Data Centers, AI and Innovation: Part 1 – Ep. 304

Welcome to the Accumulating Wealth podcast. I’m Hunter Satterfield. And I’m Judson Crawford. We’re CPAs, wealth advisors, and partners at Cain Watters and Associates, a financial services firm here to help you navigate the decisions you face every day. Today, we’re kicking off a multi-part series on data centers, the infrastructure behind cloud computing, AI, and so much of our digital world. 

Yeah, they’re becoming a bigger market story as AI drives demand for more computing power, electricity, real estate, and H2O. In episode one, we’ll cover the basics, what they are, where they’re being built, and why investment has surged. Let’s get into it.  

Okay, so Hunter, data centers have been, I think in a lot of respects, if you’re not really digging into it, they’ve been sort of the story on the periphery in the market. You hear stories about where they’re being built, how they’re being built. States shutting them down. Yeah, all kinds of stuff. A moratorium on them, all that. You’re right. 

But I also think that probably the average person doesn’t really fully understand the story, so we’re starting the story today. That’s right. And we’re storytellers, you know? We’re all about history and sort of simplifying the complexities, and I think this is something that Judson, you and I talked about doing. As soon as you start hearing places like New York that are like, “No more data centers,” and you hear about space data centers, and you hear about, “Hey, in Texas, you now have to stand up all your own power, and you have to have an energy audit,” and all these things as Texas sees more and more of them. 

And it is really a fascinating story. And what I will say, listeners, is that data centers are only going to continue to grow. And the reason is that you’re dealing now with a world where we’re about to have more and more autonomous vehicles. We’re going to have more and more robotics, and all of these things require data centers. 

But the really interesting thing is that I think people, Judson, might think that this is brand new. The data centers just started, like we’ve never had them before, right? And it’s only a result of AI. Is that true? It is absolutely not true. If we take it to a really, really small… and look, if you’re a computer engineer, I know that there are going to be some ways that this doesn’t relate perfectly, but if we think about it, in a lot of different ways, every one of our practices listening used to have data centers in their office. 

And then we went to serverless software, and we were like, “Oh everything’s out in the cloud.” Yeah, it doesn’t really exist. It’s just out there. But we don’t- it doesn’t exist anywhere. It’s in the poofy white cloud out there, right? Well, no. I mean, we still needed a place for all of that data to be processed and stored. It just doesn’t so happen to be in that little box in your office. 

Yeah. We just moved Ronnie to college a couple of weeks ago, and I’m installing his TV on the wall, and it’s just that you plug the TV in and then connect it to Wi-Fi, and he has everything he needs right there. And he’s like, “Dad, what did you do in your days?” 

And I’m like, “Well, we had to have a satellite that brought it in, and then I had to plug it into the back of my TV, and I just had the one DirecTV. That was it.” And that growth and that transition is actually part of the data center story. So folks, these are basically buildings, a lot of them now in Texas, actually, Carrie and I were just driving out to West Texas recently. Judson, your good old area out there. They’re everywhere out there, and we’re going to talk about like- There’s room. 

They’re buildings that are packed with servers and storage and networking gear, all those things that used to be in your college dorm or in your office. They’re packed with all that stuff, and that building basically runs the internet, the cloud, and now AI. But it’s not just AI. It’s all the other things. If you have a cell phone, if you have a smart device in your house, a Ring doorbell, if you log into your banking app, if you use any sort of device at home to say, “Hey, Alexa, turn on your lights,” that is all going through a data center. All of it. 

You can’t turn off one app and feel like you’re not using a data center. Exactly. Now, of course, given all of the things that are going on inside of that you do have to have cooling and water and energy. And so we are going to get into that story a little bit about kind of what is required. 

Before we do that, Judson, and folks, we’re going to talk here in just a second about why it’s expanded with AI. So, we’ll get to that in just a second. But Judson, maybe we just start with like, where are they? The answer broadly is everywhere. Right? 

Everywhere that there is space and everywhere that there’s money, essentially, right? I was actually really surprised at these numbers, and we can break it down a little bit, but the United States leads the world with over 4,400 data centers, okay? And we talked about this. This was actually a quiz that we had that I think we both got wrong originally is, one of the main areas is Northern Virginia. 

We call that the data center alley, okay? So, 4,400. And in fact, 80% of all new construction is U.S. as well. So, we’re growing faster than anybody else, but Judson, that’s not the full story because while we have more, it doesn’t necessarily mean that we have that much more square foot, right? 

That’s right. The surprising thing to both of us was that, and again, this is what’s reported, but China actually falls behind UK and Germany with somewhere between 370 and 420 reported data centers. That’s 10% or less of United States, and that really surprised us. 

What’s interesting is from what we can tell, even though they have that many fewer data centers, the square footage is only about 50% of the United States. So, the estimate right now is that the United States has somewhere around six to seven million square feet in data centers. China’s somewhere between two and a half and three million square feet. So, their per data center square footage is a lot bigger. 

Yeah, absolutely, and there is a reason that these two are the leaders right now. I mean, obviously China is a communist party that can stand up whatever it wants, whenever it wants. The United States tends to be a very innovative, capitalistic society that stands things up as quickly as possible and finds innovation and whatnot, and as a result, these are the two leaders right now trying to win the AI race. 

And so, I think that is a really good segue as far as, like, where these things are and we’ll double-click a little bit on places like Texas here in a little while. But on where these things… It’s a good segue on why did AI kind of force the story or the narrative here? 

Now, again, we’ve talked at length about how these things have been around for a long time. This is not just some new thing as a result of AI. Now, the difference is that if you think about the majority of apps that are being used, the energy usage or the data usage, I should say, is a little bit more cyclical, right? I mean, in parts of the United States where everybody’s asleep at the, the same time, Central Time Zone or whatever else, nobody’s googling stuff, nobody’s using their apps, et cetera, et cetera. 

And so, you do have a cyclicality in the usage of a data center itself. AI changes all that, right? Because you have these two components of AI, and I’m going to use some words that you might have heard on the news, but the ideas are training and inference. And so, what this means, folks, is that when you open your phone and you use Copilot or Grok or ChatGPT or Claude or any of these different models that are out there, those models have to be trained, and they have to be trained on data. 

And that training is happening constantly. So, the model’s being fed data over and over and over again, and that training is happening all day long, which means there is no cyclicality use of the data center. There is no cyclicality in the energy. It’s constant, right? 

The second piece of AI that makes it very constant is inference. And inference is when you actually use it. So, the training of the model is on the model side. Inference is when, hey, I’m actually going to use the model to create something for myself, whether it’s a stupid looking video that I come up with or it’s maybe support for a presentation you’re going to give. 

Well, that also is constantly going on. Now there are downtimes obviously, but what’s happening a lot in businesses right now, listeners, is that people will turn on their inference at night, and it will work all night to kind of create that for them. So, I think this in particular, Judson, is what spiked both the narrative but also the usage of data centers because AI is constantly working and never stops, and as a result, we do need more data center usage. 

Absolutely. And it’s not slowing down. And this is a result of, I mean, again it’s all sort of self-fulfilling, right? I mean, all of this market growth that you’ve seen in the technology area, I mean, a lot of that has to do with exactly what we’re talking about here, right? 

It’s investing so that these companies can keep up with demand of everything else that they’re building, that their consumers are using. And so, it is not slowing down. Exactly. Now, what we both agree on, listeners, Judson and I both agree on, is that the technology will continue to innovate. 

You see an entrepreneur like Elon Musk stand up a data center in Tennessee in like no time, right? You see the accessibility of energy with things like solar, and the transmission of it is much higher. So, the ability to stand up and use data centers will get more optimized because that’s what we do as an economy. 

But I agree, we’re going to continue to use more and more and more. So, even with that optimization that might occur, we might need to stand up more. And I think the fears that are out there, which we want to address, listeners, and sort of demystify some of these things, the fears are if we have to use more data centers or data center usage, we really have a handful of things that are needed. 

Number one is we have to stand them up, which requires steel, it requires concrete, it requires construction materials, labor, et cetera, et cetera. So that I think the economy broadly can handle from that perspective. But then it really uses two sort of natural resources. The first is water, which we’re going to address, and the second is energy. 

So, let’s start with water, Judson, because I think that’s probably the one that maybe people are most concerned with. Let’s put this in perspective. You want to talk a little bit about how much water do they actually use? 

Okay. Well, let’s- before we do that, why are we using water? Oh, yeah. That’s a great idea. So, we are using water to cool. That’s right. It’s all about cooling. Now, we are going to have a listener on that’s talking about future developments that are happening, and what’s actually happening now is a lot of these things are being put into the ground to create natural cooling itself. 

But yeah, you’re exactly right. That is why we’re using it especially in places like Texas where it’s super, super hot right now. Yep. Especially right now. It’s Death Valley here. 

Yeah, so large data centers can consume somewhere between one and five million gallons of water per day, okay? And obviously, again, that is different based on climates. It’s different based on facility size. And so it can be completely different based on those things. We’re already seeing advances in this. Like you said, they’re looking at more underground units. 

Unfortunately, right now there’s this thing called latency as well, which is basically response time, getting data back and forth between the data centers. So, we can’t today stick data centers in Antarctica, right, where it’s really cold and use no water. But they are getting more popular in cooler climates and again, I’m sure that’s something that will continue to change over time. But yeah, there are your numbers, 1 to 5 million gallons per day for a large data center. 

Yeah, and I think that they are using now, again, just to demystify very quickly, they’re using now recycled water, and that’s helping a ton as far as water usage. But really let’s demystify it for this for a second. So, on average right now, U.S. data centers use 17.4 billion gallons a year. That’s a lot. Like that’s just crazy, 17.4 billion gallons a year. Now, that 17.4 is difficult to break down between what is AI doing versus what everything else is doing. But the reality is we have to have it no matter what’s doing it. 

So, if you’re concerned, we’ve already made the joke about almonds, which use considerably more, but if you’re really concerned, let me put some in perspective. U.S. golf courses use 531 billion, 30 times the water of data centers. So, if we’re really concerned, maybe we should stop eating almonds and stop golfing.  

That’s going to be a household battle. It is. Whether one partner gets to keep using TikTok or the other partner gets to keep playing golf. It is. Who wins? Yeah. Arm wrestle, I guess. Yeah. 

In addition, extending it even further, lawn and garden watering, just home lawn and garden watering so we can have pretty plants in our front yard. We’re not even in billions anymore. We are in 3 trillion, so 2,900 billion gallons versus 17 billion. I can’t even do the math. Thousands of times more for lawn and garden watering. So especially as we talk about data centers, as they get better at using all these alternative methods, it’s going to get even better. 

But it is a fraction of a fraction of a fraction of water usage in the United States. It’s just new and something to attack in a way, in reality, and people don’t know how to handle it. 

So, we’re going to talk about energy, but before we do that, Judson and I love Texas. The stars at night are big and bright. deep in the heart of Texas. West Texas is a wonderfully beautiful place, if you’ve never been there before. And Carrie and I were driving out through West Texas this summer, and we saw this big data center out there, and we’re like, “That is an eyesore.” And I do think there’s an element where people are like, “Oh, it’s so ugly, it’s so ugly,” et cetera, et cetera. And I don’t disagree with that. 

But let’s talk about land usage for a second, Judson. Currently, the existing data center buildings and sites plus what’s estimated to be built in the next year, take up about 1,400 square miles across the United States. The total idle farmland is 40 times that. 

So, I think we’re okay. We’ve got the land for it. We do have to deal with, like, how do we actually get the water out there, and are we using it in the right spots, especially hot climates? How do we get the energy out there as well? But I do think we actually need to talk about energy as well. 

And I think on energy, what’s really interesting is that at first glance, and we see this in Texas now, Governor Abbott has said, “Hey, you have to do an energy audit to prove that you’re not drawing on the energy of the population that you’re in right there.” This is why they put moratoriums on them in a lot of places as well, which I think is very shortsighted. 

But what’s interesting is a study just came out from the United States government that actually areas with data center growth are seeing lower overall electric rates by about 6% than areas without. And I think it lends itself into what’s happening. This is classic Jevons paradox. Where they need power, they’re standing up power, whether it’s more LNG, more oil and gas, solar, wind, whatever, they’re standing up more of it, so it’s actually driving the cost down for everybody in those neighborhoods. 

So that I think is important. The creators of the data center should be responsible for creating their power, and many of them are. In fact, the largest data center alley, like you mentioned, in northern Virginia, they had already stood up that huge wind farm off the coast, which some people believe is ugly, and it’s fine, I get it. At the same time, they’re standing up their own power, and so what’s happening is it’s actually driving electricity costs down for those communities. 

Well, and again, I think that this is a really good example of how we will adapt and how we will advance this to where some of the concerns that some people have today will be assuaged because we will just adapt. I mean, as an economy, we’re not going to not get this to where it needs to be, or maybe going back to our prior point, we’re not going to stop using it. 

Exactly. We can’t, unfortunately. And one of my favorite things is Wells Fargo just came out with an excellent survey or an excellent data set, excuse me. These are real numbers. In counties with operating data centers currently- So these are your current ones that are already constructed. They have more housing units built. They have overall higher home values, which actually drives property taxes for these areas. They have lower unemployment. These are numbers from Wells Fargo, not me. And they have more job growth. Again, let me repeat that. More housing units built, better home values that are growing, lower unemployment, and better job growth. 

Don’t know what else you could ask, but what we see as an economy is where innovation goes, those things tend to occur. So, I think as long as our creators of these are doing so in a healthy and responsible way, which is where I think it’s helpful to see the government come in and say, “Hey, I want you to show me that you’re not pulling on the energy of the electricity of the area.” As long as they’re doing that, man, there’s some really good benefits that have come for those counties. 

Yeah. When you look at it, it’s just business. You know? If we take the emotion out of it, and we really just look at it, it’s bringing businesses to areas that can use jobs and can use these types of things. So, there are huge positives to these as well and we can’t just focus on the negatives. 

Now we’ve talked about this a little bit, and we’re going to talk about this more in our next two episodes on data centers. But let’s talk a little bit about the dollars. The dollars? Yeah. Yeah, I think that’s a good idea. The dollars that have gone into this, okay? Because it’s not small. No, no, we’re in the billions for sure. 

It’s surging. And again, I think this is consistent with what we talked about with AI does require more because of training and inference. But I’ll give some past numbers, and then Judson, maybe you want to talk about the growth going forward. But construction starts in 2023, about $15 billion in the U.S. $15 billion in 2023, $27 billion in 2024, so it doubled year over year. 

And then it almost… Well, no, it did triple from 2024 to 2025 to $77 billion in 2025. So, we’ve gone in just the last three years from $15 billion to $77 billion. Now it is important, we have talked about this at length, that the construction of data centers does require concrete, it does require steel, it requires raw labor. And so, this is one of the reasons we’ve seen small cap surge so much over the past few years. 

So that is important to see that it’s actually feeding itself into the economy, and much of the GDP growth that we’ve seen in the economy is a result of these. Now, the question becomes, Judson, can we keep this pace up, this tripling that’s occurred from 2023 to 2025? What say you? 

No. We’re not going to keep tripling this number time after time. And I think that… and again, Brad talked about this on one of our story times with Brad and he related this to when we were expanding the fiber optics in the country and that there was this race to put in all these fiber optics all the way around the country. 

And the reality is that at some point it got over-invested, and there were probably a lot of fiber optics that were laid in the ground that were never even actually used. Now, I’m not saying that we’re going to go to that over-investment point. We tend to do that to a degree. And at some point there will be some exhaustion there when we get to the point of expansion enough or technical advancement enough to where we just don’t need this to continue at the rate it’s been, right? There’s probably just not enough investment for this to continue to double and triple every year anyways. 

Yeah, it’s going to go up for sure. I mean, they’re saying $600 to $1 trillion of investment over the next 12 months, which, who knows? Does it happen? That’s what’s planned. We’ll see. 

What I do know is what we already said before is that as we have seen in the past, they’re going to get more innovative. They’re going to create models that get to answers more efficiently. And ultimately, we’re not going to just triple to the moon to where there’s a data center on every corner like a 7-Eleven. 

I mean, they will get to a spot where we’re using it much more efficiently, effectively. But I do think that we are going to see considerable growth over the coming years as we continue to stand these up, which maybe lends into a little bit about the episode we talked about of the potential growth in the stock market and small caps. We’ll see. It will be interesting. 

But assuming that we start to see the benefits from AI in the broader economy continue there’s no reason why we won’t see continued growth. But of course, we can’t triple and triple and triple and triple. That just wouldn’t make sense in the broader economy. 

Well, and I think – and you make a good point, when you were talking earlier about the things that have to go into a data center, concrete and construction and all these different things. And we’re going to talk more on these next couple episodes about how data centers… Because people- in I think clients’ minds, they’re like, “How do I invest in this growth in data centers?” 

And in a lot of ways, just by your broader investing, you’re investing in the growth of these data centers because you may own that construction company that’s building one. You may own that concrete company that’s selling concrete. You may own all of these different things that are going into this. And we’re going to talk about that more. 

So, Hunter, because we are essentially two data centers right here, I don’t want to go on too long with this episode. You want to talk a little bit about what we’re going to be talking about on the upcoming data center episodes? Yeah, so next one we’ll have somebody that’s going to come on and talk about what these look like when you design and develop them, what future ones are going to look like, what advancements are occurring. Should be a fascinating episode. 

And then we will wrap, folks, with an episode on what does this mean? Like, what does it mean for your business? What does it mean for your financial plan? We’ve led a little bit and foreshadowed a little bit on those things so certainly you probably know some of that already. But it should be a fascinating story. And again, I think we’ll bring some more context to what these look like, what they mean in probably the next three, five, and 10 years. Well, that’s good stuff, Hunter. Looking forward to the next two episodes on data centers. 

Listeners, the best way to keep up with us is to subscribe to this podcast. That way you never have to miss an episode. You can go back and listen to all the others. If you’re enjoying us so far, leave us a review. Have a question, comment, or suggestion for a future episode? Drop us a line at cainwatters.com/wealth. We really do answer these, and we love to hear from you. And if you want to learn more about what we do when we’re not recording these episodes, visit cainwatters.com to see how we’re helping almost 4,000 clients reach their long-term financial goals.  

I’ve got a question for you, Judson. Is it data or data? Data. No, it’s data because “The Goonies”. Data’s in “The Goonies”, data is a center.