Data Center Expansion Through Adaptive Reuse | Executive Roundtable

Adaptive reuse began as a plan to convert classic buildings either for their charm, or to boost the economic preservation of historical buildings. Today, adaptive reuse has a much more pragmatic purpose – enhancing our physical and digital infrastructure. As data creation and demand explodes, owners and developers are finding clever ways to adapt existing structures into data centers.

Join data center leaders and industry innovators explore adaptive reuse, what it means for the industry today and its potential for data centers in the future.

Executive Roundtable

Eric Jensen, VP and GM, Data Aire

Craig Deering, Director of Design and Construction, CyrusOne
Mitchell Fonseca, Senior VP, and GM of Data Center Services, Cyxtera Technologies
Michael Silla, Senior VP, Design Construction, Flexential

Eric Jensen: Adaptive reuse is of course not a new topic. It’s prevalent in commercial real estate, as well as residential. It also happens in the data center sector. Some of the ideas around adaptive reuse as it relates specifically to data center spaces are the market drivers. Maybe that’s the geography relative to the end user or the scale of the facility. Maybe it’s the types of facilities that are being considered for adaptive reuse. What makes a good candidate site versus what doesn’t may dictate how we prioritize among the variety of drivers. And then of course, how we execute on adapting a site for reuse as a data center is a consideration.

Let’s start with you Mitch. You mentioned that you were a believer in the future scarcity of viable data center space. So I’m curious, what kinds of data center space do you expect to become scarce, and then what options do you foresee builders considering to resolve that?

What Buildings Are Good Candidates for Adaptive Reuse?

Mitchell Fonseca: As data centers have become larger, more prevalent within different markets, the best space becomes more scarce, particularly in urban areas where we see a lot of legacy telco hotels. To build additional data centers within that space becomes complicated. Finding the right facility that has the right power and the right connectivity is a challenge. So, adaptive reuse becomes more critical as you look at the potential data center footprint, or pieces of land, buildings, warehouses, and office buildings that could potentially be retrofitted. You need to become flexible with what your designs look like. It’s definitely something that’s top of mind whenever we go out to look at properties. Reuse is more prevalent in urban areas. And, and from my perspective, a lot of the warehouse types of spaces tend to be more flexible when it comes to data center retrofits. Office buildings are obviously a lot more challenging. When it comes down to it, one of the biggest challenges we run into is clear heights. So you need the right building that has the right clear heights to be able to support your data center environment.

Eric Jensen: That’s one side of serving the end user. But I think Craig, you and CyrusOne maybe come at it from a little bit of a different angle as far as what targeted end user you’re trying to serve. What is the availability looking like for the future of data center spaces and how we might need to adapt existing spaces into data centers?

Craig Deering: At CyrusOne, I have converted two warehouse spaces based on end user requirements, and both of those were mainly driven by speed to market. In that, I could avoid a site plan application and I was able to move very quickly to building permits. But going back in time it was actually quite common (particularly based on scale) to adapt office space. Going back to some of my early client base, like Above Net and some of the academic, medical and high performance computing environments that have been successively converted to data centers…it’s almost universally been office space. I see more of that happening in the marketplace as people start talking about workloads and workloads needing to move to the edge, primarily to manage the end user experience and also the communication costs. I see it as a growing market and people looking towards the past to see how some of those problems — like clear height, generator emissions and other things were solved before we started building these massive hyperscale centers as greenfields.

Why is Manufacturing Speed-to-Market an Important Driver?

Eric Jensen: So Mike, I’m wondering…coming at it from your experience, what are you seeing as the driver for specific targeted audiences who need data center space? Whether it be enterprise or co-location or wholesale, what are you seeing as sweet spots and how do those market dynamics drive the need for reuse of existing space that isn’t currently data center?

Michael Silla: I think speed to market is a big driver. The end user is challenged with capacity planning, right? Technology is growing at such a fast pace and that’s the reason why they’re coming to us and the third party markets to help solve their problems. I think Mitch brought up available properties, scarcity of available properties. Where there’s available power today, plus what the utility could scale to. Going back 10, 15 years, or a little longer, everybody left New York and came to New Jersey. PSE&G was right there to build at capacity for a lot of the data centers that entered the region.
And we watched that follow suit around the country. Power companies know that we’re not going to take all 36 or 72 megawatts on day one. We’re all in the same neighborhoods and they’re rationing out the power and they’re building as they go. It’s a business model, too and the reality is we’re not going to find greenfields. So we have to take a look at adaptive reuse of brownfields. And again, we’re looking to find a viable facility with floor load, ceiling heights — your height slab to slab; what’s the minimum requirement that you can make work. We tend to focus on what is our design of record. And how do we adapt that? Whether it’s greenfield, brownfield, multi-story, or campus environments, we need to build that flexibility into the process, having a standard that we can adapt as we go out.

We all struggle when the broker sits across from us and says we’re going to start on your property search, describe what you need. We have to try and give them some parameters to bring us viable solutions without filling the funnel so big. We look at warehouses and distribution centers to be adapted. We did a couple of chip fab facilities that were able to convert into a data center relatively easily. It’s going to be faster to pull permits on an existing facility that’s there today. Again, speed to market is important.

Greenfields vs Brownfields – Data Center Development Strategies

Craig Deering: When I’m building greenfield, I have to cover a spectrum of expectations and requirements. If someone can really focus their expectations and have a good understanding of what their IT kit is going to look like, they can take advantage of some well-located properties. To the point when we’re doing greenfields now, we’re looking at a range of densities to cover a range of customers. There’s still a good amount of workload in that 100 kilowatt to 150 watt per square foot density. And that leads to floor loadings, where you can consider some office properties — with the right strategy. So having a strong cleaver about what your core technical requirements are opens up a lot of possibilities in the adaptive reuse market.

Michael Silla: The modularization of our MEP infrastructure makes a big difference as you look at a brownfield and the surrounding property – looking at where you can place that on the outer perimeter of the building. That opens up opportunity to us, when you start to bring that type of infrastructure. Building plants inside of the building complicates it a little more as well.

Craig Deering: If you have a flex office building – a one story with a 40 by 40 structural grid and 16 foot clear, that’s the kind of criteria where you need a strong cleaver, and then you can focus on the relationship between location and proximity to your end users, your fiber resources, and your power resources. Then you can make a pretty quick decision to go forward. At that point the property is fully entitled and you’re just working with building permits, which certainly gets you down into the desired six to seven month delivery time frame.

What Are the Challenges of Converting Office Spaces to Data Centers?

Eric Jensen: So there’s a couple of really important considerations there. There’s the modularization side of things as it relates to design and selection of site, but then there’s also the type of facility. You’re talking right now about availability of office as a prospect, and right now there’s speculation in the real estate world that there’s going to be an increase in the availability of office space as a result of the [COVID-era] environment that we’re living in. And so I’m curious for the panel, maybe we can get everybody to weigh in on office space as a consideration. What are the pros and cons? I think we’ve touched on some of them, but if you feel as though that either connectivity or power availability makes an existing office space, what seems to be the best available option for you right now? How, then, do you overcome some of those concerns you had when originally thinking, ‘well, I’d much rather have a warehouse or an existing manufacturing facility?’

Mitchell Fonseca: I think office spaces are definitely environments that are doable. We see some of the biggest data centers in the world that are converted office spaces. I think office spaces are a lot more challenging because they normally don’t have the clear heights required to be able to cool environments to the levels that we need. To Craig’s point earlier, if you’re looking at more of a retail environment, that’s going to be usually a lower density from a lot of the hyperscalers’ retail environments, which tend to fit better within that office space environment. It’s probably the model for more of an enterprise type of solution. If you have an enterprise that’s going to build out their own data center I think that it’s more doable within that realm. Once you start getting into the high performance computing, whether it’s high frequency trading — the new environments. And some of that stuff is not really going to work well within an office environment. So that’s where you start getting into pretty heavy modifications to the structure of the facility where it probably wouldn’t make sense. It’s really about the use case. What workloads are you targeting for that environment that are going to dictate whether an office building will work or not?

Eric Jensen: That makes a lot of sense, but it sounds to me like potentially the targeted audience makes a difference for whether office space is truly viable for you. With Flexential in particular, Mike, you’re probably sitting in bit of a unique position in that you’re starting to look at both sides of the middle of the road, as far as size of facility and targeted audience. What would tip the scales to prevent you from pulling the trigger on an existing office space?

Michael Silla: The look, the feel, the densities that we’re seeing today make office buildings more difficult to convert. In a previous role I looked at an office building in a dense area, not a lot of great opportunities. It just wasn’t going to work for a data center that focused on the mix of clients at the time.
The other challenge is that we need the slab heights for the design of record. We need that minimum height, and the reality is, I’ve heard somebody mentioned 13 feet, but that’s tough. We’re looking for a lot more height than that. Looking to the future in a second tier market where folks are trying to get close to the Edge and you’re in urban areas, it’s going to get tougher to find a property. So, depending upon business case and future, you would try and make that work. But right now…where we’re at today, that would be a tough one, trying to make the design of record work, trying to keep it standard.

But as time goes on and the data center market continues to grow, and properties become scarce…the edge comes about. So I think that’s the way we need to look at it.

You can’t rule anything out. And this is what I tell my team, if something doesn’t work today, okay, we park it over in the box here, but you never know when you’re going to go back in and revisit that. So you may wind up back to the future.

Eric Jensen: So really what you’re doing is you’re prioritizing the site over the facility itself, of course. You’re basing decisions on the primary drivers, geography, connectivity, power availability — things of that nature.

Michael Silla: If it’s the right price, it’s worth tearing down.

Modularity as a Component of Data Center Design

Eric Jensen: And so I’m curious, you had mentioned earlier that modularity is a component of the design. Can you touch on how you incorporate modularity and its place in adaptive reuse?

Michael Silla: When you look at modularity, it’s a term that’s widely used, and we’ve seen everything from fully modularized data centers to servers in a box. We like to modularize our components as much as possible. Consider a data center kind of as a product, like a vehicle coming down the assembly line. You need wheels, a steering wheel, a radio. And they’re all made in factories elsewhere and then shipped and bolted on. So if you think about our electrical and mechanical infrastructure as we have designs of record, and we have standard blocks of infrastructure, you can prepackage that equipment in a factory. Then it’s shipped as needed to a location.

And for all intents and purposes, it’s bolted on or assembled to the box being the data center. So your facility, your building is your data center and your infrastructure sits on the outside, or is skidded to the interior of the data center, but sits on the outer perimeter. Longevity is looking at life cycles of data centers; we’ve been through multiple generations, where the rush was to get product to market. You build it a certain way, and we’re finding that as you go back to do upgrades on those facilities, it’s a little tougher and an invasive open heart surgery. Whereas our cooling units 750 KW, you can remove it. If you need a 1200 KW unit, then you replace it with that, the same thing with your infrastructure.

It’s easier to adapt the facility long-term when your infrastructure is sitting on the perimeter of the facility versus trying to do open heart surgery inside. Engineering DesignAnd, you know, we think about that when we’re in design today. And when you’re approaching your concept, your hear everything from ‘Hey, my operator’s going to go in there every morning. And how does he park his car, walk into the building and clear security, go to work? Sales bring prospects, walks them through the facility. Eventually those prospects become clients. How do they go in and function? How do our equipment vendors come and do maintenance on the equipment? How does the fuel truck deliver fuel to the site?’ We’ve put a lot more thought over the past couple of years on the future of these facilities, because some of the facilities that we’ve built are limited to maybe retail because of the characteristics of the envelope and the ceiling heights and the floor heights. But as we’ve moved to the more dynamic data centers at the higher densities that we operate today, we have people still operating it at low to medium density and extreme high density within the same environment. And so we have to put a lot more thought into that as we design the modularized components.

Eric Jensen: So future-proofing, of course, is the panacea. You’ve got to be able to see the future in order to do that. It’s certainly no easy task. But I think modularity has a place, whether you’re thinking about a containerized solution, or a power or mechanical skid of some kind. Craig, I’m wondering, is there also potential to use modularity to go vertical?

Craig Deering: Of course there is; we’re doing it. We are doing it with our designs in Europe, but those are greenfields. I have looked at parking garages in urban locations and solved that problem. I’ve also looked at three to five story suburban office buildings and in order to go vertical, we’ve gotten away from talking about modules. We talk about provisioning and provisioning towards end user density. When we laid out my last project, which was a warehouse conversion, it was on a provisioning range depending on whether it went enterprise or hyperscale — because you’re talking about a hundred watt per square foot swing, or even more between the two users. So on that site, depending on how the building sells and gets provisioned, I can go up to about 24 megawatts and 250 watts per square foot.

Based on our topology, I think we’re somewhere around 16 megawatts in 150 watts per square foot. Plug and play flexibility is key. We look at a building like a glider kit, and if you know what a glider kit is, you know it’s a car you buy that comes with no power train, and you put in whatever power train you want, and that’s sort of the concept that you can use with adaptive reuse. And it’s also great on scaling-in a user. We have a lot of high volume users, but they do still ramp in. One of the advantages we’ve had in doing adaptive reuse is that we can get an end user in very quickly, at very low cost for that initial deployment. And this is the advantage of not having a site plan to file in that time frame. Through a series of incremental building permits in adaptive reuse, we stage in all the capacity.

A very effective prototype when we’re doing adaptive reuse, whether it’s retail warehouse office or a single story building…if I have the right setback so that I could develop the yard space to get all of the chillers, we use air cooled chillers and all of the generators on ground. It’s a very effective delivery method. Because I have this space around the building, as Mike says, just to stack the capacity on an as-needed basis.

Is Designing Data Centers to Support 5G Latency Important?

Eric Jensen: Do any of you have smaller urban data centers in design right now to support evolving 5G latency issues? Or are you already starting to build them?

Craig Deering: I’m not aware of anything we’re doing in response to 5G latency. We do have urban data centers but some of those are legacy facilities, but nothing in my region that I can speak to.

Mitchell Fonseca: We have a number of urban data centers, but we’re not really building or currently planning for an Edge use case that’s more specific to 5G.

Eric Jensen: Typically, by what factor does power demand increase when you convert an asset such as an office building to a data center, 5X, 10X or more?

Mitchell Fonseca: It’s usually significantly more than that. When we’ve had to convert buildings, we’re normally stripping out the entire power infrastructure and transformers and everything, and kind of rebuilding those from scratch. I don’t know if anybody else has a different experience, but it’s normally significantly more than 10X.

Craig Deering: So typically, if you’re picking up an office block, you’re going to be three to five megawatts on a service and you’re going up 20, 30, 40, 50, 60, depending on scale. The smallest facility I have is probably a 12 megawatt facility and we up the service from 2MVA to about 19MVA on day one to build that.

I would say it’s easily on a small facility 10Xs and if it’s a good property and you’re using the existing facility as a core, you could be looking at 20, 30Xs — depending on your ultimate development plan.

What is the Impact of Existing Power Supply or Infrastructure on Data Center Conversion?

Eric Jensen: How much does existing power supply or the building structure impact the possibility of the conversion to a data center? For example, let’s take a life science building that has a larger power supply versus an office building; does that matter much? Or if you need to be in a specific geography, are you just going to build it out?

Michael Silla: The big key is working with the power company to see what they can actually get you. Because at the end of the day, we’re selling power, right? If we’re doing our jobs well, we’re going to have excess space, but we’re going to run out of power first and that’s the game that we’re in.

Craig Deering: Yes, as far as that goes, even a question we ask when we’re looking at an existing facility is what’s at the street? So that’s a question to the power company and then, what’s at the nearest substation because we’re typically looking five to 10 years, or ramping into an ultimate power load at five to seven years. We want to get at least 20 to 30 megawatts on day one in order to build out the first section, and then you’re looking at 60 to 100 by ultimate. And most end users are now comfortable with sourcing from one substation. It’s the rare customer that is asking us for diverse substation feeds. Data centers don’t need to be the fortresses they used to be 25, 30 years ago, because the resiliency is now in the network and the information. That’s how i’s managed; it’s not in the facility.

Michael Silla: It’s rare that we have an RFI or client looking for that and when you start asking, well, why are you looking for this? They always point to the uptime and you say, well, that’s even been relaxed if you actually read it. It’s a matter of having that conversation. It’s just probably something that’s been on an RFP that’s been floating around for two decades.

Eric Jensen: I think the question centered around life sciences as an example. I think you have experience in converting chip fab, Mike. Is that right?

Michael Silla: Yes. Life Sciences or other industrial spaces are definitely viable options indeed and there is power at the street with those facilities. But then again, when you start looking at facilities of that size, we’re going to want a 36 megawatt or larger future capacity in there.

Craig Deering: Let me just add one thing though, because it’s important that everybody understands. If you’re adapting a building, I don’t care how big the power services, you’re not keeping the switchboard. You’re not keeping any of that source material. Because it just doesn’t work for a data center use; it’s not set up correctly. There’s never going to be a position where you’re adapting an existing incoming service. You’re going to originate it back out at the street and you’re going to be looking for a property that has a substation that has a double ended connection to the transmission system. And those are the key things to look for.

How is Airflow Management is Key to Operating Data Centers?

Eric Jensen: As an example of power challenges…Data Aire saw a lot around the One Wilshire project we did in Los Angeles. Power utilization was the primary driver for modernizing and centralizing the whole power and cooling infrastructure there. Also partly because it had just kind of slowly evolved over the decades which is part of what you are going to wind up finding in any kind of legacy type of office space, if it’s of any size and of any age. For anyone reading, adaptive reuse was described as open heart surgery. That’s not a mischaracterization, but there’s also a lot of people who are alive today because of open heart surgery. We’ve seen plenty of folks who need to go into office or warehouse space that is strategically located on the smaller scales. So medium to medium/small types of facilities that have substantial considerations for airflow management. Airflow management becomes the number one thing that you have to be thinking about, which is really what the gentlemen here have been talking about with the importance of those clear-heights. Some additional thoughts…if you’re not going to get the clear heights, then it’s critical that you really pay attention to how you are managing your airflow. Are you going to get the delivery of cold air where you need it to go? And how are you routing everything, whether it’s piping or layout of the infrastructure relative to the ITE, etc.?

Michael Silla: And to add on…number one, make sure your operations team is in the room during design and CFD analysis — and not the perfect environment, but the type of environment that you’ll actually be operating in. Instead of focusing on a perfect environment, we all operate in imperfect environments. Doing that during the conceptual stage will help you because airflow management is the key to operating the data center. We will always be managing that airflow and that’s the key to success; very important facet for our operators.

Eric Jensen: I think it’s a mindset — thinking about it as part of a pre-commissioning activity. Run through a couple of what-if scenarios, what-if I didn’t have that? Or what-if that occurred?

Craig Deering: There’s a user called Power Loft and they had an interesting concept which I’ve looked at adapting for office use where all of the air handling equipment was on one floor and the data center was on the other floor, and they supplied everything from below. And so, if you’re dealing with like low heights, you know if you start getting, and if you’re going to stay with an air base system, right, you can get very creative, particularly with the site source airflow designs, which a lot of people use now, you know, the fan wall design. If you start looking at using space creatively to move large bodies of air, but of course, if you are going with more of a liquid approach, you know, either this was just posted about the barge data centers, but you can certainly use the cooling concept in an office where you really just make the decision that you’re going to do rear door heat exchangers, you know, universally or in rural cooling.

And that opens up a lot of adaptive reuse opportunities. And you can actually mix those with space cooling which is actually something I used to do a lot 35 years ago in order to address, you know, kind of a comprehensive cooling scenario. So find yourself a creative engineer and be laser focused on what your operating parameters are going to be and I think you can go out there and find a lot of buildings in good locations at good prices that can work for your need.

Mitchell Fonseca: I would add the biggest challenge you mentioned there is at what point does it not become economically feasible? So we have amazing engineers in the data center world and some of the stuff that we can do is pretty mind blowing. The challenge is how much are you willing to spend to make that specific building usable? It’s always doable, right? It’s just, how much did it cost you? So that’s where there’s always a fine line between, is it economic or is it doable? And I wouldn’t say that in a lot of case, when you start talking about clear heights is where you start getting into having really unique cooling to make it work. It’s just, is it really worth it?  So there might be a specific reason why you have to stay in that building and that structure. And again, it is doable.