Podcast: AI's boom creates manufacturing opportunities with data center equipment

Typically, Big Tech and software aren't manufacturing topics. The latest data products don't generally come out of a factory in a box or barrel. However, the rapid growth of AI is having a domino effect that's benefitting large manufacturers. 

AI needs lots of computing power, creating opportunities for semiconductor makers to boost U.S. production. Construction companies are building data center facilities all over the country, creating manufacturing opportunities for construction materials and equipment. Tightly packed computer servers generate lots of heat, creating opportunities for materials companies and hydraulics manufacturers that can cool data centers. 

And these data centers are super hungry for power. So, expect more gas-powered and renewable energy plants to be built in the near future (it's already happening but expect more). Makers of switchgear and grid-improvement systems also have full order books.

Discussing this trend are four editors from Endeavor B2B:

• Robert Schoenberger, editor-in-chief of IndustryWeek
• Matt Vincent, editor-in-chief of Data Center Frontier
• John Caulfield, senior editor of Building Design + Construction
• James Morra, senior editor of Electronic Design

Below is an excerpt from the podcast:

RS: Hello, I’m Robert Schoenberger, editor-in-chief of IndustryWeek, and welcome to Great Question: A Manufacturing Podcast, where today we’ll be talking about data centers. Yes, they don’t really make anything in data centers, so it’s not a manufacturing topic, but the growth of AI and the growth of data centers nationwide is driving a boom in certain sectors of the manufacturing world. Really, if you want to look for where manufacturing is strongest right now, you look at people making electric switch gear because of the power demands of the data center industry or the people making any kind of networking or connection equipment or the microchips that power these new AI systems. There are a lot of different pieces to that. And to talk about it today, we have a handful of editors from Endeavor B2B, IndustryWeek’s parent company. We have Matt Vinson from Data Center Frontier. We have James Morra from Electronic Design and John Caulfield from Building Design + Construction. Welcome, everyone, and thanks for taking the time today.

So Matt, if we could start with you, you deal with data centers on a daily basis. What is driving this industry right now and how big is the growth trend right now?

MV: I mean, Robert, AI is overwhelmingly driving the industry ever since everybody knows about the ChatGPT inflection point that happened, I think, at the very tail end of 2022. And since then, it’s just been a hockey stick growth curve in terms of, you know, projected demand, actual demand and investment coming online. And, a lot of anticipation. So, the interesting thing, though, why I thought this was a really interesting conversation is that the computing boom definitely appears to be redrawing the manufacturing map. And, you know, the hyperscalers are just spending so much money now in their expansion.

And of course, that bleeds over into, you know, it’s the whole scale and the co-location data center operators are absorbing some of that load. That’s why it’s, you know, record low vacancy. I think it’s like 1.7% vacancy for, you know, trying to get your data center workloads taken care of in any kind of data center environment. But, you know, behind every hyperscale campus development announcement you hear is an order book. You know, it requires chillers, racks, modules, power gear, and this all is related to that data center manufacturing supply chain.

So, you know, I don’t want to go on too long right off the bat, but there definitely, you know, we’re in a data center tsunami and there definitely seems to be a manufacturing echo. With that, you need look no further than the news this week. I mean, three major acquisitions between Eaton acquiring Boyd to become a full stack, full engineering stack data center provider. Then we saw Vertiv acquiring [PurgeRite], you know, much less monetary deal, but no less strategic. allowing them to, Vertiv to tighten up their liquid cooling supply chain. And then we also saw Daikin acquiring Chilldyne, which was a big move for them. So I’ll let the others chime in, but yeah, it’s this data center boom story is definitely also a huge manufacturing story.

RS: Just a technical question for you, for those of us who aren’t in the data center world. When you say hyperscaler, can you give us a quick definition there?

MV: Sure, the hyperscalers are, you know, the big four or five, or, you know, the, you know, I mean, it’s, we’re usually talking about... Google, Microsoft, Meta, AWS, I mean, Oracle now has to be included in that hyperscaler list, what with all the data center AI campus development, we’re seeing them involved with, you know, in the context of OpenAI and the Stargate campus build down in Texas. So, yeah, when we’re talking about the hyperscalers, we’re talking about, you know, the largest data center providers globally. And it’s usually those companies that you see on CNBC, Everyday, or Bloomberg now talking about data centers. You know, it’s, you know, the Big Four or the Mag 7, those type of operators.

RS: John, you’re seeing this on the construction side. Can you tell us a little bit about some of the construction impact you’ve seen from the data center growth?

JC: Just to preface on my comments, I’m currently working on an Outlook 26 story, and I’ve interviewed about 15 architectural engineering and construction firms, and to a person, they all see data centers as they’re the leading sector for them going forward, at least for the next couple of years. There’s $20 billion in data center construction currently underway. And there’s been some estimates that the global spending on data centers could reach 7 trillion by 2030.

So I, this is, it’s interesting too, because this coincides with two things. One is that the industrial practices of these firms is starting to cool a little bit. So data centers is kind of taking up the slack in terms of their business strategy. And coincidentally as well, a lot of these firms are going all in on AI. And they need data center power to achieve their goals. And I’m talking about the big guys, Gensler, Turner, Gord Corgan. You know, these guys are seriously seeing AI as a major piece of their puzzle. And that only works if you have enough juice, and so there’s a lot of construction going on right now. There’s another factor involved too. It isn’t just data centers that’s driving it. It’s the infrastructure that’s powering these data centers. So you have building going on for renewables, for nuclear, Burns and McDonnell signed an agreement last year with a technology firm to help them design small nuclear reactors. And so it’s a factor on the construction side, both for data centers and for the grids that are powering them.

MV: That’s such a good point. I mean, the behind the meter power is, I knew we’d get into it before, but that’s been such a huge trend in the data center space this year, on-site powering or behind the meter. And yeah, there’s a lot of anticipation for John. You mentioned SMR nuclear, but really on the ground, it’s coming out in the form of natural gas. There’s a tremendous amount of development for, and of course, you know, that leads to more manufacturing because What do they need for natural gas on-site powering, air derivative turbines? What’s the bottleneck right now if you talk to people in the data center industry? Turbines. And you also mentioned about the other forms of powering gear, you know, down to the transformers, switch gear, all that kind of stuff that goes into grid powering of data centers is, you know.

RS: Schneider Electric just announced a big plan, a plant expansion to do more switch gear for data centers, for the power needs, for the grid improvements needed for this. I mean, there are several secondary effects from this also for manufacturers. And then in the short term, the higher load on the grid means electricity prices, spot prices have been going up in a lot of places in the country where these data centers have been coming in. So we’re seeing some consumer pricing increases for electricity. We’re seeing definitely industrial pricing increases for electricity. So on one hand, this is creating opportunity. On the other hand, it’s raising some cost for a few individual manufacturers out there.

JC: This was a factor, Robert, in the governor race in New Jersey, that one of the platforms for both of the candidates was the high, the 22% spike in electricity costs last summer that happened in the state. And they, of course, left out the fact that the ring, what was driving that was this data center tsunami, as Matt put it.

RS: James, let’s get over to you on the electronic design side. This is also creating some US manufacturing of specific microchips for the first time. Can you talk a little bit about that?

JM: Just a couple of weeks ago, Nvidia made, Nvidia being, obviously at the center of a lot of what we’re talking about with their AI GPUs. They made an announcement with TSMC. They’re the world’s largest and most advanced contract chip maker. And they’ve been expanding very aggressively in the US, partly driven by geopolitics in Taiwan where they’re based, partly based on the administration’s push in the US to reshore a lot of microchip manufacturing. And it also relates to what’s happening with the tariff situation, being able to manufacture more components inside the United States ends up reducing the tariff burden for companies that are building equipment with those chips or designing the chips themselves.

So Nvidia and TSMC made an announcement a couple of weeks ago that the first GPU based on Nvidia’s latest Blackwell architecture, the most advanced one they currently sell, has been the first wafer of chips, which is, you know, the round disk of semiconductors before they’re sliced up into square processors. The first wafer was manufactured in the United States. And so that sort of is a big milestone for both companies, both for TSMC to show that they’re making a lot of progress with their technology capabilities in the US. And it was also an opportunity for Nvidia to show that it’s focused on sort of rearranging its supply chains to better fit just its priorities.

RS: You mentioned the administration, that’s actually a couple of administrations now because the Biden administration was really pushing the CHIPS Act, which spurred a lot of semiconductor investment, including a lot of money to Samsung in Austin and TSMC in Arizona. Both of those companies have been and trying to push further into this AI data center space. The Intel plant in Ohio, that was one of the big wins, has actually been delayed and doesn’t look like it’s going to go online anytime soon. But those existing facilities that were able to use that money to expand rapidly seem to really be taking advantage of this moment in time to grow very, very rapidly.

MV: Yeah, and leading to the rise of, reinvigorating regional manufacturing hubs here in North America. you mentioned all the areas that we’re mentioning, but, we could kind of go right down the list of deals that happened between, stuff happening in Virginia and, the Southeast and stuff. Great points.

JM: And I just wanted to mention too that, there are obviously a lot of different forces that are pushing semiconductor companies to manufacture more in the US, including from, the CHIPS Act incentives and the tariff situation. But the growth of AI, I think, is really making it, is really creating a compelling financial incentive for them to do that as well. It’s not just the fact they’re getting these incentives from the US government and, from customers that want to maybe reduce their tariff burden. there’s a pull from the demand side as well that’s encouraging companies to double down on manufacture.

MV: Yeah, and I actually just misspoke. I didn’t mean Virginia. I meant Missouri, Texas, a tremendous amount of manufacturing is happening. There’s also a lot of, one of the big trends in the data center industry is prefabrication for both exterior construction and inside the white space. And, you know, that’s definitely a form of manufacturing that is happening. at different places around America here to streamline the supply chains. So, yep.

RS: Yeah, there are even some, I mentioned earlier, some unconventional plays that are really benefiting from this trend. The hydraulics companies, hydraulics and pneumatics, the cooling needs of these data centers are massive. You’ve got so many of these processors that are running through, you know, data so quickly, running through electricity so quickly, and that generates a lot of heat. And if you don’t dissipate it, you will have a very expensive pile of ash where you once had $1,000,000 worth of computer chips. So yeah, the traditional pump and hose companies are in on this as well and doing quite well with some of the more advanced hydraulic systems within those. Sarah Jensen, our colleague from Power in Motion, had hoped to come in and talk about that, but she had a conflict today.

JC: Robert, it’s interesting that you mentioned the cooling element. I mean, large data centers are using 5 million gallons a day to cool themselves. And you know, it’s going to be, it’s going to be a battle down the road, I think, about, you know, the efficacy of using that much natural resource for an individual facility. And so there’s a lot of opportunities, I think, for manufacturers to come up with alternatives, to traditional cooling methods. So we’ll see what happens.

MV: You just stepped back to your point on the liquid cooling and, you know, giving rise to need for, you know, the mechanical infrastructure in the way of pumps and tubes, you need to look no further than I was just at Data Center Dynamics this week and got to go inside the Boyd thermal truck where they have all of their CDUs and cold plate infrastructure. And, this of course was in the wake of their, 9th $0.5 billion acquisition by Eaton, which really turns, in some sense, Eaton into one of the, giants in terms of vendor supply for cooling and data center infrastructure.

Eaton, you know, being known as an electrical infrastructure house, but now acquiring all of this liquid cooling stuff, you go into that board truck and it’s all on display there. So a good point. And on the data center water consumption, John, I mean, you raise a good point. They definitely do consume a lot of water. Some of those systems, though, are closed loops. And the data center industry has a plan for that. But it’s definitely not to deny that data centers do consume a lot of water and power.