EP #154: Why Concrete Is on the Critical Path for AI Data Centers

Seth (00:00)
All right, and welcome to another episode of the Concrete Logic Podcast. And today I have Douglas Mouton with me. He's a global expert in hyperscale data center infrastructure with over 30 years leading design, construction, and delivery at Microsoft, Meta, and Jacob Solutions. He's a decorated US Army Colonel, retired. Thank you for your service. My pleasure. Doug has commanded units of up to 3,000.

personnel and lead recovery efforts in Afghanistan, a post Katrina, Louisiana at Microsoft, he scaled global delivery to over one gigawatt per year, have costs and delivery times and established best in class safety programs. At Meta, he drove design innovations, cutting carbon and costs by 40%. Doug now advises top firms through Muton advisory services, offering strategic insight into digital infrastructure.

AI impacts and sustainability. Serves on numerous boards. Do you have time for anything else, Doug? Come on.

Doug Mouton (01:02)
It's a busy time.

Seth (01:04)
Yeah. and, he's, sought after speaker. I'm happy to have him here. We we've been going back and forth for a long time. finally connected. So Doug, thank you for coming on the show today. I appreciate it.

Doug Mouton (01:15)
Thanks for having me. Thanks for your patience. I know we go back and forth trying to find time to connect. So glad you're here.

Seth (01:20)
So if somebody was going to guess, guess what we're going to talk about today with Doug on here. We're to talk about data centers. So we're going to talk about how important concrete is to the data center industry growth and what we should be thinking about as an industry. But before we get started, I just want to remind everyone how you can support the podcast. If you go to www.concretelogicpodcast.com, there's a homepage there. And on there, there's a couple of ways you can get a hold of me.

a ask Seth at the in the menu at the top of the navigation there. Click on that and you can shoot me an email and what I'm looking for topic or guest suggestions. A lot of our episodes are driven by listeners of the podcast. So get on there. Tell me what you want to hear, which one to learn about. If you don't like that option there in the bottom right hand corner, there's a microphone icon there. You can click on that and somebody did this the other day and what what that

does is leave me a voicemail. So you can just tell me what you want to hear or whatever your idea is for the podcast. Please do that. The third way you can help the episode is on the same homepage on the concrete logic podcast.com homepage. There's a donation button in the upper right hand corner. We typically don't advertise, but you might been hearing some advertisement lately before the episode, but within the episode itself.

I'm trying to keep advertisement out of that. And how you can help me do that is by donating to the show and that helps keep the podcast going. So click on that and you can give any amount. There's not a set amount on there. You tell me what the podcast has worked to you. That's how that works. And then the last thing is, is if you go to Concreelogicacademy.com, there's a way for especially professional engineers to get their professional development hours.

And it's built around the podcast. So same content. We just have more resources for you on the Academy. In addition to the content, also there's quizzes to make sure you're absorbing what we're talking about. And it's a little bit more formalized in the podcast. So, all right, I'm out of breath. So we're going to talk about again, how important concrete is to the data center industry's growth and what we should think about.

And, you know, work for Baker Concrete, or, sorry, Baker Construction, Ruru Brandon, Baker Construction. It's very important to me, but how important is it to the data center industry,

Doug Mouton (03:45)
Data centers need concrete. They're the foundation of all the buildings. So we can't get a data center without concrete. It's a bit of a paradox though in the data center industry because when the grand scheme of things, you look at the breakdown of all the materials that go into a data center, concrete is single digits in terms of the total cost. But as we look at the evolution of the data centers today in the era of AI, in the era of bring your own power,

We're now seeing gigascale projects. Look at Stargate in Texas with just line after line of combined cycle gas power plants on top of these massive data centers. What do all these projects need? They need concrete. And look at the scale of the projects that are being announced right now. Look at the $10 billion plus announcement in northern Louisiana and many others are just being announced day after day, it seems. And the idea of

of concrete being such a small sliver of the program, but at the same time, when you scale the aggregates and sand and cement that needs to get to site, you got to build a railroad. We're having to rethink how we scale the delivery of materials. So concrete quickly becomes on the critical path to get the foundations in, which allows us to get the structures up.

And getting dried in is a fundamental step in any data center project, because that's when the real work starts. As you know, the mechanical electrical work kind of trumps everything else with the scale of it. But we can't start any of that until we get that building dried in. So that's just one example of how important concrete is in this day and age.

Seth (05:26)
Yeah, should probably back up a little bit. Just to can you explain why there's such demand for data centers for folks that may be not familiar with data center is?

Doug Mouton (05:38)
So data center is a warehouse that we put computers in. It's not a specially complicated building. It has like four systems. It's got a power system. It's got a shell or an envelope. It's got a mechanical cooling system. And it's got fiber optic connectivity. It's really those four systems. It's not a clean room. It's not anything super scientific. It's just like a Volkswagen that you put a V8 in. It's a small, simple shell with a really

powerful powertrain in it. Then now with the advent of AI, the cooling is becoming more complicated. Over the last 25 years in North America, we've gone from zero data centers to 35 gigawatts of data centers. That is the unit of measure for data centers. We don't measure square footage or cubic area. We measure how much IT or

Information technology capacity of building can manage and that's called IT capacity and is measured in megawatts, which is a very big unit of measure in power. So that said, we look at where it's evolved. Well, what does it evolve for? It's evolved as a place originally to put compute and storage workloads. What is that? Why is that important? Well, before the cloud happened, all your big enterprises had like their own computer room and they had like

some servers in there and they had some IT staff and they had to do all the stuff to manage their own internal infrastructure, both hardware, software and personnel. And it's very expensive and it's decentralized. Everyone had to like roll their own and even small offices had to make investments. So the cloud shows up as a value proposition that says, hey, we can give this to you as a service, software as a service, infrastructure as a service, and you just pay when you're using it.

and we'll buy the hardware, we'll house the hardware, we'll have the IT staff. Suddenly, the cost for the IT component of many businesses just drops down, makes it much more efficient. This is how that whole game started, and it's grown, and it started in some of the concentrated areas that touch demographics, that touch key connectivity nodes, fiber optic nodes.

like Northern Virginia, you know, it's gone crazy. If anyone looks around that area, Phoenix, Dallas, Fort Worth, central Washington state was actually a big area. And you go on and on and cite these clusters. So over the last 25 years, 35 gigawatts of power have been installed across all the plant. And this is before we could even really get serious about AI. And AI has changed the game because

AI is another level of compute capability that allows us, and we're all talking about it today, what it can do for us, how it can change our life, how it can simplify things. So that is a spike demand even greater. So what we see from now until 2030, that 35 gigawatts on the low side of estimates will grow to about 60 gigawatts, doubling. That's the low side estimate. The high side is about 90 gigawatts in North America by 2030.

So that demand is that core compute and storage I discussed earlier about nine, 10 % of that. And the rest is these AI training centers. You see these big facilities out in the middle of nowhere. They're training, they're creating a neural network that creates a brain and that brain creates general intelligence, the algorithm, that's a grok or a Claude or a co-pilot. And then it gets monetized. It has to get monetized and it gets out to people. You grab your phone to say, hey, oh.

solve this problem for me AI. Now you're in furry, now you're using it and that's a whole other area of demand. So it's these three areas of demand that stack up, computing storage, the traditional workloads, training AI models to create general intelligence and then inferring it, putting it to work and monetizing it.

Seth (09:41)
Yeah. Yeah. So when they first were starting to build these things like Northern Virginia was kind of the place where it all sprouted from. And now what we're now we're seeing them going out to rural areas, which is getting a lot of attention lately. I don't, don't, think people that live near these projects that are being thrown out there are starting to pay attention to just because of

information that's put out there that might not be totally true, might be partially true on the resources that the data centers use. But can you talk a little bit about why we're seeing more of the data centers being built in the rural areas versus near the cities?

Doug Mouton (10:27)
Yeah. So we can go back to those three demand streams. That computing storage wants to be close to demographics that they serve, wants to have a low latency experience for people using that product. Like if you had a 911 system for New York, you want your data center really close so it's performing fast, it's one idea. But when we talk about training workloads, the AI training to create the algorithm, it doesn't need to be connected to the world. It's not serving customers.

You can have this isolated massive data center and you just dump a bunch of data into it so it can train itself. It's literally getting all of these servers in synchronization and they do a number of load profiles that they get in sync and they think and then they pause and then they assimilate. I'm not a smart guy about the exact functions, but it does not need to be connected necessarily to.

the closest concentration of demographics in fiber optic neuro centers, if you will. So that is what we're seeing right there. And we should also call out that Bitcoin mining is a form of data center as well. And they're also not latency sensitive. They are just mining coin, if you will. And they don't have to be anywhere particularly close to the historic areas. So those are some of the things driving

rural development of data centers. And I think also causing some of this an NIMBYism not in my backyard is some of them in the rural areas, particularly, I'll just say it, the miners, they're pretty scrappy, and they're kind of noisy. And there's been some negative outcomes around what that feels like to have that big development around you. You also recognize in a rural community, you're building data centers and suddenly,

out of nowhere there's just all these trucks running up and down the road, destroying the roads, making it unsafe for people, just making people uncomfortable. And it's change, know, and we're all, you know, as humans a little resistant to change.

Seth (12:32)
Yeah. So when we're moving out to rural areas, it was already pretty demanding to keep up with these schedules. The data center schedule, construction schedules are compressed. Yeah. they're, you know, got to get these things up and running so you can start collecting some dollars.

Doug Mouton (12:50)
Everyone's

going to race to make the best algorithm.

Seth (12:52)
Right. So we're moving out to these rural areas and it's even more challenging to get your labor and your materials out there. So can we talk a little bit about what is important on the, we'll just focus on the concrete side because that's all I know. So the concrete side of what I guess, first of all, what is the, what is, we were, we had a great discussion before we hit record by the way.

What is, I guess, what is the ownership or the developers that are building the data centers? What do they want from their concrete? Like what's top of mind?

Doug Mouton (13:29)
That's a great question. So it's kind of important to understand the market of data centers. As much development as you see everywhere, if you really go back to who the customers are, it's really like less than 10 outfits. And it's the big hyper-scalers that we know like Google, Microsoft, Meta, Amazon, Oracle. So these are the folks that are driving the spin of that big wedge we talked about of demand.

It's mostly those players right there. You got some NeoCloud and others, but that's where the real demand comes from. You look at what's driving their capital programs and they have two things they can do. They can make it or they can lease it. You'll see data centers being built by QTS or other digital bridge, these really big outfits that they're delivering, but they're serving that hyperscale customer.

or the hyperscale customer will build for themselves. So 70 percent of data centers are built by third parties, 30 percent built by the hyperscales themselves. Back to your question, what do they want? Well, they all have a little bit of a different personality. So some want to go really fast, some want to make sure that they're really environmentally sensitive, some want to find that balance. In general, we find that they are all driving against

some ESG requirements, meaning there is still some environmental social governance that they've been driving that they will continue to drive on because they're global companies. So they're worried about their emissions. That's why you see the investments in renewable energy. That's why you see a lot of work in understanding what the scope three emissions are in data centers, which is the embodied carbon and all the materials. So that's one piece of it. They also, they're in a race. They're in a commercial race to

build the best algorithm, the fastest they can monetize it, just like you said. So they want speed. They want cost effectiveness that can offer their services at a better rate. And they want an understanding of their materials that reduces their environmental footprint. And a thesis that's emerging among them, most of them, is that the way we can really drive that efficiency is by making our designs leaner,

simpler and really honing in on using less. What does all that mean to concrete? Again, you can look at this pie shape that would be all of the materials that go into a data center, half of its mechanical electrical equipment and wire and piping. The other half is like the last quarters of building itself, so you got some concrete in there, and then you got all the site work, the roads and the substation. In aggregate, you might get 10 percent of the total project is concrete.

The challenge that we have is we know also that concrete after steel has some of the highest embodied carbon in it. So that created a tension by the customers, the end users. Hey, how can we make this better? So there was a coalition across the data center providers that said, hey, how can we get more recycled content like fly ash and slag and other novel pozzolins to get in there and allow us to use less cement? And can we get it to cost less? And can we get it?

to be viable. So a lot of work was done with that AI was actually used. But what we learned is that the buy cycle to get to folks like Baker Construction is through the general contractors. And the general contractors don't really care in the grand scheme of things who's buying the concrete because they got so much other stuff to buy. You were telling that story earlier. It's like,

You know, no, no, no harm, no foul set, but you're not as my most important thing right now. I got to get a bunch of copper in here. I got to get a bunch of chillers and all this other stuff. Right. So, so we're not controlling the opportunity. So if we have like four different places that are using concrete and everyone's getting their own concrete. And by the way, we're building power plants next to these big gigawatt centers. we're missing a huge opportunity because everyone's kind of competing for the same aggregates, the same sand, the same cement.

It's not coordinated. So A, we can create a bottleneck of supply on which is a relatively small material, but critical for getting these buildings dry like I was talking about. And then if we're driving towards like, can we get a mixed design that we think works better and we're working with the, know, the and rises of the world or, gosh, what's the other one top of my mind? It'll come to me.

How do we get commonality of these mixed designs? How do we get that to the batch plant? How do we make sure we have this consistent material? And you're seeing it in your business right now. You're seeing these pilot projects. saying, how can we get low carbon AI designed concrete into our projects? And to do that, we have to break these silos of how we procure both concrete as a bulk product, concrete that's mixed in the batch plant, concrete that's, you

formed and emplaced. And we've got to be much more intentional. And we got to get past the general contractors in the same way they give, I think you're mentioning this, the mechanical electrical subcontractors, they get a lot of, they're at the table. They're right there, they're at the table. We need to get our concrete supply chain at the table. So we have a coordinated, you know, method to really get the best value out of our concrete and the best efficiency.

Seth (19:02)
Yeah. Well, I think a couple of things there. You mentioned that concrete and it's large CO2 or carbon footprint. And that's just based on volume because the amount of concrete that's used in the world, it's the most used manmade material on earth. If you took a unit of concrete and compared it to steel, there's vastly different

Yeah, yeah, it's something so when someone says concrete has an intensive Amount of carbon that it it puts out in the world That's partially true of partially false. It's based on volume. It's just use so much. Yes Yeah, so the The the other that what but what you mentioned there, I think this kind of solves

the challenge at hand is the design. So if they can skinny up the design, they would use less concrete or less cement. The cement is the one that gets all the attention in there. But if they would skinny up the design, they would use less concrete and that would use less or emit less carbon as well.

Doug Mouton (20:08)
and would cost less and it build faster, which are other pillars that is really important to everyone.

Seth (20:13)
Yeah. So that, that to me, that one would hit a number of those, goals that they, that they have. it, I did, appreciate someone, I can, I use, I used to scoff at these guys. I would create these new co and, know, concretes and cements and everything. But I think what, what, what someone pointed out to me was what's great about these things is that we're innovating here in the U S so where it's kind of a way of starting.

you know, reshoring manufacturing here in the U S and I thought that was a, a, a, a great point. So, you know, encouraging those guys to keep working. It's just, I, I've become skeptical cause those things are hard to scale, at least the ones I've seen out there. So there's a point where those don't. Yeah. Yeah. So it's going to be hard to scale, but I know there's pilot projects out there and they're trying to come up with new things. So it may have.

Doug Mouton (20:55)
100%.

Seth (21:03)
it may have its use case in some of these projects, but to kind of take over the entire industry, it's just totally skeptical at this point, but I look forward to being proved wrong. So yeah, I appreciate you sharing that because that's all, you know,

when we get these RFPs and it's like, we want, we want less carbon, we want our concrete done fast and quick and we want it now. So it's like when you're a concrete contractors, like, just want to beat my head against the wall. Sometimes

Doug Mouton (21:39)
Where are going to get it? Can you get enough of it? So it's Heidelberg and the other suppliers, they've got major suppliers. So yeah, there's all this innovation that we're seeing like algae concrete or some other concrete and you basically get a ziplock bag full of it. But you can't get it at scale. So I think the strategy to get low carbon concrete is to really work with those suppliers so that they have a base product that comes to you in a normal supply chain. Yeah, I want this

type two or whatever those types are that you're getting in there. And then making sure we're specifying, know, posilums and other substitutes that are available at industrial scale. it takes discipline, I think, and pre-planning. You know, I think about concrete on a gigascale project. It needs to be one of the first things you think about. Why? Well, first of we've got to get a logistics worked out. There needs to be a railroad coming in for all the stuff for that.

that massive construction project. We need to figure out where the batch plant is going to be. And if we're going to you know, substitute cementious materials, where are they coming from? Are we going to mix it in situ right there? So where's that supply chain coming from? Let's make sure that we understand the locally sourced aggregates and sand. We understand all the dynamics of it, how it's going to behave. We should be testing all of that stuff while we're doing our grading. We should be making

everything dialed in. So then when we're really go, we got our supply chain set up. We got our mix design confirmed with all the real materials that that site is actually going to use. And then we can do a few, you know, test mock-ups, make sure we all know how to finish it and everyone's happy. You like it, everyone's good. Then we just go and execute. I think that's the level of coordination and silo busting we have to do to really get the best value out of our concrete programs.

Seth (23:24)
Yeah, I agree. That was an easy sell, Is there any other thing you wanted to make sure we discussed today?

Doug Mouton (23:32)
Yeah, I would say that concrete is also showing up in the power space, obviously in the foundations. But I was just with Skidmore, Owens & Merrill, one of the biggest architectural firms in the world. And they're looking at how to solve this pumped storage. How do we unlock the intermittent characteristics of our utility, whether it's mismatched

demand with mismatched supply, whether that's renewables that aren't being accessed or producing power, we don't need it curtailment of workloads, whether it's a gas-fired generator that's pumping and staying going but doesn't have the workloads, how do we store that energy? So one of ideas is they want to build a 620 meter, that's like three times taller than a wind turbine that can hold enough water

so that it turns a turbine and creates a gigawatt of power for an hour. mean, that is like, it'll unlock a lot of the utilities that we need to grow to those big numbers I sharing with you. And some of the solutions we're looking at, gravity batteries, which is basically hoisting up huge blocks of concrete. they're learning a lot about energy vault, kind thing, learning a lot about the mix designs and the reinforcement designs of disaster at first. So a lot of more coordination needed with the industry. And then there's 600 and

20 meter tower is literally a high rise building in the middle of nowhere that is going to need a mountain of concrete and a mountain of logistics. So we talked about AI being an effective use to get more sustainable materials in. One more call out is we're using deterministic large language models to make our steel designs more efficient. OK, why does that matter for concrete? Well, if you apply the same logic to rebar designs, we use less steel, which costs less.

sets up faster, it displaces less concrete. So we use less concrete and we can build more. So I think we've got to keep pushing for those incremental innovations and make sure they're scalable as we go. I just want to illuminate some of those different use cases that are showing up.

Seth (25:35)
Yeah, what about the other building systems, the servers themselves and all that? I assume there's some design efficiency going on there. mean, these things are going to get more compact because your weight of your building has a lot to do with how much concrete you're using as well.

Doug Mouton (25:52)
I

that's right. I think the point load on the surface, on these slabs is going to get higher because the density of the servers, if you look at the systems inside, the servers are getting smaller and more dense. They're able to run at a hotter temperature, which is more efficient, which means your cooling plant doesn't have to expand as quickly to keep up with the server density. Power plant has to increase like that, but our buildings are going to get, I think, theoretically smaller, but stiffer, if that makes any sense.

Seth (26:20)
Well, this was great, Doug. Did we cover everything for today?

Doug Mouton (26:24)
Absolutely, it was a great conversation, Seth. I hope I can come back and give you updates as things are evolving.

Seth (26:29)
Yeah, did they talk about your 1800 foot tower? Yeah. All right. We're gonna, we'll get Doug's link to his LinkedIn. That's his preferred way of getting hold of them. And so please check out the show notes for that. Doug, thanks for coming on the show today. I appreciate it.

Doug Mouton (26:33)
Beast

Thank you, Seth. I admire the work you're doing. And as I got into this concrete world, I really found your podcast very informative to kind of quickly get a handle on what's going on in the industry.

Seth (26:59)
Appreciate that. Thank you. Until next time, folks, let's keep it concrete.