EMCOR Group, Inc. (EME) Earnings Call Transcript & Summary

All right. Good afternoon, everybody, and thanks for joining our fireside chat today with the executive management team of EMCOR, which includes CEO, Tony Guzzi; and CFO, Jason Nalbandian. We picked up coverage right after Labor Day this year and came out with an outperform rating, right around $600 per share, which was up sixfold from about $100 per share only 3 short years ago, driven by the incredible momentum that we've seen in the business. And shares continue to move higher with every incremental headline about industrial reshoring and AI infrastructure. So with that as a backdrop, I'm excited to have the team with us today to talk more about the growth drivers and industry dynamics. So Tony and Jason, thanks for joining us today as well as sorry, Lucas, I should have you as well. Lucas Sullivan, IR. But yes, thanks for joining me, guys.

You bet. Thanks for having us.

You bet. So I want to start out with just a quick recap, state of the union. You reported your third quarter results last week. A few of the items that really stuck with us is, one, you're continuing to see broad-based revenue growth. I think 10 out of your 11 construction end markets grew this quarter almost at a double-digit clip, I'll add. The backlog is continuing to grow #2, strongly, and that's particularly true for data centers. I think the backlog there doubled from last year, which leads me to point number three. You shared that you believe you can maintain a high single-digit to low double-digit organic growth rate over the near term. So maybe with that, can you just spend a little time, share your perspective on the current state of affairs, how you're thinking about things stepping back, like how are things playing out this year relative to your initial expectations? Anything surprising you one way or another, just at a higher level?

Yes. So look, I think you captured a lot of the high points well, Tim. And thank you for setting this up today. And everybody on the line, thanks for joining us. It's an exciting time to be the CEO of EMCOR, and I've been the CEO for a long time. But I think you captured it fairly well. RPO is up 29% year-over-year, about to $12.61 billion. 80% of that growth has been organic. I think you captured it well. You see good diversity of demand because our -- it's not only our revenue growth, we also had RPO growth in almost all of our sectors. I think diversity of demand is something that's important. Yes, we are growing well in data centers and some of the other things focused on nearshoring and reshoring and all those things. But we also have good diversity of demand at EMCOR, which I think sets us apart from some of our peers. We're growing our revenues. And Jason, when we hit this later, we'll hit this later when you ask about labor, but just to get it upfront, we're growing our revenues 2 to 3x the rate of our headcount growth in our Mechanical and Electrical Construction segments. And that's important because it ties into a point you make later. We are also seeing opportunities to grow, which to us is one of the key capacity increases is we're having the ability to -- because of our growth to grow more foremen, superintendents, project managers and other folks that help us expand and be able to do more projects and larger projects. We continue to see good aftermarket growth. That gets lost in all this, right? We continue to have good aftermarket growth, especially in our mechanical services business. And we've had really, I think, we have a record of really good capital allocation, but we continue to have really good capital allocation. We're in the process of divesting our U.K. business, which we have built into a pretty good business. We'll be selling that this year, pending U.K. regulatory approval. We've made over $900 million worth of acquisition, the latest being our John W. Danforth acquisition, which we actually signed -- will close here in the fourth quarter. We actually signed in the night before earnings when we want to announce it. And we also have done $400 million plus in share repurchases. And we've done all that by having cash flow generation that continues to be best-in-class. We expect to be at least equal to net income when we do that and we have a long track record of doing that or a little bit better. And we leave 2025 as we exit out of the year, feeling pretty good about as we go into 2026. And we're doing all that by earning the best margins we've ever earned between 9% and 9.5% on a consolidated basis, operating income. Remember, we do operating income. A lot of people do this sort of EBITDA accounting and EBITDA accounting. We're operating income. We figure everybody on the phone smart enough to add back amortization. And we continue to invest for the future and development of our people. Jason, you got anything to add there on the highlights upfront?

I think Tim hit it, right? 10 out of 11 of our market sectors grew when you combine electrical and mechanical. I think the thing for me that's always interesting is if you look at the business overall, you strip out what we've seen as high-growth sectors over the last several years. So you strip out networking and communications, which for us is predominantly data centers, you strip out high-tech manufacturing and you look, how did we grow the rest of the business in '24 and for the year-to-date period of '25, and we're still growing it at mid- to high single digits. I think in '24, it was roughly 4% to 5%. And year-to-date, we've seen the rest of our business grow at roughly 7%. And so I think that's a testament to really the diversity of our business. And Tim, you mentioned kind of our near-term growth rates, and we've said this before. I think for us, if you look at how we've grown historically against non-res, I think that's a good predictor to how we'll grow into the future. And for us, on a consolidated basis, we typically grow 200 to 250 basis points above non-res. And in our Construction segments or our Mechanical Services group, we're typically growing, let's say, 500 to 600 basis points above non-res.

And that's where we get to the 9%, high single digits, low double digits. in our Electrical and Mechanical segments that you mentioned when we kicked off the call, Tim.

Okay. That's a good level set. Thanks for providing that overview of the business. It sounds like you guys are largely thinking about it the way we are. Although I will say we raised our estimates on the most recent quarter for 2026. So although that's a good growth algorithm on how to think about it, you do kind of continue to surprise to the...

Yes, we continue to outpace it a little bit. [indiscernible] talking about, doesn't have any acquisitions in it, next year.

Good point. That's a good point that -- yes, that's right. Although my model doesn't include the most recent acquisition because you announced it, but you haven't closed it yet, correct?

Yes. And then we have the U.K. coming out. So, we have a month of Miller coming back. We've done some other small ones. We're pretty good about breaking that out when we give guidance in February.

Perfect. So I'll move on here because, Tony, I was laughing in a recent fireside chat, I think you made a joke that no one wants to talk about anything with you except data center these days, which I understand. But actually, I do actually want to start this off by talking about some of your other markets. Cool. So I mean we've seen strong growth in a lot of these areas and something we've all come to appreciate more and more is just how long and coming the push and need to reshore has been and the potential for this to be a multi-decade theme. So maybe I could start with these high-tech manufacturing opportunities, semiconductor fabs, a lot of MEP work goes into those, I think. But obviously, these are large projects. A ton. Okay. All right. That's what I thought. At the same time, it seems like the U.S. may be in just this initial push of what might be, I don't know, 10-plus, 10, 15-year surge in CapEx to get the critical mass of chips here in the U.S. How much has your team already been involved in this area? Because I don't know. I'm new to the story. And I know you guys have talked about semi, but like how much have you already been involved in this area? And how do you think about the opportunity over the next few years? Is it less about as much? Or more than you have been? I just don't, I guess, have a good level set for.

Well, I'll give you some historical perspective. I've been at EMCOR 21 years. We've been working on semiconductor plants on and off for 21 years. Now what's different this time, that was for 1 or 2 particular customers, one in Arizona, one in Salt Lake City, a little bit in Boise, Idaho. That's the historical perspective prior to this recent expansion. The recent expansion, we picked up a couple of new customers, one in Arizona that's pretty well known. We're doing the mechanical work there. We've done some of the fire life safety work there. We've done some low-voltage electrical work there. We did a lot of work on the first fab mechanically, fire life safety. We're doing some of the work on the second fab, the same scope going forward on the mechanical side. As you go to one of the other big fabs built in Texas, and we don't give customer name, they don't want us to give customer name. Another big fab built in Texas, we did the CUB, Central Utility Building. So just put -- is it a bigger mechanical project? It's a huge mechanical project. It's 70,000 tons of cooling. So just to put that in perspective for you, 70,000 tons of cooling, take a major medical center that may span 2 or 3 city blocks. 12.5x as much cooling power goes into that semiconductor fab at a place they're planning on building 5 that would be in that 1 location. So that just gives you an idea.

How many times?

12x of a major medical center. So go down to Northwestern Medical Center, you're in Chicago, looking at vast medical center, the cooling capacity that they will use at that one semiconductor fab that we built in Texas, just the central utility building, 12x the size of what they'll use at Northwestern Medical.

So 70x when it's all said and done, if this is...

70,000 tons is 12x. So it's about 5,000 to 6,000 tons at Northwestern Medical Center. This is 12x that. I'm just trying to size it for people. And there's other phases to that, right? We did that there. places, we'll do the tooling install, which is more of a unit price job. In other places, we'll do what's in between the tooling install, which actually moves the air and moves the liquids, and we do that. That's actually what we're doing out in the desert. You pick which part you're going to do, you really can't do it all, no contractor really. And they don't want one contractor to do it all. And so we think that's a 5- to 10-year opportunity, too. We're well entrenched in Arizona. We'll have the opportunity to potentially do the Texas job when they do the next phase. And then we have to balance that against other opportunities we may see in the same geographic market in Texas because there might be something better to do that we'll go do there, maybe a more remote data center that's going to take 7 years to do, and there'll be more quick-term projects to do that. And you say, why don't you have the capacity to do both? Well, we could have the capacity to do both, but we'll see if that's what we want to do. To the central question, though, do we believe this is a good long-term market for EMCOR? The answer to that is absolutely. Is it going to be a big drop into our RPOs like it was when we did started the first plant in Texas? It might not be. They might let it out in phases, and that's more what we're seeing in Arizona, and that's what we've seen through time as we became more entrenched with the customers. Jason, you had some good data on this, I'll flip it over to you to add.

Tim, I think the key here when you look at the data is the point you made about this being a little bit lumpy for us relative to some of the other sectors, right? You finish the first phase. There's an inherent lag before that second phase starts up as our customers try to get these facilities up and running. But if you just step back and you look at the work we're doing today in this space relative to what we were doing just a few years ago, let's say, 2021, for example, our revenue today is 5 to 6x in high-tech manufacturing, what it was just a couple of years ago. So we're down maybe year-to-date. But long term, we're certainly seeing elevated demand, and we think we can continue to grow off this space.

And it's really all trades. It will be everything from sprinkler fitters, fire alarm techs, low-voltage techs, traditional electrical, mechanical, and we won't get every job, every phase, every trade. And then millwright actually because one of the semiconductor equipment providers have actually hired one of our companies to be the national person that goes around and sets all the semiconductor equipment regardless of whose plant is going into. So this is a broad-based opportunity for us. It's something we will participate in. It can be a little lumpy or it will be steadily coming into things. We like the work. It's tough work. There's less people that can do this tough work. But again, when you're working in a sector like this, you're working for some of the most sophisticated customers in the world. And some of these folks came over and either hadn't built a semiconductor plant in a long time in the U.S. and one of them had never built one in the U.S. So you go through those growing pains. But to be fair, we did quite well on that first one with a new entrant in the market for the U.S.

Got you. That was a lot of good color. And when you say it's tough work, not everyone can do it, Tony. Is it your technical capabilities? Or is it the fact that you have the manpower? Which one is the bigger one at the moment that...

I think it's the intersection of both. The manpower are pretty highly skilled. I mean this is a lot of high-end welding, high-end fitting, electrical, this is pretty high-end work, right? These are very experienced people. In our case, we try to have at least a percentage of that 20% to 30% of those people have done this work before. We're not like cycling. The second thing that I think is the amount of interaction you have to do with the person building it. The amount of BIM, Building Information Modeling, virtual design construct, design assist we're doing here before we ever build anything, coupled with the amount of the precision on the prefabrication. And then if we move -- if we're doing the middle phase, the middle work or the tool installation, the middle phase is all clean room work. So we actually have basically the chambers to do clean room work like a pharmaceutical or anywhere else to do the welding and everything around that. So it's very specialized. The further you go into the fab. Central Utility Building, maybe not as much, but as we're more into the fab, it becomes very specialized, very -- you need some facilities to do it. You need training around those facilities, that kind of welding. You're doing a lot of high alloy welding, a lot of stainless steel welding, the further you move away from the central utility building.

Got you. Okay. If I stick on the other one that everyone thinks of when they think of high-tech manufacturing is battery plants. Is that also -- I can't remember exactly what you said about them in the past, but I think that it's something that you've mentioned in the past as being you playing in the EV value chain in the battery plants as being a growth area. Is that something that you're seeing steady growth in? Or is it pulling back at this point like we're seeing in some areas of the EV value chain? Or where are we at?

Let me tell you how we thought about the EV value chain. So maybe the cynic of me was proven right. So like we're doing -- at the macro level, we're doing both capital allocation and resource allocation all the time. And one could argue the resource allocation is actually harder than the capital allocation right now. They both take a lot of thought. The capital allocation is about money. The resource allocation is about people and fabrication facilities. It's about our -- and it's people all the way from the BIM designers all the way through the tradesmen on the ground installing the installation. So when the -- all these opportunities were coming out 3 years ago, data centers were starting to grow more, and we're going to talk more about that later. You had reshoring, you have our traditional markets like health care and institutional, you have EV chain. Now we were involved in some of the first electrical EV projects with Tesla in Reno, Nevada, especially on the millwright side and the fire life safety. And so we had some experience around that. We also have a lot of experience with the auto companies. And they can be very difficult customers and they can change their mind in an instant. So as we were looking at all these opportunities, we said, okay, how do we want to think about how we're going to participate in the EV value chain? Well, first of all, we always are thinking about how do we maximize margin and how do we maximize the opportunity for our people to have long term. And then is that going to be a customer that we're going to have multiple opportunities. So quite frankly, we did make decisions in some case to not be the go-to contractor with some of the Big 3 and some of the locations where they were going to build them out from nothing, and they wanted a commitment on dedicated resources over a 3- to 5-year period. Quite frankly, in some of those cases, we were not the guy, and that was by design. But we did participate broadly with fire life safety. We did participate probably on specific scopes. And we did okay in EV value chain of all the places that we're growing and continue to grow. This was the one we emphasized the least, but it's still a substantial part of what we did. What we see today is I personally believe there's going to be a lot of battery plants in the U.S. because I think a lot of people are going to follow the Toyota model. I mean, remember, we're one of the biggest fleet operators, too. So we never understood sort of that we're going to be all electric, can't carry the way. We did a lot of work around that. But we do know hybrids work. And we know that, that could be the solution for the future. And so we think there will be a lot of batteries built, and we are participating in that. And especially both fire life safety, mechanical and electrically, we've been involved in quite a few of those projects, and we'll continue to be involved. Is there as much demand today in the EV space as there was 2 years ago? No. But is there a steady demand in the battery space? The answer to that. Jason, maybe give some numbers around that.

I think Tony hit it. We participated in a very select way. What we're starting to see now though is a shift away from full EVs and towards more battery. But when you look at it on a combined basis, there's still growth there. I think right now, battery is the demand driver, but we're still up 10% year-to-date in the EV and battery space overall.

Yes, it could be interesting in the future, too. And I actually want to ask you about the power infrastructure later on, but maybe batteries are going to be used for more than just EVs and...

Yes, we've done -- Tim, we get to talk about, we've done big battery, like not where they're producing them, but where they're using them for power infrastructure. I think that has a place, but that's not going to power -- with the amount of power we're talking about here. That's -- if you did a 1,000-piece jigsaw puzzle, that might be 3 of the pieces.

Got you. Got you. Well, maybe we could just move to that. I'm going to skip a few of these. Well, I did want to ask you about life sciences and if you think it's that whole thing. And if you think it's important, please skip back to that. But I'm going to skip that so that we have enough time.

Yes, I think I can answer that really quickly for you 2 seconds. Long-term market for EMCOR, actually a long-term player in the aftermarket for EMCOR. We have people at these facilities all the time, electricians, pipe fitters doing [indiscernible] moves and changes. We also are there when they put in major lines, and that's across all trains. Good long-term market has a lot of the same characteristics of other good long-term manufacturing markets. Sometimes they grow 1% or 2%, sometimes they grow 12%.

Got you. And this is like GLP-1 manufacturing stuff.

Yes, we're doing -- supporting that. Yes, in a couple of places. Yes. And also life sciences and some of the BioMed and we're you're supposed to be. We're in Southern California, we're in Indiana doing it. We're a Research Triangle Park, and we're in New Jersey doing it and some of the primary research we're doing supporting the facilities up in Boston.

Got you. Okay. That's helpful. And yes, so lastly, on the reshoring theme before we get to more of the AI infrastructure. Well, this kind of is AI infrastructure with the power because I've always wanted to ask you this, and I did it during my due diligence on your company, is that I really wanted to talk to you about was the power infrastructure build-out that needs to happen in the U.S. And there's 2 conversations to have here with you. And one is about the potential bottleneck that's coming to AI infrastructure from it. I'm going to set that aside for a second. And I want to actually ask about do you play here? Like do you -- does EMCOR build power plants? Do you build export terminals? Is there an opportunity for you to participate in this "unleashing America's energy independence" that we see happening right now?

So some of the investors that are on the call now, I'm sure, have heard me cynically talk about over a period of time have heard me cynically talk about the way we were actually approaching this problem in the last 4 years, right? Because you're not going to do this with intermittent power. I mean this is -- it was a fallacy. All of the above was silly. I mean it's not -- yes, it could be part of the solution. Go to your battery point earlier, it's not the solution. But EMCOR does play. I mean we have built utility-scale solar farms. We're contractors, right? We're going to go to where we think we can mix the labor the best, get the best opportunity, build capability if we think it's a long-term market. We do have utility scale solar experience. Secondarily, we also participate in the aftermarket in power plants. We have a business that's called PPM. It's part of our -- it's part of -- it's in the mechanical business. It's part of our industrial platform in the Southeast, not oil and gas platform, but industrial, that PPM stand for power plant maintenance. That's what they did. They can do coal plants, gas plant they used to do nuclear, we sort of moved away from that. But -- and so we do the aftermarket. We also participate upstream. We are emphatically not an EPC power plant producer. We're not going to guarantee someone output. We're not going to design build a large-scale power plant. We may do a combined life cycle cogen plant at a hospital where they're going to take the waste heat off of a turbine and turn it into cooling through a chiller. We will be involved in that. But -- and we may design build that with a partner. So any large-scale utility scale power combined cycle plant, that's not us. However, we do participate in the balance of plant work, both mechanically and electrically. We will do some prefabrication around it if we're going to install it. When the last power with the IPPs happened in the sort of the '05 to '08 in California, we had some pretty good success on balance of plant work mechanically and electrically. We do substation work selectively around the country. So we will be part of that. But is it going to be the major driver like some of these other things? Probably not. But we do participate. And it's like any other job a contractor thinks about, is this the best opportunity I have to execute in the next 12 to 24 months? And where that's the case, we will execute for our customers. Now, linking back to this AI infrastructure, we're going to have to build, right? Combined cycle plants are sold out from the big 3 to 2031. So a lot of gas plants are going to go into the next 5, 6 years. And so what they've done to fill that gap in between the time, and again, if you listen to me over a long period of time, I said 5, 4 years ago, we were going to go on a quest for stranded power. And they are now on their quest for stranded power. We could talk a lot more about that when we talk about data centers.

Yes, I would love to dig into that because that's been a big theme here with my partner, Jed Dorsheimer, who covers a lot of that stuff and talks about the power infrastructure bottleneck a lot and the mistake that we made going after intermittent power versus baseload power. I think you guys were friends actually. So yes, I would love to dig into that more. But now that we're kind of on the data center theme, maybe we could just dive in. I think you've actually communicated really well about this market over the last few years. So we don't have to spend a ton of time here. But for those on the call that aren't as familiar, and that kind of includes me. Again, I'm newer to the story. Could you provide a brief history of EMCOR's footprint in the data center market, how you -- and I'm thinking about this like how your capabilities have evolved over time, but also how you've expanded geographically over time because you asked novice, right? And I'm like an electrician as an electrician, as an electrician. I know that's not true, but I don't know what the differences are. Can you pull someone up over here, put them in a data center. When you say I have a data center market or data center capabilities or we're establishing that, I don't know what that entails. I was curious to find out.

So we have a long history in data centers, going back to the first data centers that were built. I'm here 21 years. I was with Carrier before that. The first data centers that really mattered were built in the late '90s, early 2000s. We actually -- I helped sell the job as an air conditioning guy to an EMCOR company called Poole & Kent down in D.C. that did the original Equinix data centers and the original AOL data centers. So that's all. And then the electrical contractor on that job, I wasn't part of EMCOR was actually Dyna Electric DC. And they were sort of some of the first data centers that were at scale, and those data centers were between 5 and 10 megawatts. So today, a hyperscaler doing AI is doing 200 megawatts. They're 20x the power need to do an AI data center is the first AOL data centers and a cloud storage is about 5x that size. So just we're sizing these right now, 200 megawatts, give or take, could run a 15,000 to 20,000 person city, depending on the city or 5,000 households, you figure 2 to 4 people per household. And it would take -- go back to the PowerPoint, it take about 1,500 acres to get the same output solar wise in 600,000 panels, unless it's the new ones, it will be about 400,000 to get what combined cycle plant can get at 250 megawatts, 1 turbine and usually put for. So it just sort of sizes it for you. So EMCOR has got lots of capability here. So we went from there. Then we went to started building when I first got here, just the big data centers that people like the Bank One before it got acquired by JPMorgan built the 2 biggest financial institution data centers in Chicago and in New Jersey. We built those. And then sort of we were just building them for owners, right? Bank of America, we're building them, mainly financial institutions, some utilities, some REITs were building data centers. And that sort of happened for about 10 years, and we were building them then. Things started to change around 2017 and '18. Cloud storage became a thing. Amazon started building data centers. Google started building data centers. They all started -- and they were mainly building them in either -- for us, they were building in Chicago, where we had the first data center builders, and they were building in Portland where we built capability, and it all started with our group in D.C., in New Jersey, and they taught these guys in Chicago and these guys in Portland how to do data centers. So we were maybe doing until about 2020, 2019, 4 data center markets, data center sites. [indiscernible] market as sort of a big city region or a state. And then we flip to now today, we're doing 17, Lucas, electrically?

Yes, 17.

We're servicing 17 different distinct data center markets today and versus 4 in 2020, 2019 and versus 2 or 3 mechanically, we're now servicing 7. And fire life safety, we can service every data center market. And so what does it take to go and get into a data center market if you're EMCOR.

Yes, what's it take?

So how did we do this? And I'm going to give you 3 distinct examples. okay? One is get bigger, right? You had great capability in D.C., you had it in Chicago. They're the founders of the EMCOR in New Jersey. They're the founders of the EMCOR data center business, take those people, teach other people in those places, just get bigger, get more capable, serve your customers better, attract more customers because these are big data center markets. The reason for that is they have great connectivity, they got power, right? So just build scale, right? So the D.C. business is 4x the size it was 5 years ago. And they did that also by growing their traditional business and then growing their data center business quite more than our traditional business. And did those first 3 or 4 markets, did that in D.C., did that in Chicago, did that in Portland, Oregon, out through Southern Washington, okay? Then, [indiscernible] how are you going to serve more customers want you to do more, you want to do more. You're talking to your customers. So okay, now we got to figure out how to go to Columbus, Ohio. We have no capability in Columbus, Ohio mechanically and had no capability really electrically. We do an acquisition electrically of a traditional electrical contractor, bought a smaller one that had some day 2 work in data centers. Now we're in the data center business with great industrial electricians that maybe we can teach to do data center deforming versus just traditional electrical work. So now we're growing in Columbus, Ohio. Mechanically, what we did is we took 2 of our best companies, one through acquisition, Bachelor and Kimball, who had big data center experience. That will be the second way we do it. And Poole & Kent, they formed an internal joint venture, went up and formed a company up the mechanical the service at the request of one of our customers on a GMP contract, which gave us both some protection to go up and do mechanical scope in Columbus, Ohio region, to Albany, Columbus all the way up. The second way we do it is take one of our existing companies like we did with Portland at one time and teach them how to do it, right, is we take one of our existing companies in Indiana, 2 places in Indiana and say, you guys are great electrical companies. You've got great capability. You do automotive work, you do steel mill work. We're going to teach you how to do data center work. So how do you do that? Well, you bring some folks over from Chicago, you bring people from all over the country to help them estimate the job. You're going to be working for a customer you already know. You can bring the means and methods. You take them out to a site you built. We got like, okay, we can do that. This is what the prefab plan looks like. Here's what the learning curve looks like. So instead of the learning curve being like this slow linear growth, the learning curve looks like this, right? Because we take some experienced people, intermix them with some of their just great foreman and superintendents and project managers after half of a build, they know what to do. And then, of course, this peer group we have is sharing all the time. The third way we do it is acquisition. And it might be of a small company that's sort of in the data center business, mainly doing scope for somebody else and we say, "Hey, you have some resume, you now have our resume, and we're going to basically treat them just like we treat one of our other companies that we've owned for a long time and say, we're going to put you in a data center business in a major way." And it's going to take us a year or 2 to do that because now we got to know -- unlike our current company that we've owned, we know what their capabilities are. We bought this. We think we know what their capabilities are. We're going to be a little more careful when we do that and then to ramp them up.

How do you do that if they -- if you're not sure, is it just you make sure that there's protections for you and the customer in the contract?

Well, we're going to be sure because we're going to put the right supervision on the job.

That's how it is, yes.

But we're going to be sure. The customers buy -- no, the customer is not going to give us a pass. We're going to take the right contract structure, but the customer is not going to give us a pass because no, I hired EMCOR, an EMCOR company. You guys know what you're doing. And then you buy people that are the best in the business like Batchelor & Kimball and knew how to -- we bought in 2019. And they then helped our mechanical business grow quite a bit in the data center business because they were known in the Southeast and in Oklahoma as one of the best data center builders in the country. And then on the wrap around all that, that's electrically and mechanically and fire life safety between the assets we have at Shambaugh and Comunale, they can service any data center market in the country, and we've built those through acquisition and mainly organic growth.

Got you. Okay. Sticking on this data center idea, sorry, I'm not distracted. I just -- there are people on the call that are e-mailing me left and right saying, "Hey, could you just squeeze this question in," and it was distracting me. They're all related to data center, by the way. But I'm assuming you don't comment on competitors. So I'm not going to ask half of them.

No, I don't comment on competitors.

So they all want to know how you think you compare relative to...

I saw in some of the notes or somebody's note that Comfort did X and we did Y and Comfort is a good company. Because they grew faster, doesn't mean they took share from us, right? These markets are growing in different markets. I don't -- like we've never thought of our business in terms of market share. It's just a general comment.

All the time in other markets as well, and it doesn't make -- especially when you're...

It's an irrelevant metric.

Yes. Yes, you guys both are performing at a high level on the data center front. I mean, I think people do think about you, too, as the leaders in the space with robust capabilities to serve these markets.

Yes. And we're very rarely in the same market, by the way.

Right. I mean the way I think about EMCOR, and please correct me if I'm wrong, is you guys are in a lot of Tier 1 cities and you have a unionized labor force and Comfort Systems thinks about themselves as in Tier 2 cities with a nonunionized labor force, you're not going to cross paths, but how would you correct with that?

I think for general work, that's true. But the more you go to some of these data center markets and the more you're in -- some of this manufacturing work that no longer holds true. But very rarely are we overlapping on a job. I mean we're competing for the same work. I think it just happens that way. I don't I think it's a big market. And different customers.

How about this liquid versus air? We have -- we've read that AI data centers with liquid cooling requires even more MEP content for data center for cloud. So that's true.

That's true.

Okay. I wasn't -- I mean it makes sense to...

Very much -- a lot more tonnage, more megawatts, drives more tonnage to cool, drives -- and the chips are more hot. So therefore, they're trying to -- they're still blowing air across the servers, Sure, right? That's where the Vertiv products and all those people come in, right? And there's custom air handlers, and that's all the chillers are lined up, and these are massive chiller plants, either air cooled or water cooled. And then what liquid cooling really means. So a guy like me thinks water cooled versus air cooled. Water cooled means that's just the loop that you're using to cool. What this actually means is you're taking the server, you're putting it in a jacket and you're doing heat exchange from either some liquid medium, you're going to either run a pipe assembly down and run it through there and cool that water that way that the server jackets in or you're going to do it through other medium. I mean that's what liquid cooling means here. And it's usually a fabricated product that you're putting cool water through -- or some combination of cold water or glycol or something in to extract the heat transfer out of the liquid medium that the server jackets are in.

Oh, really. So it's [indiscernible] liquid.

Just think about like a tea bag. And think about the server being the tea bag. You're dipping the server in and maybe it's not just water all around it, but there's a jacket and then there is some kind of piping system around that jacket to extract the heat out and then take it out and expel it through the loop usually through a cooling tower outside. That's what it means.

That's what that means. So that's part A is, is it more content? And the answer is yes. And part B is, well, where are we on that journey? Because I think about it like probably inning 1 on the liquid cooled side. And please, again, I don't know if I'm talking about, Tony. Correct me when I'm wrong.

I think we're on inning 1 or 2 at best in building out the AI infrastructure based on the work I know we're doing. And the mix in our thing. I think we probably look at the same industry studies, you guys probably even created a couple of them. What we've seen is sort of cloud storage is going to grow CAGR over the next 5 years, 9% to 11% and AI is going to grow north of 20%, which gets you to a mid-teens growth rate, right? I still think the majority of our backlog today is still cloud storage. Jason, you got -- you have some facts and figures around that.

Part of the growth you're seeing, if you look at our Mechanical segment versus our Electrical segment. electrically, we're still doing more data center work. Our revenues are still greater on the data center side in our Electrical segment than it is mechanical. Even dollar-wise, electrical is still growing faster and some of that's just a larger base. But if you look at growth rate, you're seeing a stronger growth rate in mechanical. Just for example, year-to-date, electrical is up 80%, mechanical is up almost 120%. Some of that is this increased cooling content. Some of it's just a lower base in mechanical. But we're starting to see that in the numbers.

Okay. So it's starting to show up in the numbers. That's interesting. Do the AI-dedicated data centers require additional capabilities on your end? Like does it require more prefab, more EDC?

It will -- yes, there's more prefab. There's more fire life safety content. There's more voltage coming in, right? So there's more megawatts coming in, so there's going to be more electrical feeds, bigger switchgears. All those things are true. I think about it simply, an AI multiplier right now is about 1.2 on electrical versus traditional cloud storage and maybe 1.4 to 1.8 mechanically depending on design. But go back to your PowerPoint, I'm going to give you an interesting figure here. So we're doing a very large data center campus for somebody down in Georgia. And we'll probably do the whole campus mechanically. We're not there electrically. But the data center campus will be 3 gigawatts. That's 3,000 megawatts. So think about the nuclear power plant they just built in Georgia. I think it was 3,500 to 4,000 megawatts or 3.5 to 4 gig. That one data center requires the power -- 75% of the power of that first nuclear plant we've built in the country in the last 25 years. So...

That's ridiculous.

In perspective, that's one data center campus in Georgia that's being built today.

Well, okay. So let's talk about this, the energy wall, as we call it, the energy bottleneck. We just don't have enough electrons for these things to run on. I'm curious about your thoughts on this dynamic and when something like that may occur. And people think about this. I actually don't think about a DeepSeek moment as being a risk to your data center backlog because if we look historically, there's this thing called Jevons Paradox, where if you increase like the miles per gallon of a car, it doesn't mean we're going to use less gasoline. We're actually going to use more gasoline because it's more efficient, right? So like we just drive more when we get those efficiencies. And if we get efficiencies in data center, we'll probably just build out more data center stuff.

Here's all I know, right? I think you heard me talk about this on the earnings call.

Please.

Here's what I know, okay? About 7 weeks ago -- and look, we follow this closely, and I read everything I can. And I think these are some of the best planners, the best they can be. And they have some of the most technical people in the world on both sides, both on the utility side, thinking about how they're going to supply and on the data center, how they're going to use, right, and how this all works plays out and where they're going for it. So the big picture, what I personally don't believe is data center operators are going to put on-site power to power their data centers as a major part of what they're going to do. You never say they're not going to do it. Is it going to be 10% of the market? Is it going to be 15% of the market? My view, and I could be deadly wrong on this, I don't think it's the majority of the market. I think because the minute you do that, you absolve the utility of any responsibility of supplying you power when you need it. So I think that's a leverage point. I think they will do it in some cases, in very remote areas. But in general, they'd much rather the utility be on the hook to supply them the electron from other places and guarantee them power. So then you get to -- I also think that because they went to these stranded power places, there's no reason -- there's a reason why they're working up on in Indiana on the shore of Lake Michigan because what used to be there, steel mills, car plants, big industrial infrastructure. There was a lot of power there. And in places like Indiana and Ohio, they're going to extend the life of coal plants and they can build gas plants and they're going to extend the life of the nuclear plants that they have, right? So they're going Iowa, right? Iowa has wind power, but they also have a big fossil infrastructure. Oklahoma, Texas. I mean, intermittents make more sense in Oklahoma and Texas than they do in most places because they got wind. And then they have to put the fossil with it, right? So all that stuff fits together. So what we've learned from our customers is they feel really good for the next 5 years, 2025 to 2031, 2030. Yes, is there interconnect problems? Can the utility keep up? All that's going to be true, but they're not putting these major projects ahead where they don't think they can get power. I mean there'll be maybe little blips, but they -- if we don't have the ability to put the next round of power in with all these orders for the gas turbines, then we're going to have a problem. And as far as nuclear, I mean, again, I have the ability to talk to more utility executives than not, right? And I have a couple I know very well. That is a -- they're looking at it. And modular nuclear maybe is the future. But the reality is that's still 5 to 10 years out, and we're not putting one in your neighborhood. They're going to be built in centralized plants, maybe modular in nature, and they're still going to be going out over the same -- they're going to put -- it's going to be easier to put T&D lines along existing T&D lines than all the stuff we're doing with intermittent energy. So we're going to have to figure out -- and all that's going to play a role. But I think most of what's going to get the next increment from 2030 to 2033 is going to be gas turbines, right? And that's a known technology that works and we have lots of natural gas. And people in states that aren't going to use natural gas are going to be left behind in this AI race because they're not going to have the latency in the data centers they need to be able to use the models where the knowledge workers are. And there are some major places that could have an issue with that. So I think that, that is going to happen. Now I will tell you, though, a couple of -- about a month ago, we had our 80 or so top data center builders, ours, EMCOR together in one place, down actually Miller Electric, our new acquisition. We had us all together there. And we were talking about means and methods and projections. And our customers found out we were all going to be together. And they invited themselves or 4 of the big people that you would think would be there, can we come and present to you our plans because we want you to know how important you are. We are -- you are to us and our ability to get things done. So through the general contractor to us to make sure -- and their plans were quite remarkable of what they plan to build. And...

Does that ever happened before where you have the end customer like that asking to come in? Or is that quite rare?

Well, I've not seen it happen too often. I mean that's pretty rare. It happens all the time at the local level where they'll say, okay, we got this job. We're going to work on it. We want you to be our team. We want you guys to think about this. But to happen on a collective basis like that. And to me, the thing is to watch these major capital spenders and the person in charge of their capital spending, knowing 50 of my 80 guys in that room by their first name. That's a little bit rare. So the next [indiscernible] well, then you guys should be able to get more margin. Well, that's not how it works either. I mean, because sometimes there's contracting mechanisms might not be fixed price initially. We may ask for a GMP contract. They may want to do everything GMP, which is a target type price we work to because it makes change orders easier and they do change or they want to have transparency of cost. And so when you think about this data center space, there's 3 or 4 different contracting mechanisms. There's 10 or 12 at least major customers, maybe more once you bring the colos in. There's, at least in our case, 17 to 24 geographies mechanically and electrically are working in, even more than that. And so they're balancing risk reward, certainty, we've got a fixed price. We may not want to give them that in certain markets because the labor force is going to be. They may not want to give us in certain markets because they don't want to get in a contentious relationship on a new build as they get ready to build out the site because they want us to be part of the team. Other places, we've had to come in and finish someone else's job, and that has a whole other thing to it. I mean -- so this is a very -- it's like 4D chess, but yes, it is very rare when they'll come in and share their plans with you. They may share the next 2 jobs with you, but for them to talk about, hey, this is what it looks like for the next 5 years, and we didn't have all in there together. We came one after another. But -- and it's respectful. I mean it's sort of like, hey, we're sort of in this together now folks and how are we going to build this infrastructure because we right now can't keep up with the demand from the people that are selling our product on the supply -- we're the factory. We can't build the factories fast enough to keep up with the supply side. I mean, with the demand side.

That's why they're doing it. They need to know that you're going to be there for them or they want to know that you're going to be there for them. That's why they're coming. They're breaking through the general contractor and they're coming direct to you. They just need to know it. It's too important. not to risk it. Yes. That's really interesting. Look, we got 5 or so minutes left, and I got through about 10% of the things I wanted to talk to you about, but this was an awesome conversation. And I'm glad that we did this instead because I learned a ton just focusing on some of your main markets. But if I wasn't to spend the last 5 minutes, I'd probably move to labor because it's a question I often get. And the reasoning it goes as follows. My question, the way I set it up here because I prewrote it, it says, where would you characterize labor shortages as being most acute? I've heard you say in the past that it's the leadership roles, the foreman and the superintendents and the project managers, and I get that, and there's a time component to developing that, that you can't rush. But if you Google electrician shortage or plumber shortage or welder shortage. I mean you'll get a dozen articles from the trade associations representing those folks arguing that there is a major shortage. So are you really not seeing a shortage there across those skilled trades folks?

Of course, there's a shortage, right, if you take the macro level overall. I think it's purely a tribute to our teams in the field. So you think about EMCOR and our operating model, right? The people that run our companies know trade labor. In fact, over half of them are run by people that started as apprentice electricians or pipe fitters.

That's Right. It's like 60% of mechanical...

The other half of project engineers and spent their time around it. And that's both -- and if you think about on this big work we're talking about, for the most part, we're doing most of this big work union, which I look at as a real strength. The unions, the IBW and the UA have been partners with us in this and helping us develop our workforce. And that goes all the way from training, which we play a big role in, all the way through the workforce development. And we're usually one of the biggest employers in whatever geographic region we're in. So we're partners in the development of that workforce. We also have -- because of who our people are and it's a trades-based company. It's not, for lack of a better word, it's not a bunch of suits running the companies at the local level. They're less like me and more like the highly skilled people that do the work instead of the overhead like me, right? And so they know where to find it. They know where to develop it. They know how to work with the union to get the different classifications to get their blended skill mix on a job. And in most of these fast-growth markets, we have that ability to work with the unions to get the right classification. So we're not all just bringing in apprentices on the way to Journeyman. We have wiremen, we have CWs, we have CEs, we have all those things. And then we put capital to work to think about how, think about that 2.5 to 3:1 ratio of revenue growth versus labor, and Jason can talk about that in a minute. And we're attacking it with capital. This year, over an 18-month period, we're going to add more than 400,000 feet of fabrication space to get more work out of the field, prefab, modular, whatever you want to call it. And if you think about -- if you just boil it down what a trade guy cares about, and people have heard me say this repetitively for 15, 20 years. First thing they care about is there no particular order. Am I going to get paid every week? And if I'm a union trade guy, is my benefit fund going to get funded, too? That's the health care, the pension, the annuity, whatever it is. And with EMCOR, that's a check. The second thing is, are you going to keep me safe? And again, these aren't in any order. In the 21 years I've been either the President or CEO of this company, I've never turned down one request for safety equipment, safety capital spending, anything. Guys in it, we have the attitude, the folks in the field know best what they need to get productive and keep their work for. We're not trying to buy the cheapest helmets. We don't have Accenture coming in here and doing a supply chain management study on. You can have everybody wear the same helmet and look, you can buy them and they're cheaper. That is not where the leverage is in this organization. It's buying the best stuff for that region, what they need to be successful. That could be something as simple as the right gloves to do the task.

I think you have pretty low incidents.

We have a Six Sigma type safety record versus the industry. The third one is, am I working for people to know what the hell they're doing. And they know what I have to do to get that job done, and they're going to keep me productive. Fourth one is, if I do a good job, can I become part of your permanent team and continue to grow my career with you? And if I want to grow my career, can I become a foreman and maybe someday I can even be the CEO of this company at the local level. Because you know what, they have an example sitting right there that started with -- started the same place they did. Jason, maybe you can real quick before we end the capital spending we've done, it's different versus historically. Still tiny, but...

How do we attract labor and then what we're doing to become more productive, more efficient to help with that tight labor market. And so Tony said it, right? We're not a capital-intensive company. For us, CapEx is somewhere between 0.5% or 0.6% of revenues. But if you look over time, we've made a ton of investment in prefabrication and construction technologies. And if you just look at our CapEx, let's say, take a 3-year window, right? Our revenue CAGR over that 3-year window is something like 14%, but our CapEx CAGR is 2x that at roughly 28%. And those are those investments that we're making in prefab, BDC, BIM, things that make us more efficient, things that allow us to do more with less and the investments that have allowed us to grow revenue at 3x our headcount growth over time.

Yes. Okay. So maybe it was a little bit of a dumb question. Of course, there's a labor shortage. That's not the issue. It's -- these are the reasons that you guys aren't struggling to find labor because of the investments that you made, because of your union relationships, because of your safety record, because of the promotional opportunities, because of all the things is the reason where you're able to still grow your headcount, not nearly as fast as your revenue, but you're growing your headcount every...

And I'll leave the other point, right? No good specialty trade contractor in a good market chases every opportunity that's out there.

Right.

So the second point is we're pretty good about thinking on labor planning on what we need -- we're not going to outrun our headlights. We're not going to commit to a customer and say, we can do that unless we thought through what that labor plan looks like. So like they said, we've been growing revenue CAGR of 14% probably, what, 2/3 of that is organic or more, Jason? Yes.

Yes. Probably slightly more, Tony.

And so slightly more, right, maybe 75%. And then we're able to -- that's all organic headcount that comes with that and everything else. Our guys are thinking about workforce development every day. And then we reinforce that with our leadership training programs all the way from frontline supervision to CEO. And we've been doing that a long time.

I think that's probably a perfect place to end it. I wish that we had another hour to dig into more of these topics, but I can't thank you enough for the time today.

Thank you.

This was a great conversation. I'm very excited to be covering the company and excited to learn more along the way and look forward to talking to you guys when you report your fourth quarter here in a couple of months.

Yes. And we very much appreciate. We like that you guys are covering us. And anybody that's on the call, thanks for your interest in EMCOR today. We're pretty proud of the company and the folks we have working at it.

That's perfect. Thanks so much, Tony. Thanks, Jason. Thanks, Lucas, and everybody else. Have a good day.