response to meeting the unknown. But probably known exponential yield growth of the industry.
Martin O’Neill (25:22):
Have you seen much Nick, in terms of to go into a little bit more on your question, Jim, like emerging cooling, right? I mean, we’ve heard everything, we’ve seen responses that we had to put together and build out now up to 600 watts a square foot, but then we’ve heard what could be coming in the future, one megawatt racks. And so, do you see that near-term on a grander scale? I know it’s happening on a micro scale at this point, I would say, but do you see that happening across the industry as land becomes more expensive, water becomes less available?
Nick Marzorati (25:52):
Yeah. Submersive cooling. So I think it’s definitely becoming a very expansive topic as well. You know, I think that’s the cop out answer I’ll give right now is the best way that we’re planning for that is bringing water as close to the halls as possible as a philosophical way of the heat rejection, right? So, we feel if we do that, then we’re prepared for much. And if we kind of reserve some portion of our land to expansions and cooling plants, then I think that we can kind of get there.
Martin O’Neill (26:24):
Yeah, definitely agree. It’s a very it’s a loaded question, right? Because I mean, everyone’s, everyone’s touching that, that topic in the industry.
Nick Marzorati (26:30):
You start to tow the line with what’s practical again, and now you’re at the speed to market you know, everything is a tradeoff and then you start to tow the line between what speed to market going to look like for something like that, right?
Martin O’Neill (26:40):
I think from what we’ve seen certainly in the last six months or so, is challenges with, again, power generation, right? Backup generation. So, I think part of the future will be generator-less data centers, and as you just said, bringing gas to the site to be able to do that. But I think that’s definitely something that’s going to push the envelope on a lot of developments.
James DiNoia (27:04):
I think we’re going to go through a period of, try this, try that, try this, until we kind of find an avenue between the balance of how much power we can get and how much, how efficient you can still get the PUEs and everything else down to where the most efficient data centers really will be at the cream of the crop.
Martin O’Neill (27:23):
Yeah.
James DiNoia (27:23):
And I think we’ll end it there for today.
Martin O’Neill (27:26):
And then everybody steals that design.
James DiNoia (27:28):
Yes. So Nick, thank you for being part of the podcast. We appreciate all the time you put in for us. Martin, of course, thank you for participating here. I appreciate STOBG’s support here, and obviously, Sentinel Data Centers. So thanks a lot everybody and have a great day.
Martin O’Neill (27:50):
Thank you.
Nick Marzorati (27:50):
Thanks.
Outro (27:54):
Thanks for listening to Building Conversations. For more episodes like this, you can find us on Spotify, Apple Podcasts, Google Podcasts, Audible, and the STO Building Group website.