Episode Overview :
The moment its construction begins, a new data center becomes an old data center, given how fast technology and use-cases are evolving. Retrofitting an existing data center is the only way to save it from obsolescence. But retrofits are more difficult than you might imagine, a panel of experts tell Cool Vector.
This episode of Cool Vector includes Bo Bond, Vice Chair of Cushman & Wakefield, Brian Jabeck, VP Business Development at Bennett & Pless, Phillip Koblence of NYI, Critical Ventures, UIA and Nomad Futurist, and Nabeel Mahmood an industry Top 10 Influencer, CXO and co-founder of Nomad Futurist.
Among the key takeaways of the conversation:
• Data center retrofits will remain a moving target as technology advances. The types of workloads data centers must support are evolving too quickly for any one design to remain optimal for long. “The beauty is that technology and innovation, which drives our industry, will always cause change," says Cushman & Wakefield's Bond. "Can we take something that was built before, improve it so it serves something today or for the future, or does it cost more and save us more on time to be able to come out of the ground new? And I think that cycle’s always gonna turn."
• The data-center retrofit queue is growing. Rising rack densities and global power shortages mean more operators are upgrading existing facilities to meet AI and high-performance computing needs. "There is a significant amount of data center and compute space that's available in the market space that needs to be retrofitted to meet the existing compute demands," says Mahmood.
• Even new data centers can become obsolete almost immediately. Technological shifts, such as liquid-to-chip cooling, are forcing operators to modify facilities that are only a year or two old. The decision between retrofitting and rebuilding often comes down to rapid tech obsolescence—not the building’s age. "You've got one- to two-year-old buildings that are drilling holes and making penetrations to run a bunch of liquid lines that they didn't plan on," says Jabeck.
Follow Cool Vector on LinkedIn: http://www.linkedin.com/company/search-party-channel/
About Cool Vector's editorial advisors:
Phillip Koblence is a strategic executive and thought leader in the data center and interconnection space. He co-founded NYI in 1996 and has successfully evolved the company from a Web design and hosting provider to a facilitator of robust digital infrastructure and connectivity solutions with executional capabilities in key national and international markets. He also serves as CEO of Critical Ventures, and Managing Director at United Integrity Advisors, agencies that provide multi-disciplinary consulting services to a broad range of real estate and digital infrastructure firms. Phillip is Co-Founder of the non-profit Nomad Futurist Foundation and Podcast, designed to demystify the world of critical infrastructure and inspire younger generations to join the industry.
Nabeel Mahmood is a globally recognized futurist, technology executive, and board member guiding innovation across the intersecting worlds of AI, quantum computing, data infrastructure and automation. With decades of experience shaping digital transformation strategies, Nabeel serves on multiple boards of publicly traded and privately held companies, where he influences decision-making at the highest levels. As the co-founder of the Nomad Futurist Foundation, a 501(c)(3) nonprofit, he’s leading an international movement to democratize access to education and careers in digital infrastructure, particularly for underserved and underrepresented communities. A top 10 global influencer, Nabeel is known for his bold perspectives, thought leadership, and ability to connect complex technology trends to their real-world human and environmental impact. He delivers keynotes around the world that challenge industry norms and push for a more equitable, inclusive, and sustainable future.
Transcript:
David Snow: Hello and welcome to Cool Vector. I'm David Snow, your host, and today we're joined by Bo Bond, vice Chair of Cushman and Wakefield. Brian Beck, VP of Business Development at Bennett and Plus, and I'm joined by two editorial advisors to Cool Vector. And the Cool Vector editorial advisors, Philip Koblin from NYI, critical Ventures, UIA and Nomad Futurist. And also from Nomad Futurist Nabil Mahmud.
We're gonna talk about an interesting trend in the data center world, which is to retrofit data centers. The headlines are largely going to these gigantic greenfield projects. $20 billion a clip starting with flat dirt, you know, building out to a brand new data center.
And yet there are, old data centers out there that, looking for a second life, or at least some of their investors want them to have a second life. Why don't we start with Bo, from Cushman Wakefield. Why would someone retrofit a data center? What is the need and what does the landscape look like?
Bo Bond: Well, first and foremost, the reason you would retrofit a data center is some of the componentry within the data center starts to, reach its end of useful life, right?And that starts usually with the mechanical, the electrical systems that run the facility. Today's applications that are running on these facilities, are greatly, greatly different. Than when the building was probably conceived. Those buildings that are being contemplated through Retrofit at the earliest would've been delivered in the late nineties and you know, early aughts.
So in comparison to other real estate asset classes they could have been built in the early 19 hundreds or what have you. These buildings are 25, 30 years old, usually at the oldest. So it really starts with updating the things that run and power and provide the cooling, extract the heat.
That's really where you're starting the retrofitting today.
Brian Jabeck: I think we're gonna see a fair amount of people just keep running 'em into the ground.
You can't say stagnant the way you operated 20 years ago. I mean, two to three years ago we're talking, you know, 250 to 300. Pounds a square foot from a floor loading perspective. And, and now we're, we're, we're pushing 450 pounds a square foot, two to three years later.
So to think of what was that, 25 years ago? we're just talking a different level of facility and a different level of operation. And if you're not growing and evolving you, your business starts to stagnate. So I think as Bo mentioned, there's, there's an appetite for seeing how can you leverage existing footprints and existing facilities That's gonna give you some, some better performance for your business,
David Snow: First of all, what does old mean? Does old mean it was built in the eighties? Does it mean it was built pre-chat GPT.
Phillip Koblence: I think old in, data center parlance means turn of the century, right? So the data center of the kind of, uh, first generation of internet, 98, 99, 2000, vintage for the most part.
There are some that are older than that. in many cases, those are in buildings that were built, you know, significantly. Further back the vast majority of kind of carrier hotels exist in places that had high floor loads. So you think printing press buildings and stuff like that.
Part of it is what the use case of that data center is, right? If you are in, a kind of tier one NFL sitting, you're running a, a carrier hotel that is not necessarily, trying to attract. The AI workloads of the present and focused more on interconnection, and network density.
You might not necessarily need, to retrofit that for the purposes of being able to support, you know, the a hundred kilowatt, 600 kilowatt rack. densities
Phillip Koblence: Old data center sites make good new data center sites
The underlying infrastructure that powers. Data centers is essentially, you know, copper touching, copper glass, touching glass, and, cooling, environments, these air conditioned rooms.
A lot of that underlying infrastructure is still relevant today. And the hardest part about it is finding access to, you know, the power and the space and trying to. To get it to market most quickly. Many times you'll find that the existence of a data center at some point in, in the life of a piece of real estate, makes it a really good place to start, to develop a future data center and retrofitting is the way to kind of get that done, more quickly.
David Snow: Nabeel Mahmood, from your perspective, what does the landscape look like out there as far as the population of data centers at different stages of life? What does the population of the older data centers look like? And what decisions are people making about how to extend the life or, end the life of some of these, uh, places?
Nabeel Mahmood: I get the most interesting question. So let's start back with the, the definition of old. Anything that's been delivered as of today is old. Uh, I mean, computing, uh, per kilowatt rack is actually significantly changed.
It's not that long ago that we were actually designing, building data centers roughly about one, one and a half kilowatts of rack density to now we're talking 1.2, 1.5 megawatt of rack density.
So what does that tell you? The load density per square foot has increased significantly.
So there is a significant amount of data center and compute space that's available in the market space that needs to be retrofitted to meet the existing compute demands. But then it comes with structural issues. Uh, the other challenge that we are facing in the industry is the power constraint.
We don't have enough power globally, so we've got to go back into the existing infrastructures and environments to retrofit and upgrade to meet the existing compute demand. So, a relatively large number, is in the queue right now to be retrofitted. Jerry Seinfeld has
Phillip Koblence: Jerry Seinfeld has a bit that he does in his, standup routine when you buy something at a store.
It's basically in various stages of throwing it out, right? When do you eventually get rid of this thing that you purchase? And the same thing is true with data centers. The, the answer is probably a hundred percent.
It's just a matter of, you know, is it being retrofitted today or is it gonna be retrofitted tomorrow or the next day, or the next month or the next year? So I think building data centers today is really about introducing modularity so that you can, retrofit if the right that's even the right word, without disrupting, uh, the existing workloads and understand that the evolution of digital infrastructure is about the flexibility and the modularity. Of trying to keep pace, not trying to build a data center today that is going to answer all the questions 10 years from now. 'cause that's not the way this technology works anymore. It’s evolving at such a rapid pace that you need to like, essentially.
Go in eyes wide open that as an appeal, says the, the second you put it on the lot, the second you drive the car off the lot, it's a used car. The second you put it into development, it's already, a legacy facility.
Nabeel Mahmood: Yeah. And, and just to add to what Phil is saying, right? So there is that carbon implication as well.
The existing sites are not carbon ready. We are at that place in point that carbon accounting is dead. We really have to do something and that means that you have to upgrade the equipment that's there to begin with. You have gotta deploy process and policies and technology to be able to account for that carbon impact that it's got.
And you've gotta actually get your efficiency ratios to, to the level where they will. Help us scale up to the levels that we need to be. one of the other points that I would wanna, uh, get into is the cost of new builds. is not where it needs to be. It sometimes is actually cheaper, easier, and better to retrofit versus go out and do a greenfield.
Brian Jabeck: I mean, there's year old facilities that. need to have liquid to chip added to the facility they didn't have it, you know, on the plan a couple years ago when they started and now it needs to be thrown in. So you've got, one to 2-year-old buildings that are drilling holes and making penetrations to run a bunch of liquid lines that they didn't plan on.
And you've got that same scenario. Maybe you had a, a three building campus set that, you know, you've got your first building done, but you've seen direct entities jump way up. You may just say, Hey, I've got my power approved for. 250 megawatts with the new densities. I can do that in two buildings. Let me just rip down the first building I already built, and I'll build two more dense buildings and start over.
A lot of it comes down to geography. You know, what are your permits? What power do you have? And then, you know, what's your appetite for? Doing work on a, on a live facility versus, tearing it down and starting over again.
And I think we're gonna see clients for a variety of reasons, pick and choose both those scenarios. And, and there may not be a right or a wrong answer,
Nabeel Mahmood: A part of the driver, Brian, just to add to what you're saying is the,
Capital expense and the operating expense discussion. Mm-hmm. Right. It's, it's really monetarily some of these decisions are really monetarily driven and availability from a human capital to power, power availability. I mean, where do we have enough power right now to put a gigawatt data center anywhere in America?
Tell me. I can can't a few places. There's probably a, I might have a few, some,
Brian Jabeck: There's some small, you know, there's somewhere in rural America. Is is Abu Dhabi in
Phillip Koblence: America? Yeah. No, they can't. They
Brian Jabeck: gotta be in. Yeah.
Bo Bond: If you're talking about real estate asset classes, I mean, are houses the same? is an office building.
In Embarcadero looking over into the Pacific Ocean in San Francisco, the same value as a one story office building in Des Moines that a call center is running in, does a million square foot distribution center. It's 40 foot clear, the same thing as a 40,000-foot, 12 foot clear, in town, last mile distribution center. So if you apply that logic to data center, there's a multiple of the variations of what we've seen in a very small timeframe of the life of the data center. If you would contend that the first computer centers might have been built in the seventies. In the eighties on corporate campuses with three inch raised floors with filled with tape decks, right?
And a lot of very substantial network that was driven by copper, not even glass at the point. And where we've come through the nineties to where we are today from bringing those computer centers out into facilities because the MEP was outstripped in those office buildings, bringing them to a purpose-built facility, bringing out really those.
Carrier hotels that were mentioned earlier, those were for connectivity purposes only to connect the world through deregulation of long distance through connecting the world through copper to glass, connecting the world through mobile devices.
Bo Bond: So we've changed so much, and the applications that resided in these different variations of data centers is dramatic. The first enterprise builds for the banks, the insurance companies, the healthcare companies, the retailers. They went out and built their own data centers. There were two megawatt data centers. So there will be retrofits that will fit certain applications.
We will improve some of the infrastructure that supports them. But you can't just classify even the AI data center today. There is inference data centers, there are training data centers. So the world will continue to evolve. Data centers, they will continue to be retrofit,
We're looking at, well, did we just retrofit the second reactor of Three Mile Island we did for a data center. Did we just take some of the cold fired plants that are being pulled offline for all the right reasons? And are we retrofitting them to serve significant capacity scale data center campuses? Absolutely. Because we have the transmission coming in, the generation exists, so Retrofit will serve a lot of different masters as we look forward, so it's always great to go down one of these paths, but as soon as you do, the beauty of technology and innovation, which drives our industry, will always cause change, which will always have a step back and say, is this a CapEx? Is this an OPEX decisions? Is this a speed to market? Is this, I gotta be in front of technology?
All those levers that get moved drive this change, which speak back to is, hey, can we take something that was built before? Improve it so it serves something today or for the future, or does it cost more and save us more on time to be able to come outta the ground new? And I think that cycle's always gonna turn.
David Snow: Maybe a question, for Brian, since we're talking about sort of the, the economics, what are the main costs of retrofitting. And then how do those compare with the new build? Because you know, ultimately, people have to decide is it less expensive to just knock this thing down and build something new?
And, you know, to Phil's point, maybe there is already some power infrastructure there that makes it a good site for data center, but the, perhaps the actual building itself is not suitable.
Brian Jabeck: One of our first things for older facilities, we find that drawings were never kept up and the facilities been living for 25 years with a bunch of mechanical equipment moving all over the place, and no one's updated any drawings.
So a lot of the first step is where is everything? Is it where it should be? Is it where your drawings think it is? You're looking at the envelope and integrity to make sure you're cooling what needs to be cool and keeping out what needs to be kept out, like water.
And then I, the powers, you know, a huge discussion. What sort of equipment, and I think, you know, a big thing, especially for older facilities is the height and the space that you've got between columns. And do you have the height in the ceilings? 'cause if we're gonna be running cabling. And if we're hanging liquid from the ceilings now to run to, to chips, do you have all of that?
So I think just a base overall evaluation on all of those, you know, upfront to say, is it worth. the investment to upgrade this facility? Can we do it? You know, can it be done structurally? Do we have the power, do we have the space to fit all the equipment that we'll need to add in now?
The great thing is 3D Modeling's come a long way. So, so you can, if you've got all the right information, you can 3D model out of facility and, and you can get yourself a very good picture on what needs to change and then what are, what are the cost implications of making those retrofits?
We’re seeing a ton of, people looking at retrofits than they're actually executing retrofits today. But I think, I think we're gonna see that the retrofit momentum continue to grow. and the analysis obviously continue for a number of different sites going forward.
Nabeel Mahmood: Well, besides just the technical and the, the financial element, there's also the environmental element to consider as well, right? I mean, some of these buildings are either in the right zone or region or not, we don't know.
Absolutely.
David Snow: What kind of a property or a building, if you could sort of caricature or, or make a generalization would just be an automatic no for a retrofit. You take a look at it, there's some element to it that is just, a deal killer
Bo Bond: A data center footprint that might have been built into a multi-tenant office building back in the nineties that might've been dot coms of 5,000 feet and a hundred thousand foot building.
We get these calls a lot. You know, can I retrofit this into a data center? Technically you could, but the value that's in there is traditionally zero because the amount of power that's brought in, it's probably was built on, old switching technologies so AC versus DC power.
It just doesn't have what the future's going to need based upon the applications that are gonna be running in there, the servers. So those are a perfect example where if somebody wants to make it more than it is, and it's probably best suited just to go back to being office space because 95% of the building is office and this little 5,000 foot computer room that was built out in someone's space, those are usually the quickest ones to tear it out, move it on. It just doesn't have a lot of value. I think there are a lot of different opportunities, and I'll let the team talk about it here, but that's the first one that comes to my mind where we say zero to value, scrap it, move on. The building was originally built to be an office building that was just a suite that had a computer room, rip it out.
Phillip Koblence: It could have a use. As a kind of meet me area for that building, but really focused on the kind of network element of it as opposed to compute. There is no one size fits all.
There is no one thing. Um, there is no one type of data center. Like is it gonna be a data center or not? there is value in centralizing connectivity, uh, for distribution in an office.
But certainly not to house millions of dollars of, you know, H one hundreds or H three hundreds or whatever. It’s just not where you would put that, you know, kind of millions of dollars of computing, gear.
If you are talking about structural integrity and you're talking about ceiling heights and all those things, if there's even a question about whether those. Would fit, a compute requirement of today, then it's clearly not an effective, long term, piece of digital infrastructure, particularly for compute workloads.
Brian Jabeck: If it's less than a hundred, 150 pounds a square foot, like. Just, you know, move on to the next one. It's not a data center.
With the advent of enterprise AI coming on, I feel like the enterprise dedicated data center is coming back from a, I don't want all my AI applications living in the cloud anymore with all the other AI to have access to what I'm building as a proprietary tool for my business.
So I do see a swing back. And I'm curious to this group, do you see that, you know, do you see the enterprise dedicated facility for, you know, a bank builds their own AI technology and we want to have that now in our own dedicated, six megawatt facility. Is that swinging back around again? Potentially?
Phillip Koblence: History does tend to repeat itself, right? So a lot of the, the AI wave has similarities to say the cloud wave of migrating all the applications to the cloud initially, and then repatriating a lot of those workloads
So there's no doubt that AI is following. A similar type of evolution where, you know, you start with it, playing with it in. A public cloud environment an AWS or an Oracle or whatever, you are going to likely find a hybridized way.
Where you control as much of your infrastructure as makes sense, as is practical, and then interconnect it from an inferencing perspective or maybe a large language model perspective,
But, um, I think repatriation of AI workloads will follow a very similar kind of path.
Bo Bond: Yeah. And those, those could end up in two places. And I think we've started to see it. Some of the enterprise buildings that were built, first generation, highly redundant, two m plus two data centers. big razor wires, setbacks, you know, 'cause tanks, were gonna run down the street and go through 'em.
Airplane, concrete roofs, airplanes were gonna fly out. I mean, the things we used to worry about, rail lines got, now they're full of fiber and you want it next to it, right? Those will probably be repatriated [00:19:00] and reconfigured. We're stripping out that redundancy and utilizing some of that power because you don't need that redundancy with some of these AI workloads.
I think you'll start to see some of those early enterprise builds, reworked. MEP changed as long as we put the floor loads correct and we can extract the heat. Those will be properly placed in the hands of AI for small workloads. And then I secondly, look, we've seen it when we saw some of these speculative.
Bitcoin, facilities built and regions. When we saw stuff come in from out of the country and hit the United States, and then we saw some companies roll into bankruptcy. Those things have been repatriated and they are housing a ton of AI workloads right now. And they took an asset that was was mining at that point.
Now it is serving the AI industry We all read that one in the paper, right? Mm-hmm. So that is a perfect example of those were assets that performed a data center function. But vastly different. And we are going to see that.
We just don't know how that future's gonna be.
Phillip Koblence: Yeah. Just for anyone listening, uh, we're referring to the, uh, core we’ve, acquisition of core scientific, both of which were born out of, cryptocurrency mining. If there was ever, an example of an industry I.
That, you know, retrofit, repurpose, whatever you want to call it, that is the telltale sign of an industry that went all in on one, specific type of thing. And then had a change in use case and a change in fortunes, the likes of which, I mean, look at Nvidia stock really over the course of the last, you don't even have to go 10 years, five years, um, to understand that, there was not a clamoring for NVIDIA chips.
Until this use case, was born and maybe, Jensen had this, amazing master plan. Or maybe like most successes, it was just the luck, um, is right.
Bo Bond: 4 trillion, 4
David Snow: Trillion bucks. Four, $4 trillion worth of luck.
Bo Bond: That's exactly right.
David Snow: We’ve talked about, properties. For various use cases that just are not gonna work. Are there any kinda holy grail types of properties out there where you come across it and someone say, I'm thinking of putting a state-of-the-art data center here, and it's perfect.
You're like, hot damn, let's do this. I asked, because I was speaking to a data center company recently and they said they built a brand new facility in a bottling plant. Bo have you come across examples of that where it's just.
Fortuitous that you come across some sort of facility that just is gonna be perfect.
Bo Bond: In today's world, it all starts with the power from day one. It has to have scalable power that you can get with that utility over a time period. I've sold, newspaper plants, that are now data centers.
See Chicago, I've sold chip plants that are now data centers. See Texas, I've seen old grocery stores that have turned into Bitcoin. but, you know, to the point that was brought up earlier, it's gotta have floor loads. It's gotta have ceiling heights.
You gotta be able extract the heat, you gotta have the power. power is everything. And the people that say, I'll go anywhere as long as you can get me power. That's not the case anymore because the collateralization of the infrastructure via the utility that you have to pay for upfront through your LOC through your kayak is definitely separating the speculators from the data center developers or operators or AI and or the hyperscale users, which is there's.
It's a short list of really the people taking and using the customers using this space. It, it really starts and stops with the availability to be able to get the power, to deliver the power and have the capacities in the financial wherewithal to be able to pay for that because the utilities are not funding that at all today at all.
Brian Jabeck: Power's a key and it probably leans to a bunch of heavy industrials. Heavy manufacturing where, where you've got substantial power brought to a facility
But if you've got, older facilities like that, that are structurally they were built 'cause they've got heavy equipment on 'em, they've got, you know. Reinforce concrete. They've got steel structures, the bones to support the weights and they've got the power, you know, provided to the site.
I think that's, that's what you're looking for. Even some mine sites are built out in that way. It's um, I got invited to a conversation here next month about using mine sites for some of this, and mines have a ton of power brought in. So there's, there's opportunities in those locations and I think it is, to Bo's point, it's all about, I think, power first.
And then see if you've got a building that supports it. If you found the power, then you can either retrofit a building or, you can rip one down and start over again.
Phillip Koblence: that’s The point. it used to be that people looked for buildings that might be, good candidates for a data center retrofit, and now they're just looking at where the power is.
So if you look at a lot of these, uh, data center deployments, these major ones, they're happening either adjacent to or right on top of former power generation, facilities you know, former smelting plants and, and all of these things that were really designed to, to create power. Secondarily, they'll wanna make sure that there are rights of way for fiber and connectivity and all those things.
People are no longer even assuming that a building has. Anywhere near the level of power that we're looking for coming to it. If you looked at a, you know, a printing press building that has a couple megawatts that might have excited you in, you know, 2003, it is not exciting you today.
So, you know, you have these campuses that are being contracted, the plan is 400 megawatts of power. That’s insanity.
David Snow: Does this mean that industrial sites and former industrial sites are being viewed as a place to put a, data center?
Bo Bond: I think we've always looked at industrial sites, flex sites, and industrial over the course of kind of speculative builds as well as, you know, enterprise driven builds.
You have industrial developers, publicly traded, industrial developers that are opening up significant data center development portions of their companies, globally that are delivering these facilities. I think they are looking more at their controlled pieces of land that sit in a heavy, heavily powered area. People are building gigawatt campuses and there are utilities that do not have that in their service territory.
Their service territory is less than what some of these campuses are. So just 'cause you see a bunch of high. Lines running through your ranch or your neighborhood, or your industrial park. There's such a qualitative portion that needs to go through. And these studies now with these utilities are taking 6, 12, 18 months.
The deposits that go forward are significant. The collateral you must place to be able to get through the studies of the utilities. It's separating. The speculator from somebody that really has the investment. So I mean, we have six or seven states right now that have passed this summer, or in the legislature trying to pass certain either regulations or opportunities within their states to be able to control or manage how the utility.
Allows for this utility growth because it needs to have some form of conformity. 'cause right now it's a wild, wild west. There's thousands of requests and they can't necessarily understand what is true demand versus speculative demand. Or duplicative demand, three people chasing the same project on the same site.
Well, that's counting three times in the stack of utility. So we have gone a long ways in the last 36 months, and I think we're coming out of it with state and utility regulation to find out really what we're going to be. But I think on the scale developments from the large operators to the large hyperscalers, it's less about the facility and more about the.
The power available to be put on a purpose-built land campus.
David Snow: Give me a example of a project, a retrofit project that goes poorly you build it and then you realize. Far too soon that you built the wrong thing or it cost too much. What, what would be some disaster examples that you could, uh, share with us without naming names of any particular project necessarily?
Phillip Koblence: Yeah, for sure. Save that for me. Save that for me.
Brian Jabeck: You, you say it. I'm gonna try to name the project. It’s a mixture of you don't have the space, you don't have the structure required for the weights, you don't have the power.
There’s a ton of conversations now around cooling and what's required for the racks and to the chip and not just the liquid of the chip, but then the air required for everything else. I would think if you don't have, you know, a facility that breeze well enough to cool the, you know, you server room, you're running at a, you know, a thermal runaway situation that would cause a massive problem for your business and the, those are probably, I think, some of the major, the major issues, structural power, you know.
Making sure you've got the room to cool.
I guess the added piece of, if it's a live site, you gotta make sure you're not disrupting the operation that's going on as well, is a big piece of that picture. So, those are our key areas that we focus on when looking through retrofits.
Nabeel Mahmood: At the core of it, you have to understand what the, the demand is and how are you gonna load the infrastructure, how are you gonna cool the infrastructure?
I mean, there's been plenty of deployments out there where we have pushed the boundaries and end up with a total cluster. with electrical faults or, you know mm-hmm. Hotspots within the infrastructure and or resulting into fires by not, uh, addressing the electrical code. So regulatory compliance gaps, uh, could also shut down the projects.
David Snow: Philip, uh, Koblin, you've never seen a disaster in your career. So if you could imagine, that's the only one.
Phillip Koblence: I think. I think it, the only one that has his middle name is
Nabeel Mahmood: Disaster. Yeah,
Phillip Koblence: Exactly. Um, the Curb Your Enthusiasm theme music follows me wherever I go.
The people that build data centers for a living understand how to build them. So it's not the same level of disasters of like the initial data centers that were built in garages and, you know, the wind, kicked up or, or, or the rain came in and, and the internet was, completely disconnected.
At some point it's the speed with which we are growing as an industry, that will put added pressure on, people to deliver, on timelines in a way that doesn't necessarily get them to dot every I and cross every T.
As these facilities come online and you start seeing not. The first couple megawatts go in, or even the first 10 or 15. But once you start getting up there and start, you know, really pushing that infrastructure to, the limits that it was designed to, you're gonna see some of these, disasters in, you know, the UK a fire started at power plants that, that took down significant portions of, Heathrow Airport you saw.
Uh, huge disaster in Spain and Portugal with, you know, the power grid being pushed to its limits. There's also the people up this chain that are delivering, that are trying to get, uh, to these timelines so that they can start delivering these things so they can start, you know, reaching their KPIs, which might be across purposes with.
Making sure that you're doing everything the right way. But no, I've never seen a disaster before in my life. I'm sure this is all just theoretical.
Bo Bond: There's so much wild speculation in this industry.
I do believe there is a sense of dotcom that will come into play where there is capital being raised and we're talking significant barriers to entry where there's capital being raised to wildly speculate on development, whether it's just. Horizontal development of land and power and or speculatively building these facilities based upon this insatiable demand that's out there,
It's a very small group of people that are actually consuming these large loads, and so therefore, I think the disaster could be that there are going some people that are playing in the private equity world or world, or investing in some of these new developments, new technologies, all based on the comm.
Build it and they'll come. We've heard that before. We heard it in the nineties. We heard it in the S. We might see some people giving back, you know, some facilities to the bank, land holdings to the bank, I think we'll see some potential disasters as it pertains to that
There’s going to be some train wrecks with some people that think they can get into this market that has a high barrier to entry, not only on the investment, but the knowledge base to be able to build it and run it and operate it successfully.
David Snow: When there is a distressed data center opportunity, I definitely want all of you to come back and we're gonna talk about it. Uh, but sounds like that'll happen at some point. We just don't know when and who's gonna be on which side of the, uh, of the negotiating table.
Well, thank you so much to Bo Bond from Cushman and Wakefield.
Brian Beck from Bennett and plus. And the Cool Vector editorial advisors, Philip Koblin from NYI, critical Ventures, UIA and Nomad Futurist. And also from Nomad Futurist Nabil Mahmud. Thanks everybody, and I definitely want you all back on a Cool Vector program in the future.