Commercial Real Estate morning News Show
Leaders in the commercial real estate industry discuss the latest news and provide insights that you will get nowhere else!
Commercial Real Estate morning News Show
Commercial Real Estate Morning News Show - Navigating the Data Center Revolution
The episode delves into the complex and rapidly evolving world of data centers, highlighting their significance in the current digital ecosystem and their interaction with technologies like AI. We explore topics such as energy demands, the need for legislative support, and the importance of location and sustainability for the future of data centers.
• The role of data centers in storing and processing vast amounts of information
• Importance of co-location models for businesses
• Rising power demands due to AI and quantum computing
• Innovations like modular designs and liquid cooling systems
• The push for sustainable energy solutions in powering data centers
• Legislative developments affecting data center growth
• Trends toward building in nontraditional markets for data centers
• Future implications for co-location versus in-house data center operations
Okay, everybody, we are live, All right. Welcome everybody to the Commercial Real Estate News Show. This is the Data Center Edition, and I've got these cast of characters on my show that are all talking about walk-up music and all these other sorts of great things, and we want Howie to go open one of those cases, don't we, Howie? We want to go open one of those cases, alright. So we're going to talk data centers today, aren't we excited? Yay.
Speaker 2:Yeah, let's do it. What's a data?
Speaker 1:center. I don't know. Isn't it just like computers and stuff? I don't know?
Speaker 2:Just kidding. Who needs them?
Speaker 1:Alright, I'm going to let my folks introduce themselves. Howie, you're talking already, so go ahead sir hey, howard, berry, avis and Young.
Speaker 2:I am a data center real estate guru focused on all things data center real estate worldwide. You name it, you name it. I'm based in California, silicon Valley. I am not in LA, thank goodness yes sorry for the people down there.
Speaker 1:It's horrible exactly all right Bill hey everybody, I'm Bill Picard.
Speaker 3:Prime Data Centers been in doing data centers for 25 years. I'm the vice president of our alternate channel, so I work with Howie and other brokers and agents across the US and internationally to drive business to prime. We focus on building massive scale hyperscale data center campuses worldwide.
Speaker 1:Awesome, very cool, and Jay.
Speaker 4:Hi, my name is Jay Hussey. I'm with H5 Data Centers. We are a wholesale co-location provider based in the US, and we have lots of space and power available in different areas. I'm looking forward to talking to everyone today.
Speaker 1:Awesome, welcome, welcome, welcome. All right, we'll kind of get into it and I like to kick it off with a crazy story. So who's crazier than trump? All right, so trump has a plan for data centers and it looks like a small plan. Howie, it's like 20 billion. Like, really, is that enough for? I mean, I was like 200 billion. Maybe that might be enough for trump, but 20 billion, huh, did you?
Speaker 2:read that what he's doing. When a hundred megawatt data center costs a billion dollars to build, it's in the grand scheme of things it's. It's a. It's a small piece of what's happening in the market, but you know where there's. There's gigawatt deployment of data centers right now in in multiple areas of Dallas and in other markets, but yeah it's it's helpful it's helpful for sure, I don't know, I don't, even I didn't.
Speaker 1:I kind of glanced over that article. I don't, I'm not really sure, but anybody, I guess who's supportive of the industry is always helpful, right, because we don't we don't face that in our regions do Do we Bill? Do we face it?
Speaker 3:So first I'm going to to be fair to everyone open kimono. I'm not a fan of our president-elect, so understand that in my comments. At the end of the day, how he's being nice, $20 billion is basically a rounding error in our industry of what we need and where we're going, and he he made it very clear in his comments that he knows absolutely nothing of what he was to which he was speaking. We also have to take a look at the gentleman who is providing the money. He is the second richest man in the Emirate and he has been attached to to Trump since 2016, when he was at a new year's Eve parties at Mar-a-Lago because Trump was building hotels in his country. So this is just more of a president-elect trying to exploit the presidency for his business than doing anything to really help our country or our business. I will say nothing more on that level. I'm sorry.
Speaker 2:No political views are accepted in this show.
Speaker 3:I'm tapping out on politics.
Speaker 1:I started out on politics. He's kind of an interesting. I'm just surprised that he went. I just you know, Noel, I don't understand why I should be surprised by anything that Batman does. It's really kind of an interesting deal, I don't know. But.
Speaker 2:You know I think it's interesting, kind of an interesting, uh, interesting deal, I don't know. But you know, I think there's. It's interesting, amy, because there's so much money coming into the data center market that companies I've never even heard of I've just on a website of a group that popped up that they go we're data center developers. I'm like what have you developed? And I? They don't own a data center or have they ever developed, and I see that multiple times a day. Or I get calls from groups that want to invest. We do support and we're educating investors on what is a data center and we're finding that really difficult to explain how data centers operate, how the market works. It's so different than traditional real estate.
Speaker 1:Yeah, for sure. So, jay, when you're, when you're for your company, you're focused on leasing out the rack space, right? Is that where?
Speaker 4:you're at. Yeah, we're a wholesale COLA provider and we do lease out space, rack space and buildings, whatever someone's looking for?
Speaker 1:Do you own the buildings that you're in or do you do both?
Speaker 4:Do you like operate some and own some, or we have a kind of a mix of owning and operating, but we're mostly an owner operator of our sites. In some cases we have some sites that are owned by others or that are operated by us and owned by others, but we are we are primarily an owner operator.
Speaker 1:OK, all right, very cool. So let's go through, because we have someone asking the question of can we explain the data center business? So, howie's right, no one knows what it is. So let's go through. There's like edge, there's co-location, there's hyperscale. So Bill's focused on hyperscale, you're focused on co-location, and they have all these different terms that a lot of people aren't familiar with. So I'm not going to say y'all have been in the industry longer than I have, so someone, take that one, take that I'm lobbying that one out there.
Speaker 2:Well, I'll start. You know everything's dependent on power and the amount of power. So what do you guys rate a retail tenant, anything you know, 250, 250 kw and under or lower. Yeah, yeah, 250 kw and under. Anything above that, up to four megawatts is uh wholesale. And above that, above four megawatts, would be hyperscale okay, so so explain to me.
Speaker 1:so for what does that get you? Four megawatts? How much? How much computing am I doing? Give me a typical tenant size. Is it like an insurance company? Is four megawatts? Is it? Is it a like car dealership? Can you help there? Someone can explain that to me. It's a good question, amy.
Speaker 3:So if you have to quantify it in one of those ways that you threw out, let's say it's an insurance company, at the end of the day we are in co-location operating other businesses' compute platforms, as my kids call me. I'm a real estate agent for the internet and because somewhere all of these servers and systems have to operate to produce all the data and produce the output whether it's websites or quotes or whatever it is that they're doing and connect it to the internet. So hyperscale is just large scale co-location. So Jay's company does lots of different sizes of co-location from more retail cages of clients that have different sizes maybe five, 10 cabinets in a cage to maybe hyperscale clients as well. So but at the end of the day, for the majority of us in the industry it's all co -location and it simply means that the client is co-locating their gear in my warehouse. But my warehouse is highly secure, has redundant cooling and is designed to ensure that their product, whatever it is they're putting out to the world, doesn't go down.
Speaker 1:So, oh, I totally get it now. So their servers are in your building.
Speaker 2:Correct.
Speaker 1:And so they come and service their servers. You don't have to do anything with that. You just make sure that their servers are secure and safe and they get access to them Right.
Speaker 3:We might have techs that can do what was very often called smart hand services, which are which really is very limited Maybe power cycling a server for someone that's that needs to reset it, something like that. You go down the stack to two other data center providers where they have managed services, and there's plenty of great companies that do that as well. But for us, larger providers, co-location is yeah, the customer runs their site.
Speaker 1:Self-storage for computers.
Speaker 2:Okay Now do you guys offer any cloud platforms?
Speaker 3:At Prime we do not and most of the large hyperscaler and large wholesale platforms will not offer our own cloud platforms. But what we will do especially if we're going down stack and doing some retail and smaller enterprise deals, we'll bring on Megaport and Packet Fabric and other companies that create these on-ramps to allow our customers who are doing a hybrid deployment between cloud and Colo.
Speaker 4:And those are the value add services that add stickiness to our business. When those customers come in, they look at us and they say, hey, they have all of these different items that we can do and all these different services that we can work with. They have partners in the industry that make our product more valuable, and then that's how they choose who they're going to work with and then that's how they choose who they're going to work with.
Speaker 1:Okay, so let's talk about how data centers are kind of transforming because of AI, right, and there's more power demand because of AI. I think somebody equated, is it? I don't know what it was like. You do a Google search, it requires this much power. If you do like a chat, gpt search, it requires 20X. I don't know, can I?
Speaker 3:actually give you that answer. So my good friend, bill klyman gave this stat at a trade show and I love it. I share it all the time. If we go online right now and do a google search, we type in whatever you want and we want to type in you know, I want to research, research howardberrycom, because who doesn't? And you hit enter and it comes back with all your Howard Berry images. That transaction used enough power to power an incandescent light bulb over my head for 11 seconds.
Speaker 1:Okay.
Speaker 3:Everything's going to AI. Even our Google searches are AI enabled now. Every AI query that we do uses 800 to 1,000 times that much power. Yeah.
Speaker 4:Oh, interesting, Okay, and you know to tie that into one of the articles that you gave us. The personal AI computer that's three grand is going to have that kind of capability.
Speaker 1:Yeah, right, exactly what he's talking about is. Jensen Wang, basically at the CES show, showed the little new AI PC that NVIDIA is just coming with and he can hold it in his hand and all those great things. So does that mean that if everybody gets an AI PC, then the power load is now at their house, their electric bills are going to go up, basically, if they're using those AI PCs. Right, basically, if they're using those AI PCs?
Speaker 4:right, If NVIDIA is smart, their power supply does not exceed what is in computers today, Right oh?
Speaker 1:okay.
Speaker 4:So the search itself is going to go out to other computers and run those computers to get that search information. That's where your power is being used.
Speaker 1:Oh, okay, so that makes sense, because everybody was saying, well then you don't need a data center anymore because I got a little data center in my house. But I don't think. I think that's not the case no, no, no.
Speaker 2:You need all the data. All the data is stored in llm large language model, where it's stored in a silo that just it goes and queries it from that, that silo interesting, okay, so.
Speaker 1:So basically, nvidia just blew up the data center industry even more. Because you know all those tech geek first adopters are going to go get their AI PCs as soon as they can. You may have a pretty cool price point. I think there was like what, $560? One was $580. It's pretty cheap. It's really kind of interesting, very cool pretty cheap.
Speaker 2:Uh, it's really. It's really kind of interesting. Very cool, yeah, with ai and and and. The amount of data we use, the amount of everywhere we go is recorded. Every camera, everything's being recorded. Our whole lives are being recorded. That that, all that data is constantly being stored, so the amount of data we have now doubles every year yeah.
Speaker 1:Do you think it'll eventually triple? I mean like I can't. I I think if you get somebody with a, somebody is starting to use an AIPC and understanding I just think there's more ways. What AI is going to allow us to do is capture more data, right. I mean, like, if you come back to real estate, the things that we do, we capture, building data, like the stuff that people don't capture today because it was a pain in the butt to do, you can do fairly easily with AI. And so now that you're doing that, now you've got more data you need stored and all these other things, and so I think that's what AI allows us to do, right. I mean, heaven to Beth, my mother had pictures for 10 years on her iPad Like I just can't imagine what everybody else is doing and she saved them all. I was like really trippy when I went and saw her iPad after she passed away.
Speaker 4:I was like gosh that's like crazy town.
Speaker 1:That was like when my kid was born, my oldest who's 28 or whatever. All sorts of crazy pictures. But can you imagine that multiplied by 300 million people in the us? Like how much. And that's just pictures and videos. All other things, right uh?
Speaker 4:yeah, all that stuff.
Speaker 1:Yep, yeah, yeah, no, it's. That is super crazy. All right, let's talk about the next. Let's move on to the next story where all these uh, there was three stories that came out. They were kind of they came out today or yesterday or within the past couple of days, so just depending on the area of town you're in, right? So Michigan, basically, is extending tax breaks for data centers. Then you've got one that's going to tax them more. What was the other one? I'm looking for it. And then you have one that basically says no, it's in Pittsburgh. Is it Pittsburgh?
Speaker 4:Pittsburgh, oh, pennsylvania, yeah, pennsylvania.
Speaker 3:It's a county in. Virginia, I believe.
Speaker 1:It's a county in Virginia somewhere. This magical county is probably where Dracula is housed Pittsburgh, Pennsylvania.
Speaker 2:They deserve the money.
Speaker 1:But they just said that no good data centers. So it's like all different people are embracing the data center trend. What are y'all saying in your neck of the woods? What are y'all saying as far as the legislation stuff happening?
Speaker 3:Well, I'm in northern Virginia and so we're kind of at the epicenter of data centers really for the US and arguably the world of. At the epicenter of data centers really for the US and arguably the world. Data centers were the bell of the ball for Loudoun County in 06 and 07 when the county ran into a financial crisis and the gentleman who ran economic development, his good friend Buddy Reiser, realized hey, let's capitalize and grow this data center community because of the tax revenue it can bring in. And it saved the county and I can tell you my property taxes are 30 and 40 cents on the dollar lower than anybody else's of my friends in other counties.
Speaker 3:But different areas have different motivations and different areas either have run out of patience. You know we have a lot of data centers and as much as it frustrates me that many of my neighbors don't get it and don't truly value what we bring to the area, I do get that. Everybody has their tipping point. I can't speak to some of these other counties but at the end of the day, loudoun County sees upwards over a billion dollars annually in tax revenue. The numbers of jobs that are created by data centers is astounding People say oh no, it doesn't create a lot of data center jobs.
Speaker 3:There's nobody in those buildings. Well, one, yes there are. And two, there's plenty of jobs being created that come and go from the data center. Or are you know, around the data center, from restaurants to you name the retail. So it's a shame that the benefits that the data center brings to the economy are sometimes lost.
Speaker 1:Yeah, for sure.
Speaker 4:And Bill, beyond that, the trades that are affected. If you're not pulling in those types of businesses, those trades that build and that plum and that electrify things, they're not working and, like you said, the county was dying. This is from people to say, hey, we want to buy land that's powered in your state so that we can bring this to one of our partners or customers and say, hey, here's a place that you can build. So you're seeing that in Montana, which didn't have a lot of data center activity at all, in Virginia, tons of data activity and folks that live there are kind of saying, hey, you know, we don't want to live next to a data center. So you're seeing that kind of move its way up the chain to the folks that manage and legislate the county and it really just depends Michigan. You know they're saying, hey, let's get more people, let's, you know, let's get more people, let's make it easier for data centers to expand here. That, to me, is a good move.
Speaker 1:Yeah, I mean from a tax revenue perspective at least right. But we were talking about the other day. Data centers operate as almost emergency operations. At some point they can stay up right when everybody else goes down and like tornadoes and whatever it might be. So it's good to have a couple of data centers around right, because then if you need to have a war room because you just got lambasted by whatever a hurricane or whatever at least you've got something that's running for at least a little while. You would think.
Speaker 4:I don't know. Yeah, there's a good portion of companies that just go into a co-location for disaster recovery purposes only and they just they just deploy some core servers with core operating on them and that's what they'll run if there's a an issue. So the gamut of customers is is very, very wide that we have to work with.
Speaker 1:Yeah, so back to the AI point. What is it? What is it? Is it changing the data center? Like are you? Are you going to have to rebuild data centers? Are you going to have to expand your centers? Are you going to have to bring on-site power generation to them? What is, what is the big changes?
Speaker 4:because ai is coming all of the above is that I was gonna say all the above. Yeah right, say the same bill, yep all of the above.
Speaker 1:So can you do that? Can you expand on your existing footprints as they are today? I mean, did most data centers are on plenty of land, I would assume, so you could just build on some you can.
Speaker 2:you know, a majority of data centers in the us are all which is surprising to people we consider like one megawatt, maybe, maybe two megawatts, and it's been like that for years and now we're building we're only building new data centers that start at a hundred megawatts. So it's the the, the demand for power from AI, because the GPUs that drive AI have a higher power. Demand is increasing every year. So we still are going to need all those one to two megawatt data centers for edge deployment. Edge is what is the connectivity to speed up your connectivity to the self-driving cars, to your taxis that have no drivers in them. You know we need data centers deployed everywhere for connectivity and low latency.
Speaker 3:Interesting. You know, amy. Yes, what's happening right now is such a change in power density. Our industry is strapped for space, for capacity. If you look at Northern Virginia, the vacancy rate is sub 1%. I mean that's insane. So we're building and pre-leasing data centers that will come online in 26 and 27 and writing contracts for them now and planning for them now. That is actually. That's good and bad, and here's what I mean by that.
Speaker 3:It's challenging that we're so low on inventory. However, one of the articles you shared with us earlier talked about you know, can you really change or upgrade an old building and retrofit buildings and or data centers and expand them? You can, but it's usually very costly and you usually find a gazillion problems you never imagined going into it. So it then becomes even more costly. Some data centers can easily have power and cooling added to them if they're a modular design, and I could bore you with what a modular design is, but I won't. But the good news is is now that we're catching up on inventory. In my opinion, now we're able to actually build for the future. We can build for the present and the future, because right after AI, we've got quantum computing coming.
Speaker 3:And that's going to make AI look like child's play power wise from what I'm reading. Ai look like child's play power-wise from what I'm reading. Yeah, I know. So, even though it's a challenge that we're low on inventory, we're able to start building in liquid-to-chip cooling and rear-door heat exchangers and eventually emergent cooling will really take hold. It's still got a ways to go, but while're, while we're playing catch up, we're now building data centers for what we need for five, 10 and 20 years down the road.
Speaker 1:Yeah, Do you think? Do you think you can build a data center today that would suffice in 20 years? Or, you know, you think it's going to change in three. I mean, you're hoping it lasts 20 years, don't you?
Speaker 4:I mean, I mean hindsight, sorry Bill, hindsight tells us no, hindsight, I mean hindsight, sorry Bill, hindsight tells us no, hindsight tells us, no, you can't build that, but you know. But, as Bill alluded to, modular building is really the way to make that happen. You plan for growth and you build your building in such a way that you have your stub outs ready for power cooling, all those different things, and just keep building and building and building as you go. H5 is big on that. We do a lot of buying buildings that we then standardize to H5 standards and we find success in that.
Speaker 4:So there are ways to do it and Bill made a very good point. It is very expensive, know.
Speaker 1:we feel like there's a lot of of opportunity out there to take what's there and recycle what's there yeah, so, uh, talking about quantum computing, ibm has made a big bet on quantum and they have uh, now you can, as a quantum startup, you can use their quantum cloud, um, and you pay basically a hundred dollars a minute to use it and a hundred dollars a minute, but the processing power that it has, you can do whatever you're trying to get done in milliseconds.
Speaker 1:Like so it's. It's just super, super cool. We're starting to, you know, tinker with it now, but it's super, super cool. But in order for IBM to let you play with the big dogs, you have to have like a PhD. You know quantum computing, you know somebody on your staff, because they're not just going to let little me get up there and go. I want to learn.
Speaker 2:So you don't even need a minute. Yeah, you don't even need a minute. I want to leave, so you don't even need a minute.
Speaker 1:Yeah, you don't even need a minute.
Speaker 1:Basically, the minute has become the hour, like at the end of the day, that's what we just did right, and so having to spend an hour on now becomes a minute or less, which is just super, super cool. I mean, I think it's going to be interesting. So that'll you know, if people start more and more people start figuring out how to use quantum computing, leaf is just computing, you know. I think IBM took a big leap over AI and just went okay, now we've got to get a quantum right, yeah, but who knows, maybe it'll blow up in their face. I don't know. Ibm's pretty smart. Yeah, there's a couple. There's just a couple.
Speaker 4:A few smart guys over there, A few smart people over there. I should say Very cool.
Speaker 1:So, Jay, what part of the world are you in?
Speaker 4:I live in Oakland. I live in Oakland, california, north of all the problems down south, so I'm good here.
Speaker 2:Down, howie right down, by san jose yeah, yeah, I'm with you, I'm near you, yeah yeah, he's, we're not far.
Speaker 4:We're about 40 minutes away every now and again we have lunch together, things like that, yeah all right, interesting.
Speaker 1:So what do you think? Where do you think the next, besides virginia, I mean, since they don't want you there anymore, bill? So where's the next big hub of data centers?
Speaker 2:Where isn't it? It data centers are going everywhere they can find power. You know, I just looked up a statistic right now co location, which both Jay and Bill do is that the market is about 16 gigawatts. There's the vacancy in the co-location markets 1.9%, giving us like 304 megawatts available in the market. There's 600 megawatts under construction and 29 gigawatts planned for development and I guarantee you probably that 29 gigawatts is already taken by somebody.
Speaker 1:Oh, you mean they're already spoken for.
Speaker 2:Yeah, there's 16 now running. Yeah, go ahead, Jay.
Speaker 4:I was going to say if it's not already spoken for. People already have it on their radar and they're starting to talk about it.
Speaker 2:Wow, that's just co-location and double that capacity is cloud. Cloud is very similar footprint, which is google, am aws, uh, you know, microsoft. They're the cloud providers and they're a similar size footprint yeah, it's it.
Speaker 3:Yeah, that's just crazy yeah you know, what happened in northern virginia is never going to happen again anywhere else. It's a very special story and we could do a whole show on it sometime. If you really want me to pour the hell out of you, I can do that all hour, I promise. But but what we saw in the formative years at Equinix, when I was there early on and afterwards, is we saw this push of data center cities to what we call the NFL cities. So you had your Chicago's, you had your New York, new Jersey area, you had Nova, you had Dallas. Later on, you know Northern California, but then came Phoenix, then came Atlanta and some other larger markets.
Speaker 3:The reason for those big cities and these big epicenters, if you will, of data center traffic was all about latency.
Speaker 3:Because when you're dealing with cloud-based and cloud-enabled technologies, when you're looking at what we're doing right now voice and video over the web early on it was hard to do near the peering points, which are very often where Equinix is to be able to be as close as possible to the dark fiber networks so that you can get your content in the lowest latency time possible to the end consumer, which is the eyeball networks and those of us who are downloading Netflix or whatever we're doing.
Speaker 3:Now that we're transitioning to a lot of AI content and a lot of AI work, traffic on the web, latency isn't a big issue anymore. Latency issues for AI are actually in the development clusters, in these data centers, where we have to have all these packed up GPUs typed together while they're doing their learning. That's where they can't have latency, but in taking the finished product out to the world, they don't have the same latency issues. So now we're seeing Reno and Vegas and you name it, utah, salt Lake City, places that aren't big peering points anymore becoming maybe those next hubs that you're asking about, because we can and we can go to places that have more space and open community to our existence there and power.
Speaker 2:Dakotas, the Dakotas, yeah. Dakotas, dakotas, yeah.
Speaker 4:To the Montana article. Right, montana is going to be another place. I mean anywhere, like I stated before, anywhere that there ample uh land and ample power. And even if there is an ample power, get to work with the local utilities and have them build power for you. That's where these, that's where you're going to see these, these, these things rise up well, there is, there is.
Speaker 2:There is a tendency, though we see that that, um, staying away from natural disasters, I, I can't get any hyperscalers to look in Houston or in Florida no, they're like, no, we're not going there or in Kansas, for that matter, tornado Alley. Tornado Alley is a big stay away, and that's one of the reasons Virginia became number one market too, is it's it? There's low, uh, natural disasters, of course.
Speaker 3:I mean, just moved to town but ironically though you know, data center alley is adjacent to dulles airport and planes almost touch all the buildings every day, right?
Speaker 2:yeah, yeah, that's crazy?
Speaker 1:um, so does. So there's a big cluster of data centers in Northern Sweden and they they did that because basically they don't have to cool the damn thing, because it's like frozen near the Arctic circle, and they got there like I can't remember who was it. There was an article saying that the PUE for those data centers in Northern Sweden is like 1.05, which basically means that what they're using to power the building is minimal compared to what they're using to power the servers. Like I'm trying to explain, but you're powering the server, you can build back to your tenant, but you can't build back the power they're using to power the building, right, and so that's usually your right relation and all these other different things.
Speaker 1:So so I'm trying to explain PUE to people but, but yeah, so it's an amazing thing when you like that. So people were talking about doing them in the oceans and doing it in all these these weirdo places. But if you do onsite power generation, if you go there, then you're not having to stick it up at the Arctic circle, right Anymore. Cause that, or is it still an issue? Is PUE going to be an issue still?
Speaker 4:even if you're doing on-site power generation. You think I mean, even if you're doing on, yes, there's always going to be a PUE. I mean I'm thinking about this as we speak, but yeah, I mean there's got to be, there's not going to, there won't, wouldn't be a parity. I don't think on that. No, you're going to have a PUE regardless, and PUE encompasses also cooling. So part of the reason they're in the Arctic Circle or close to the Arctic Circle is because they're using outdoor cooled air to keep that expense as low as possible.
Speaker 3:So that's really driving that PUE lower Right for sure. But you know, amy, there's so much we have to consider in all of this. You know, whether it's Arctic Circle or wherever, you also have to have a trained data center workforce. The men and women that run these data centers. You know we seek to hire retiring Navy nukes because they know how to run mission critical facilities. You know the person who runs an apartment building or an office building is not someone that is qualified for a data center business or a data center job.
Speaker 3:So you have to have that, but you also have to have connectivity. I can't speak to what's going on near the Arctic Circle, but I don't think that they're ever going to have the dearth of fiber capacity and options in networks that you find oh I don't know in Northern Virginia or Chicago. So there's a lot of pushes and pulls in that kind of an equation.
Speaker 1:Yeah for sure. Yeah, that's, that's super interesting. I mean it's, it's this whole push just to sustainability. I don't think it's going to go away. I mean, obviously there's a reason why nobody wants a data center around in their backyard because they're afraid all the power is going to get sucked away. But you know, I don't know how do you get, how does the industry kind of move the needle, or the PR needle, that they're sucking all the power away? How does the industry do that from a when they're building? If you're building a new one, say, I mean, texas, we don't care, come on, we love you here. They're building a 786 acre campus outside of Dallas, right, we love you in Texas.
Speaker 2:Yeah, amy, it's got to be self-generation on site and we're waiting for small modular nuclear to be generating, you know, on site with. It makes sense for the data center to generate the power and then feed back to the grid. The power source should come from where the data center is and we have not. A lot of people know this, but there are 57 nuclear power plants in the United States and of those there's 97 nuclear reactors in the United States already operating and we never talk about that. We're scared about nuclear, expanding nuclear.
Speaker 2:But nuclear has been perfectly safe and most of these facilities were built in the 70s and 80s and they're still operating. So you know, nuclear is safe. Small modular nuclear is going to be safe and it doesn't make any sense to build a nuclear power plant 100 miles away from a data center and then an extension cord all the way over and then a storm comes in and breaks that extension cord. Now you got to run generators to keep the data center running. It's build the power at the data center and that's the way we should be looking at this. But the data center industry doesn't adopt any new technologies very quickly. It takes a long time to change from what we do traditionally to anything new. Because it works, let's not change it. Let's do what we're doing.
Speaker 4:And it's very expensive and it's expensive and most others can't.
Speaker 1:Yeah, that's true. So did you see the article about Deep Vision and how they partnered up with you're going to? I don't know they partnered up, they said they've done a small modular nuclear site at a data center, which is super cool. I just so I guess it's is it, I don't know. They were touting that it was the first one, so hopefully we finally got nukes at a data center. We'll see, we'll see how it goes, but at least we got one pilot. We got one pilot to see how it works. I don't know.
Speaker 3:So here's some, doesn't the?
Speaker 1:Department of Energy have to say don't they have a say in all of this?
Speaker 2:They do they do Absolutely, they approve.
Speaker 4:Yeah, and there are a lot of regulations and laws and rules and all that around running a nuclear plant. So we'll see, we'll see how it goes.
Speaker 3:So you actually got me very excited earlier, amy, when I saw that you wanted to maybe talk about nuclear at all. Fortunately, my next door neighbor, a very good friend, is the Deputy Assistant Secretary for Nuclear Reactors at the DOE and he's transitioning out of that role over to what's called the Idaho National Labs where he's the Director of Nuclear Reactor Development. So I've been teaching him data centers. He's been teaching me nuclear words I don't know what they all mean, but I know a lot of words now but he's also become a leading speaker on nuclear power and the transition into our data center industry. He's been asked to speak regularly at our data center events. His name is Brian Smith and the reality is Jay you nailed it very aware of the importance of energy resiliency for our economic stability as well as our military stability, and it feeds into AI as well.
Speaker 3:So the DOE is leaning in hard right now into supporting this transition, or transformation, if you will, of our power to more, to nuclear and not just for data centers. It's going to be really across the country we are behind, I believe, most of Europe on this, and the SMR technologies that are out there right now are exponentially safer than, as in how we mentioned. We've got so many reactors already. But what they're building now in these SMR Phase 3 reactors, it's amazing. But the PR campaign to not have the NIMBYs losing their minds when all of a sudden hey, I don't want any more data centers in Northern Virginia.
Speaker 3:Oh, but bring them nuclear reactors. That would be great. I don't see it.
Speaker 4:Yeah, amy just another point about Go on it or something Make it pretty.
Speaker 3:Put it over there by my kid's soccer field That'll be fine, there's room.
Speaker 4:Another point about companies that are dealing with nuclear is Microsoft. They're working to restart Three Mile Island. I know why they want the power.
Speaker 1:Yeah, I know, I think the DOE quashed that I mean. So no, it's funny because Microsoft went to go rent one and then soon, jeff Bezos. The millionaires had to like fight amongst each other. Well, I want one too. And so then Jeff Bezos was trying to get one first, he had a first. Oh Bezos had it first.
Speaker 2:They bought a nuclear power plant first.
Speaker 1:Yeah, oh, interesting. Yeah, that is crazy, tom, that's where we have to go.
Speaker 3:You know you asked about on-site generation and all that. That's all very feasible. The challenge we have with it, in my opinion and you guys tell me if you think I'm right or disagree but there's two challenges that I see primarily for onsite generation. One is cost. Simply, if we can use the grid, if we can use the power company for lack of a better word I can't think of the utility right now that is the better cost to pass on to our client. But then the other challenge simply is space on that land. When we take on X number of acres, it's because we want to develop X number of data centers that are revenue producing. If we're foregoing a lot of that space to our own on-site generation, it does change the math on your cost that you now have to pay to your client on the data center.
Speaker 2:I challenge you, bill, I challenge you on the data center. I challenge you, bill, I challenge you. They actually. On-site generation eliminates generators and air quality problems.
Speaker 3:It does if you're using the grid for your backup.
Speaker 2:Oh no, the generation has its own backup. You have redundancy in your generation. So, instead of using generators, now your on-site generation is your redundancy because you have multiple generation sources on site, so you can eliminate all the generators and the cost of development and buying generators. Sorry, generator industry, but that's the way it works. Xai in Memphis, tennessee, in something like three, four months, but he spent $7 billion in three or four months to get his data center up and running 100,000 GPUs and $7 billion and he used onsite generation with gas to get it up and running before he got tied to the grid. Just insane. Don't get in his way, he'll make it happen.
Speaker 1:Was he using like a genbacher generator or what.
Speaker 2:I don't know what he was using, but it was natural gas self-generation on site, something like that oh interesting. Yeah, that's super crazy well I mean that's half of his gpus. He's gonna deploy double that size interest.
Speaker 1:That's crazy time. So I'm assuming he got. He got, uh, eventually memphis power and light or whoever it is, it's going to bring power to him eventually to bring power.
Speaker 2:He's right near a self-generation uh natural gas power generation power plant, so it was pretty just across the street ben they. I think they generate a gigawatt of power there.
Speaker 1:Oh, interesting.
Speaker 2:Yeah.
Speaker 1:Yeah, don't put it past Elon Musk to do anything, because he's probably going to do something. Who knows, who knows? So let's get back to renewable energy. I guess everybody thinks nuclear is renewable, so that's fine, but the only way. So people ask me this all the time about. Well, why can't they use wind farms or why they can't use solar? We're going to go back to data center 101 here for a second. Solar is still a very expensive way to charge a battery, and so there's no possible way you can especially on the data centers that are 100 megawatts or more that you can ever use solar or wind, right, I mean, we're limited in the power generation that we can do. So it's going to have to be natural gas, there's going to have to be small-nodule nuclear, right, that's it. That's all we got.
Speaker 2:Go ahead. Jay.
Speaker 4:Solar generation. Back to Bill's point about the geography or the land you're buying. Solar generation takes a ton of space. So if you're buying land to build data center, the last thing you want to do is put a bunch of solar there, unless you put it on your roofs or things like that. But you know it is. I think it's a developing industry and a developing technology that's going to improve over time. I think it's something that'll eventually be part of the mix there for for large generation of power.
Speaker 1:Yeah. So when you talk about so, when you talk about there was this article about how expensive AWS is and a lot of people are doing bringing back, like getting off of AWS and bringing back to their own servers does that then feed into the co-location industry? Because I'm assuming 100%.
Speaker 2:Repatriation. We love that Repatriation correct.
Speaker 4:Yeah, we love it when they do that Come out and get off the cloud. Get off those bare servers, those bare metal servers. Buy servers, put them in our location and manage them.
Speaker 1:Okay, I see. So that's what's kind of leading the. Because AWS is expensive as hell. I mean it just is, especially for us little startups like me. It's really expensive. I mean it's not something that I can just go do. It's a huge expense for for a startup like me. No joke, I could easily get into like a million dollars in a heartbeat of just hosting If I was using AWS like and like that. And I'm just like you know I can't scale and grow that way. So it's, it's, it's really so. I guess everybody else is recognizing that. I'm assuming it'll force AWS to drop their prices because they've met the top of the line no, I don't know that it will.
Speaker 3:No, Because, if you think about it, the amount of data that we produce and consume is rising exponentially annually.
Speaker 3:And so all of these forces will continue to play an important part. So when I was with a couple other companies that were not as much focused on hyperscale companies, much like H5, that did a lot of commercial and enterprise and all that, we absolutely saw some amount of a movement of clients away from Colo into the cloud. And one of the reasons why we wanted those ramps, those on ramps, was because we wanted to be able to monetize that, that movement, but at the same time also to make it easy for them to come back because, like you said, you know it's expensive. They found out quickly it's really inexpensive and easy to put all my stuff over there, but when I want to take it out, when I want to massage it and work with it, oh no so so the pendulum swings both ways and eventually comes back to the middle, and that's what we're seeing now.
Speaker 3:There's right station, um, but, but. But I, I don't, I, I, I don't know, I don't, I don't follow aws closely enough to say, oh, it's going to change their pricing. No, because, because there's still companies that are still moving to cloud for the first time and still out there to learn the lesson.
Speaker 2:Oh, yeah, yeah. And the difference is they look at it. They don't own or operate their own data center. They don't have the cost of refreshing new servers. All the equipment, the employees, the operation, the cloud looks really desirable, but it's like a gas pump. Anytime you're pumping gas or moving data, you're paying for it.
Speaker 1:Interesting, yeah, so I guess maybe it's just a scale thing. At some point you outgrow it and you can afford to have your own servers. Right, it becomes an economy scale At some point you can just do it yourself, right.
Speaker 2:Put your servers in their data centers yeah.
Speaker 3:And we're going to see the same thing with AI, because right now, this boom of AI, this is all these companies like CoreWeave and Lambda Labs and all these companies that are building these basically cloud-based AI services, and everybody's going to go there to use their AI. But over time, corporations and organizations are going to become AI savvy and start incorporating an AI solution and program into their own infrastructure and move, move, pull back from just a cloud based AI service.
Speaker 1:Interesting. So there was this article about talking about services, this article about how Equinix this is in Ireland. They bought BT's kind of assets and BT was moving away from owning assets. Is that a thing that you might see happen in? Will there be something like that happening in the us, or or is there going to be? I mean, look, I'm a, I don't own data centers, but I'm a former real estate owner. It's, you know, not for everybody, so you know it's like a kind of interesting thing, do you?
Speaker 1:don't think it's going to happen here?
Speaker 2:what, what's that dig? Go ahead, though. For people it'll like, you know, like they's like a kind of interesting thing, you don't think?
Speaker 1:it's going to happen here. What's that Go ahead Bill For people that don't like? You know like they'll be the people who just want to be the data center services people and not actually own the data center.
Speaker 3:So the answer is it's already happened and they refer to it as an asset light strategy. And BT decided hey, we're a telecom company, we want to focus on telecom services, because that's what we've done.
Speaker 3:That's what we were born and raised to do. We got into data centers and we realized you know what? It's not what we do. Verizon did it. They sold off all their data center assets. They gobbled them all up and then they sold them all off. At&t has done it, Zayo most recently got rid of their Zcolo platform, and now DataBank has 99 data centers because they bought them all up. So yes we're seeing this retrenching of focus of what we all do, but it's been happening for some time and I think it makes sense.
Speaker 1:Yeah, yeah. So get people to specialize what they're good at, right? Yeah, I mean it's the same reason why.
Speaker 3:Why, uh, enterprise companies sell off data centers and there's companies like you look at element critical. They they go out and do sale lease back strategy. You know, um, you know, a pharmaceutical company built their own data center. They realize we don't know how to run a data center. We don't want to run it and by the way, we only need about% of this data center that we built, so let's sell it to somebody. We'll lease some of it and everybody will be happy. Yeah, sure.
Speaker 2:We sell those all day.
Speaker 3:Yeah, yeah, that's what keeps Howie off the streets at night.
Speaker 2:Yeah Well, thank you.
Speaker 1:We want to keep Howie it's not.
Speaker 4:I've seen him on the streets at night.
Speaker 2:I've seen it too. It's not pretty. It's not pretty. We all have our likes. Don't get my yum.
Speaker 1:Yeah, for sure. All right, pull up your crystal balls. Pull up your crystal balls. What happens in 2025? What do you see happening this year? Pull it out. Mine's broken, so I'm not going to say anything, but pull it out. So if you had to pull out, put the Swami hat on, think Johnny. If you have had to pull out, put the Swami hat on, think Johnny Carson, put the card on your head. What do you think happens in 2025? This is such a hard question.
Speaker 2:Yeah, it is A lot of the same. It's 24. A lot of the same at 24, a lot of deployment, a lot more buying of power or figuring it out. You know. Hopefully we understand the small modular nuclear and we get some deployment there. And it'll be interesting to see what happens with these large, as Bill mentioned, the GPU deployments. Maybe there's consolidation of the GPUs.
Speaker 1:Do we run out of power? Is there like so much demand that we can't even just fill it? I mean, there is now right, we can't even fill it today, but do we actually hit a breaking point in the power grid, like, do you think that happens?
Speaker 4:No, I think the power grid is going to recognize there's a lot of money to be made doing this and they're going to buy in it's taking time because they're slow to move.
Speaker 4:It takes a lot for them to do anything, but it will catch up. It's got to. You know business won't survive if it doesn't. So that's going to happen. And for 2025, I think we're going to see a lot of the same stuff that happened in 24. We'll see a lot of pre-leasing. I think there's some facilities that will come online to give you know to provide some critical capacity. That will come online to give you know to provide some critical capacity. But I think 25 is still a year where things are ramping up again, are still ramping up to kind of meet the demand that's out there. But you know, end of 25, I think we'll start to see a lot more activity and a lot more capacity coming online. 26 will be like that as well A lot of capacity coming online.
Speaker 3:You know what Howie said a lot more of what we saw in 24, but I'll take it another step, but much bigger. We're already seeing it. Tract is a company that develops powered land. They just announced the startup company that they're going to spin off of Fleet, and fleet is going to build fleet sized data centers, as they're calling them, on the land that Tract is developing. Avio Digital is building massive hyperscale campuses and what we're seeing is these companies want to build these massive campuses for one tenant yes, right tenant yes.
Speaker 1:Right, oh, okay, interesting.
Speaker 3:So that's a change and the reality is whether you're an AI startup, because they've got money they've got NVIDIA and Supermicro and all of the behind the scenes money and then you know, but we still got the AWSs, the Metas, the Azures, all of those guys scaling aggressively as well.
Speaker 2:We're going to see a lot of these super hyperscale campuses going to one, maybe two tenants, but yeah, yeah, and I think there'll be some M&As this year and there's some mergers of some companies, consolidations of operators.
Speaker 1:Yep oh of operators of operators. Yep.
Speaker 4:Operators, you know. On top of that, I think you're going to see a lot of smaller, newer companies try to break into the industry this year. That used to be real estate type of plays, but now they're saying, hey, data center is where the money is in real estate, and I think you're going to see quite a few of those pop up as well. I don't know how successful they'll be Depends on their choice of partnership with an operator or whatever, but at the end of the day, I think we'll see quite a few of those start popping up.
Speaker 1:How well has Prologis dipped its toe in data centers? How well has it dipped its toe? Has it started? I mean, does anyone know? Can we comment on prologis?
Speaker 3:no, yeah, I don't know a lot about them. I have some colleagues that are involved, so I hope they're doing really well.
Speaker 1:Uh, yeah it's kind of interesting, uh, you know they're involved.
Speaker 2:They have data center, real estate, um, they've had it for years. It's part, you know, industrial portfolio and they're expanding and where they can. So yeah you know, they're world's largest industrial owners.
Speaker 1:So I think uh, I think they could figure it out pretty quickly, right?
Speaker 2:yeah, yeah, they've got a really good team.
Speaker 3:They've got a really good team there yeah, I can tell you the design build team working with them is one of the best in the world. Oh well, look at that so they have people who know how to build data centers, so they went out, used all their billions and found the right team.
Speaker 1:Well, look at that, that's what happens when you have money. That's cool. You can go buy the right people. That is super cool. Well, thanks for coming See. The hour's ended.
Speaker 4:Look at that.
Speaker 1:My clock says we have eight minutes left. Is this the Howie roast portion of the show?
Speaker 2:No, no, no, no, jay, aren't you working yourself out of a job? Are you getting any space left?
Speaker 4:Am I doing what? No, do you have any space left?
Speaker 2:Where do you have space, if I can plug.
Speaker 4:H5 for one minute we do have space in the country and we do have quite a bit of power coming on, quite a bit of critical capacity coming online mid-year end of year, capacity coming online mid-year end of year, and we are actively acquiring more sites and more capacity. So we're in the market, we're making waves and we're ready to take customers.
Speaker 2:Can you talk about what you guys bought? Excuse me, can you talk about what you bought recently?
Speaker 4:Yeah, we bought the Yahoo facilities in Quincy, omaha and Buffalo Huge yeah. And we've got big plans for those facilities to open and grow and expand.
Speaker 2:Yeah, interesting, and if I could throw it out there.
Speaker 3:Prime is aggressively expanding in the Elk Grove Village area 180 megawatts, 240 megawatts down in the Phoenix market, and we're also expanding in Madrid, denmark, frankfurt, all over Europe. So there's a lot coming from Prime in the next couple of years.
Speaker 1:Wow, look at that, that is awesome. So educate me. Data Center 101. You say, hey, I got like five megawatts available. How fast does that happen? Like people snap that up, is it like?
Speaker 3:you had me at hello that's pretty funny wait, I actually, I, I.
Speaker 2:How long does it take to get a lease done in a data center? 100 megawatt data center?
Speaker 4:tell me that's, that's the yeah, that's, that's kind of the long, that's the long ten pole is going through the paperwork, going through the terms and conditions, going through the commercials. It takes time, what's that a year. There is the power of the LOI. So if somebody is very sincere about a location, an LOI can do a lot of good for that. Yeah.
Speaker 1:For sure. All right, ricker. All right, let's do our outtakes. Tell people how to get a hold of you, gentlemen, if they would like to go, you know, co-locate some servers at your locations, go ahead.
Speaker 4:Sure, I'll start. Jay Huzzi, H5 Data Centers. I'm at jhuzzi at h5datacenterscom or you can call me at 415-933-6244. And my phone is always in my pocket. I will always answer unless I'm preoccupied.
Speaker 1:Well, your AI avatar will answer for you too, correct?
Speaker 4:I've got my AI to do it Bill.
Speaker 3:All right. So I'm B Picard, B-P-I-C-A-R-D at primedatacenterscom. Please get in touch with me. I'd love to turn you introduce you to our whole team. I specialize in working with the brokers and the master agents of the world, but we've got a phenomenal team, so looking forward to speaking to people and I hope everybody watching is going to be at PTC. I know Howie's coming down.
Speaker 2:Yeah, pick me up on the way over.
Speaker 3:You know I will Jay you going to be at.
Speaker 4:PTC, Fortunately, no. We have a good group of folks that will be there and so you know, welcome to stop by and say hello, David will be there.
Speaker 3:I know David will be there, Of course you know you're welcome to stop by and say hello, david will be there. I know David will be there. David doesn't miss it.
Speaker 2:Yeah he'll be there. Guess me, howard Berry, avis and Young. You can reach me at howardberry, b-e-r-r-y, at avisandyoungcom or just Google best data center broker in the world and you, there it is is 408-375-6853. Too fun, guys had a great time. This is good. We'll do this more all right, thank you, amy.
Speaker 3:Really appreciate the thanks, amy, amy all right, thank you guys, all of you, amy, bye-bye thanks bye.
