AI infrastructure—What it takes to build at scale

0:02Hello, and welcome to PwC's Accounting Podcast.

0:12Today we're talking about what it actually takes to build the infrastructure powering AI and why this massive build-out matters for business and finance leaders navigating investment decisions, infrastructure constraints, and long-term risks.

0:42Today, we're talking about what it actually takes to build the infrastructure powering AI and why this massive build-out matters for business and finance leaders navigating investment decisions, infrastructure constraints, and long-term risks. In a recent episode, we explored how geopolitics is reshaping infrastructure. In this conversation, we're building on that conversation and taking a closer look at the growing demand for data centers, power, equipment, and construction capacity needed to support AI and how companies across the market are responding. Joining me are Matt Cardamon, a principal in PwC's Capital Projects and Infrastructure Practice and Hugh Lee, a principal in PwC's Energy and Utilities Strategy Practice.

1:26We'll discuss what it takes to bring these projects online, where companies are encountering challenges and what this evolving infrastructure landscape means for business strategy, capital investment, and risk management. Matt, Hugh, welcome. Thank you, Heather. So maybe just as a quick introduction, my background is actually in civil engineering and construction, and that's what I was doing prior to joining PwC. I've been with PwC now for 16 years working in this capital projects and infrastructure practice and focusing on helping our clients with their big infrastructure investments and such a timely topic now spending so much time in the market in the utility energy sector and then also the digital infrastructure data center sector so looking forward to the discussion yes i'm excited that's a big sales job in the introduction so hopefully you're ready and how about you yeah thanks for thanks for having me and yeah uh excited to be here uh so bit of background on myself uh as you mentioned uh in our energy and utility strategy practice, been primarily serving our clients in the energy sector for about 16 years, about the same time as Matt.

2:39And Matt and I have actually collaborated quite a bit in the market because naturally as things have evolved, especially in the digital infrastructure space, big piece of that has been power and power as a critical constraint to a lot of the build out that that we need. And so our team has been serving this intersection of power and digital infrastructure for a number of years now. And so it's just been a really interesting time to see how we can kind of accelerate that time to power. You know, a lot of people talk about accelerating that time to intelligence now. Well, and I think for our regular listeners there, they know that I come, my background is powering utilities audit partner.

3:23So different direction, but definitely thinking about power demand. And otherwise, it's so interesting just to look at what's happened from an industry perspective and how interested people are in power and infrastructure now. So it's like a perfect time to be in these two sectors. So definitely. All right. So for our listeners who have not been so focused on power and infrastructure, I think it'd be super helpful just to take a step back and start with the big picture and why the infrastructure behind AI is such a business issue that extends beyond the technology sector. Thanks, Heather. PwC, in collaboration with Oxford Economics, actually just recently published a 2050 Global Infrastructure Outlook, where we forecasted infrastructure investment spend to be around$30 trillion just here in the U.S.

4:13And that spend itself is not just the actual data center shell, which is just a fraction of it. But really, that capital is needed, you know, not just for the chips, but it's the power infrastructure and everything that supports this ecosystem. So the infrastructure extends this sector into real estate, into construction, power, fiber, including subsea now, or more subsea, cyber resiliency, but then also the finance and insurance that comes along with all of it. And so these projects takes years to develop. And so these investment decisions now must account for the changing technology and demand that's out there.

4:55I think a lot of our listeners are probably pretty aware that the top 10 stocks make up roughly 40 %-ish of the S &P 500. And a lot of that is tightly coupled to the technology sector. But this investment super cycle has so many implications around U.S. competitiveness, energy security, affordability, which is why so many people are closely monitoring not just the AI demand itself and where that's going, but also just the infrastructure behind this demand and can we meet it in order to really achieve those goals around competition and security. Well, it definitely seems like you have known the question of just the massive build out and otherwise, but financing that much construction.

5:41So maybe a little later I can come back to that part of the question. But before we get into that, I think it'd be helpful just, again, to provide some additional context of why decisions that I'd be making now matter so much. Yeah, I think that, I mean, the time to build a data center, right, was, you know, a pretty short amount of time to actually do that part of it. But everything that supports it, the power infrastructure, right, a lot of those things have been pretty widely reported around just the delays that have happened, right? You have time for equipment, interconnection, mobilizing, engineering, construction, workforce, all those things.

6:17All those things naturally in and of itself take time to do, and that's why decisions now matter. But really, we should really also be addressing just some of the community pushback and moratoriums that we've been seeing around AI infrastructure build. You know, can't be remiss without talking about the midterms that just happened, right, where, you know, we've seen very recently you have, you know, approximately 10 states where there are either data center bans or, you know, 40 or more local moratorium instruments that have delayed data center build. And so that's something that I think we're closely monitoring now.

6:57Post midterms, there's a lot of rhetoric leading up to it around these types of moratoriums and really what's going to happen now afterwards. And can maybe there be a little bit more balance and maybe the heat of it cools down a bit to where we start getting into, okay, how do we actually go build this infrastructure? Well, and I have a question, I guess, can be for either of you, and maybe we'll get a bit more into this. But one of the things we've talked a lot about here on the podcast is just supply chain constraints as well. And so you mentioned this in passing, but just getting the equipment that's needed.

7:33So obviously you need all the siting approvals. You have all the community issues that you're talking about. But then just getting the equipment you need, that seems like it also has to be part of the challenge in this build out. Yeah, definitely. Yeah, for sure. I mean, we're seeing that both on the energy side and then the chips and sort of all the equipment that goes into the data center. All those long lead items are definitely a challenge, and that's delaying projects. And so we're seeing companies really think about how do they order those as early as possible, which means you need to know what you're designing for or what you're building in order to order them.

8:07And so when they order those ahead of time, they're then doing it sort of speculatively because if it's a developer, let's say, they may not have a tenant that's signed up already. And so they're ordering equipment before they even know who's going to occupy the building they're building. And then they end up allocating it to the project that is ready for it first. And so there's creative ways that companies are getting around this, but it certainly is a constraint and a challenge and adds to the complexity of delivering this kind of capacity. Yeah. And then I guess from a power perspective, I mean, building power plants takes quite a long time.

8:42And so if there's a huge need for energy, but then you're trying to figure out where the projects are going to be built and how to contract, like how is that intersection working as companies on both sides are trying to make those decisions. So you don't know where the project's going to be built. You don't know where the energy capacity is needed. That seems like quite a conundrum. Right. Yeah. And it's been interesting because there are many places you go and you can talk to a lot of utility providers that are out there, you know, and they'll say they're open for business. And that's right. There is, you know, spare capacity on the grid that allows projects to get connected.

9:16But it's always in the details. And it's always around, you know, where are things going to be specifically located? It really comes down to also just commercial terms. And so how will actually the commercial terms work for getting some of these kind of large loads connected? And that in and of itself just takes time because you can say, okay, maybe we have spare energy on the grid. okay we need to start planning for how does that actually work how do we operate it traditional energy providers just don't operate at the same speed to be able to make some of those decisions and allow for that and then of course just the the back and forth around okay well am I going to provide new generation for this large load customer whether it's a data center or otherwise am I going to build out new grid infrastructure for them as well okay well how does all of this get paid for?

10:10Because, you know, in the end, we're trying not to impact, you know, our, you know, you and I and everyday rate payers with that. And I think for the most part, you have a lot of the data center companies, you know, hyperscalers and otherwise that are willing to pay for the infrastructure that they need. It's just that the mechanics of it take take time to kind of figure out. And I think that's where there's a bit of a dissonance between perhaps maybe what we read in the newspaper or what we hear in headlines versus just the reality of working through those details to make it happen. Well, and I think that was actually one of my follow-up questions is just that I do think there's this headline of that, oh, this build-out is making everyone's power costs go up and it's being passed through the rate payer and it's this fair and all these types of questions.

10:59But what you're describing is a bit different is maybe that the data center providers or other large customers are willing to pay, but maybe it's how the rate structure is set up or it's just that's how the headlines are set up. Right, right. Yeah, I think it's a bit around the headlines in terms of that's what people are seeing. And it's because they're, generally speaking, rates are going up for a whole host of reasons. And so it may not be specific to, oh my gosh, this data center that I think is going to be in my backyard, but it's a whole host of other things that have just been around broader grid infrastructure and other things like that that need to be updated because they're decades old.

11:41Or we're retiring old coal plants and other generation, and we're building in new ones, and that in and of itself needs to get paid for. So some of this also starts to kind of get blended in with just all of this massive build out that's happening. And so both kind of top or like kind of issues get convoluted a bit. And I think people just associate or they can just point to, oh, that data center is what's really going to be the problem or that's what's really driving up my energy bills where, you know, it's a bit more nuanced than that. Yeah, and I think that's so interesting because, and it's something we've actually talked about here on the podcast, but I hadn't made that connection is that, you know, this big push, retire coal, maybe bring some nukes back online.

12:24But you do have like a hugely aging infrastructure from a power perspective. And so that's all happening at the same time. And you're right, it is easy to say, oh, well, look, this, you know, this new power plant is being built. It has to be because of this data center that's being built. Right, right. Yeah, exactly. But some of that, what we're seeing is leading to this behind the meter movement, right, where a lot of companies are starting to say, well, I can get power faster by building my own generation plant next to a data center, which then takes the pressure off of or takes out the discussion around does that increase rates for rate payers in a certain area.

13:02but it does go back to that affordability and the community engagement and then how do you get speed to capacity goes into all the decision making around what is the power mix that's on a particular campus or next to a particular data center so that's a another big i guess movement and we're seeing more and more of that behind the meter at least a hybrid version of that happening in the market right now okay i'm gonna ask one more question then we'll go back to some of the more key issues, but then are you seeing then that's the utilities themselves that are entering into these behind the meter contracts as more independent power producers?

13:37It's both. It depends. I'm looking at both of you. Yeah, no, I mean, yeah, I think it's really actually kind of new entrants. So it's independent power producers. It's a lot of new entrants coming into the market. You know, when we're talking about trillions of dollars of spend, you know, that's going to attract a whole host of people to come in and solve that problem. And so that's where we really see it is kind of these newer players coming in. They're creating new partnerships, right, with existing players in the market or others that they're pulling in where people are solving kind of different problems within the value chain.

14:18But yeah, I think, Matt, to that point, I agree. like the kind of the behind the meter or the kind of bridge powered solution, I think will just continue to increase. And I think just a whole host of things to solve, I guess, so to say, will be needed when that happens. And maybe a whole separate podcast topic, but I think that also complicates how the grid thinks about planning, how the different utility players think about it. All right, let's go back to focus a bit more on the data center. So we talk a lot about power. I did jot down a few more podcasts we could do on this topic. But what does it actually take then if, you know, okay, we know we want the data center.

15:01We talked a little supply chain. We talked a little power. Big picture, what is involved then in getting a data center ready to go? Let's call it that. Yeah, I mean, well, we talked a lot about all the different stakeholders involved, right? I think that's first and foremost is getting all those stakeholders together, understanding where do you want to build it for business purposes, where can you build it based on the constraints with power, with fiber, where a community will allow it and where there's positive regulatory environment, there's permitting allowed. So all of that goes into it. And then orchestrating all the stakeholders around the equipment you have to order, the contractors and design firms that need to be a part of designing it, the tenant that will ultimately or the company that will ultimately take the capacity, and then all the technology equipment that needs to go into the data center.

15:52So all of that needs to get orchestrated. And then when you're talking about campuses, that's done typically in phases. And so you have to orchestrate the timing of when data center capacity comes online and when energy capacity comes online so that they're in sync. And then if there's a tenant involved, thinking about when they need to do the fit out and commissioning of the data center, that all needs to sync together and is often a very challenging process to get completed. So if I'm listening and more from a perspective of I'm either a customer or a company that's a user of data centers, but maybe not actually someone powering them or otherwise, what are some of the risks I should be thinking about in terms of what happens if this doesn't work?

16:39And sorry if I'm jumping ahead. I just think it's helpful to give this context of why people should really care about this process. I'll use the word working. Yeah, I mean, I guess a whole host of reasons, but I think for the call of the end users, it all still goes down to, you know, this is all about compute and the demand that we that we have as individuals or as enterprises that rely on the compute that's needed. And so whether that's, you know, obviously we talk a lot about AI right now, but this was happening well before the AI boom, right? A lot of processes were being moved to the cloud, right?

17:15You have a lot of that. You had streaming, right? Netflix and all those types of things as well. So there's already a huge demand for a lot of this compute that's happening within these data centers. And so if these don't get built out, these things slow down, your processes don't work, especially if you're an enterprise and you have a lot of these things. that are reliant on that data center capacity, those can be critical processes that you need to run your business. And so that's really where people will care. I mean, as you guys think about all of the challenges and the massive investment, then do you think there's high risk that we won't be able to build out the infrastructure that we need?

17:55Or to, again, I'm cutting ahead, but I want to really focus in on some of these key issues. I mean, yeah, I think that, and we talked a little bit about it at the outset, but I think if you think about, you know, just U.S. competitiveness and, you know, what we, what we're striving for in terms of being a world leader in AI and, you know, where that all goes. Yeah, there is a risk there. And so that is a bit of a concern, which, you know, I'm hoping, like I said, post midterm, maybe some of the rhetoric starts to die down a bit and maybe we start to find, you know, more balanced solutions. um but uh but yeah i think that there is that that risk there if um if we don't kind of build that out then are we uh going to be behind from wherever you know the this this technology can can take us because we're already seeing you know tremendous impact in terms of just not only our daily lives but what what this is also turning into across enterprises and and things like that and the whole host of other opportunities that this unlocks, whether that's research, whether that's robotics, whether that's autonomous types of solutions.

19:05There's a whole host of other things that we're seeing this technology hopefully unlocking. And so that's the risk that we have of not being able to meet the build out. Yeah. I mean, just the indicator of that demand is also looking at how much CapEx is being spent across the market and what the hyperscalers are committing to, even when there's maybe pushback from the market, they're still putting out huge numbers that they're planning on spending, which I think is an indicator of how much demand they see in the market as well. And so the constraints and risks that we're talking about, about delivering that become very real.

19:38And I think every company is trying to figure out how do I get through that and be the winner or one of the winners in the market. Yeah, so then from a practical point of view, sticking with that, But are we seeing then a lot of new companies or a lot of companies either changing what we're doing or new market entrants that are getting involved to try to solve some of these issues? I know we talked briefly about some of these new entrants on the power side. But how about just on the infrastructure build outside as well? Yeah, we're seeing a number of things. I mean, we'll talk a little bit more about partnering, I think, is one.

20:10As you see, the CapEx spend is so high and investing or the financing, all of that takes a lot of dollars. So the partnering across different types of companies to address the complexities of power and data center infrastructure and share the risk and share the financial commitments and burdens around that is one thing. I would say another is engaging the community sooner. I mean, we're seeing, obviously, community pushback, but some of that was because a lot of these data centers were being built sort of behind the scenes and not really talked about. And now it's more front and center, and I think we're seeing more companies proactively figure out how do they address the real concerns that communities have.

20:53We talked a little bit earlier about procurement strategies. I think that's another is data centers are always a very standard sort of design, and then technology evolves so quickly, and requirements from different companies are changing quickly. But I think we're starting to see more standard-ish type designs that will allow for earlier procurement and allow for common procurement across different data center locations, which then allows you to order and allocate later to a project. And then I would say the bring your own power we talked about, that's another changing dynamic where most data center capacity was connected to the grid, and now we're seeing more where you're bringing your own generation power.

21:36And the last one maybe is around the EPCs in the market or the engineering and construction companies getting involved earlier and getting into more strategic partnerships with these data center developers rather than sort of bidding on projects and just jumping in in the construction phase. But I think, Matt, where you're getting at is really around the reason where you see a lot of these different movements or partnerships or, like you said, the EPCs going kind of more upstream is that there's these, it's what we were calling kind of seams of friction in the system where, you know, people weren't necessarily solving it before.

22:11That was what was really delaying a lot of this build out. And so you have different players kind of figuring out or trying to solve these different areas of friction to help kind of move the whole process faster, right? Everybody benefits when, you know, the whole process goes faster, right? Rises all boats. And so it's actually interesting because so Matt, myself and two of our other partners recently published an article called the Full Stack Data Center Orchestrator. And where, you know, we argue actually in that in that like the current state of like alliances and partnerships are just they're not going to be enough to win in the market.

22:52And so we have kind of this these four stages of maturity that we've seen or that we are seeing in the market. And so you had the legacy kind of viewpoint, which was around fragmented delivery. EPC did this, data center developer did that, you know, hyperscale customer did that. Like everybody kind of solved their own piece of that chain and it worked fine, you know, years ago. Today's reality is more around these alliances and partnerships. And, you know, you have examples like Engine 1 and GVernova and Chevron teaming together, you know, Invenergy, Emerald AI, NVIDIA, they're teaming together.

23:28There's a whole host of just different creative types of solutions that people are doing because they're solving different parts of those seams of friction that are happening today. Really, where tomorrow is going to be or what we're starting to see is that we're seeing more of these kind of infrastructure platform type of play where you have people that are solving both from the digital kind of technology side, but you also have people solving it from the power side of it as well. Google's recent acquisition of Intersect Power is a good example of that. Google, with a very large balance sheet, can do these types of things.

24:05But really, they're starting to move into that space because I think they're trying to basically solve the frictions that they're seeing in the market. And then finally, the final stage that we argue for is really around this kind of full stack orchestrator. And so it's kind of how does a almost singular party drive kind of repeatable execution of reference design, schedules, acceptance criteria, etc. But how does this kind of full stack orchestrator really solve all those problems to make the whole process move faster? And then also win more value in the market, right? We see that as something that somebody will get paid for.

24:43And so how do they solve that? or that's why we have to read the article yeah yeah I mean it's it's around some of the things Matt that you were kind of just highlighting right I think there's there's certain parts of it where if you're it I understand like not every design can be standardized but there's there's going to be certain pieces of it that I think people are starting to land on and I think you can kind of get get some some level of clarity amongst the different players within the value chain along that. I think there's other things just around being able to control. You talked earlier about supply chain, right?

25:19Being able to have better control of your equipment and when that's coming in. So it's just better ownership and control of that schedule. Yeah. And even when you just talk about the power side and the data center side, those are typically different stakeholders. And even when you get to the contractor level, that's actually constructing the infrastructure, it's typically not the same general contractor that's going to be building your energy infrastructure that is also building your data center infrastructure. And if you're building your own power, it's likely natural gas. And so now you need to be coordinating with a natural gas provider and who's building that sort of last mile of pipeline to get to the site.

26:00So all of that needs to be coordinated or orchestrated. Oh, that's why you need your full stack orchestrate that's right so all of these things need to come together in order to deliver on time and on budget and deliver to the requirements that the tenant or the hyperscaler is expecting to get from it and while you're doing this a lot of times these projects are being started without knowing what who the tenants going to be and what the requirements are going to be and so you're doing that in parallel to getting all these other pieces lined up so that's where this like orchestration and these seams that you talked about are so critical and why the complexity is as high as ever and you know with the volume of work there is for any particular company and the constraints in the market that you're seeing so are you seeing then that there are this orchestration is happening the seams are getting addressed or it's really depends i'm assuming on the project and otherwise or what are you guys seeing in practice in the market Yeah, we're starting to see that and we're starting to see companies emerge in this space.

27:06And it's interesting, they've come from different angles of it, right? So whether you come from more of the kind of hyperscale kind of technology angle, or you come from more of a energy angle, or you come from more of a data center developer angle, or even actually, to be honest, even private capital. So you see a lot of the portfolio players that have pieces of this within their portfolio, and then they realize, oh, I need to bring this all together now. And so I'm going to go out and make some targeted acquisitions and then actually build up that platform. So we're actually seeing that as well.

27:42So I think you have a lot of people that see it, and I think you have some early movers. But it's still a little bit early right now. I don't know if we're quite there yet in the market, but we're getting close. And I think that's where you're starting to see a couple of examples where people are going and kind of finding the right ways in which they're solving this. Yeah. One stakeholder that you mentioned was the data center developer. We're starting to see some data center developers that are really building that muscle around building energy infrastructure as well. So maybe historically they always just built the data center infrastructure.

28:19They worked with the local utility. They got connected to the grid. Yeah. Now they're buying turbines and they're figuring out how to build generation plants and transmission infrastructure next to their data centers. So that's a whole new capability. Sometimes that's through partnerships. Sometimes it's building the capability themselves. Sometimes, as you said, it's working with other portfolio companies that are part of the same investment umbrella. And so that's definitely an area we're seeing some of those seams close. And then I have to ask this question because we spend a lot of time working on sustainability, or that's where I spend most of my time focused now.

28:58And I think one of the questions that comes up with the power build out with these data centers otherwise is the sustainability, the emissions and otherwise. So as you're thinking about this, as you're talking about all the parties involved, is that a focus at all? Or are people really more focused on, I think we talked about this earlier this is like a need from a economic and otherwise for us as a country and so yes sustainability is maybe important but there's all these other priorities yeah i guess i would start with saying that i don't think the hyperscalers have backed off of their commitments sustainability commitments but with that said it's certainly clear that they're prioritizing speed to capacity right now.

29:43And so they're finding different ways maybe to offset the emissions that they're creating with the new solutions. So I would say it's a little bit deprioritized, but still a topic that's considered. And I think as they think about the long-term viability of where this is going, they are still considering some of the sustainability commitments they've made and how they address those. Yeah. I mean, just to build on that. And I think that's where you see a lot of the announcements that have been made around nuclear and have been made around other low-carbon types of solutions that can provide the scale of power that's actually needed for this amount of demand.

30:25And but but, yeah, I agreed with with with what Matt was saying is that I think you still have especially certain regions as certain places that benefit either from kind of existing nuclear that might be there or or the ability to build out more sustainable or lower carbon types of types of solutions. And so I think that will just continue to continue to evolve, continue to persist. And another point about that is, you know, there's also an argument around like, well, you know, power pricing as well. And so, you know, certain of those lower carbon solutions actually have are competitive from a pricing perspective, too.

31:03So I think those are things that will be there. I think, yeah, right now we're in that moment where we just need to build things as quickly as possible. And, you know, a lot of that, you know, is a bit on the back of natural gas. But we're also seeing people come in and try to solve for, you know, trying to minimize some of the emissions, leaks and those types of things, carbon capture as well, coming along with even some of those gas solutions. Yeah, because it seems like if you talk like true like zero carbon type of solutions, obviously then you run into potentially reliability issues, which I think is foundational for these data centers or the infrastructure that you have to have highly reliable power.

31:46So there is does seem to be some conflict there, but it sounds like people haven't forgotten that that's important. It's just there's also other priorities that need to be focused on right now. Is that fair? summary. Yeah, that's right. Yeah, agreed. Agreed. Yeah, I think, again, the reliability piece is huge because, again, if you and I are relying on, you know, that compute for, you know, our financial processes, as an example, right? We don't want... No, it's offline. Yeah, we don't want our main data center going down or anything like that. No, exactly. And I guess then that actually is a great lead-in to the next question, which is really...

32:20So if we take a step back, we talk about all these parties involved, we talked about the fact sometimes you don't even know where you're building, you don't know who you're partnering with. And so what, like Hugh, from your point of view, what are we seeing in terms of how developers are choosing, like who's involved, the location, the partners, what infrastructure to build? Like how are companies thinking through all of that? Yeah, I mean, it's interesting because there's always a kind of question around like, okay, location-wise, are we more open or flexible to that. I think you're seeing that with certain data center announcements or buildouts that have happened in what we would just call like kind of non-traditional locations.

33:01And part of that, I think, is mainly because of a power constraint and a need to just get speed to power. And so that's where you've seen some of those instances of where that gets solved in West Texas, as an example from Texas. And so see a lot of that happen. That being said, I do think that you still have a lot of core demand in what I would call are the primary markets. So where there has been a lot of existing, call it IT, fiber infrastructure, still a need or a demand to still build around that. I think there's still a lot of benefits and economies of scale when things are still built around those main areas.

33:40But those have been constrained from a land, from a power. Transmission, I'm assuming. Correct. Yeah, all sorts of things that limit that. So, yeah, so while there is a little bit of that kind of trade-off there, I think we'll still see a number that are built in, call it, non-traditional markets or even things that might be behind the meter, as Matt was mentioning earlier. But I still think that the traditional markets are still going to potentially win. What I'm kind of interested to see is that I do think you have couple of emerging regions especially like in the midwest as an example uh where i think you have a whole host of uh benefits of building out there uh that that could contribute to kind of maybe a new kind of epicenter of of data center build out i guess so matt from your point of view when you think about feasibility what are some of the things that you're that you really would focus on yeah i mean first what i think is interesting is some of these companies are actually going out to as many sites as they can when they don't intend to actually build at all the sites.

34:51They're trying to see if I can lock up a site and figure out the feasibility of it and then move forward with the one that will be the fastest or the most feasible. So you are seeing some of that. I think there's a little bit less of that because of how some of the utility providers are requiring more money up front. And I think the communities are getting more engaged, but you are still seeing some of that. But in terms of feasibility, you know, obviously we talked about power availability. So how do you get power to the site and how quickly can you get that? That's a big one. Some of the other surrounding infrastructure like water, fiber, you know, that's still critical and part of the equation.

35:31Then the economics of it. I mean, can you get laborers there? Do you have a contractor? Oh, yeah, because if there's no housing or whatever. So yeah, that makes sense. a big criteria. And then obviously the community's sort of willingness to allow it and the local government and regulatory bodies, if they're willing to give permits and what does that look like? What kind of constraints are in place? So those are some of the big ones that, you know, lead to the feasibility of building something. And I think what we haven't spent a lot of time talking about this, but the labor market constraints, you know, in addition to the supply chain constraints we've spent a lot of time talking about are a big deal.

36:11And especially when you're in these remote areas, thinking about how you're going to get the hundreds or thousands of people that you need to go actually build this infrastructure is really hard. And then on top of that, even if you can get them, you're typically maybe housing them there. And then what kind of strain does that put on the community, like a small community's overall infrastructure that may need to be supplemented as well. So there's a lot of considerations that go into the feasibility of these projects. And so how are companies thinking about some of those challenges? And it's funny, I was watching a documentary about building the Hoover Dam.

36:46So if you think about it, there's a lot of same constraints because no workforce was in the middle of the desert and they talked about building entire housing and otherwise. I mean, is that the path companies are going down or are they just importing a workforce or are they just trying to build in these places where there is existing infrastructure, which I think, Hugh, what you were saying before. Yeah, it's a combination of all of those things. People are building housing specifically for the labor force that needs to be there. So that's definitely happening right now in the market. And then in other places, they have maybe strategic partnerships with the EPC vendors, and they can get enough of a labor force in a particular area.

37:25And it might be an area that they're building in repeatedly versus a net new area that they haven't built in. And so that's a risk factor that we see with these developers as they're scaling. is it a market that they're familiar with or is it a new market they're entering and i think that you know changes the dynamic of how feasible the scalability is how about i guess from a skilled workforce though are these mostly skilled workers they're looking for or is it just people who can wield a hammer or i maybe should know more about this i know there's more with building power plants is it the same with building a data center as well that you'd need fairly skilled laborers Yeah, I mean, in a lot of cases, a lot of the trades are pretty skilled in their area of expertise.

38:11So like an electrician, the electrical contractors have very specific requirements that need to be built to for a data center, as an example. So those are skilled laborers that you need to get from somewhere. Yeah, so if you think even the utility workforce, I know there has been for some time questions that there's too many people retiring, not enough people who have these skills and knowledge. And so then this, I would assume then it's compounding from that point of view. Yes, very much so. And so even on the power side of the market, that was a continuing trend to begin with. And so you have a lot of training programs, apprenticeship programs, and things like that where people are trying to solve for to try to get as many more people in as they can.

38:58But to Matt's point, there's a lot of competition for that talent of those that still exist. And so, yeah, I think the electricians is a good example of it because you not only have it on the call at the data center digital side of it, but you still need to have that connected to some type of power source. And so you're going to have to have that type of skilled labor as well. and so I think yeah just the shortage of you know things like that shortage of skilled like from an engineering perspective as well making sure that you can solve for all the right designs and things is also an important area so yeah I think you have kind of limitations on a lot of fronts especially from the workforce perspective.

39:39So we one of the things we started and talking about impact on community and the fact that there's a lot of communities maybe have said no to data centers. I think you said 10 states or otherwise, or at least where there's holds and otherwise. But if I think about then this labor question, there also could be the flip side. It's like, well, it's bringing all jobs and the skilled workers we just talked about. And so when developers are looking more broadly at impact on communities, what are sort of pros and cons that companies are thinking about? Yeah. I mean, it's certainly the economics of it are important and it does create jobs it does create some economic value to the communities and you know potentially in the form of taxes or just in job creation um so that all gets part of the consideration and i think what balances the the questions that communities are asking around you know our electric rates going up what kind of water usage do you have those are balanced with what are the economic benefits for our community and you know does that help us work through or balance out some of the things that maybe we're more concerned about so i think that's all part of the equation for sure yeah i mean just building on that you know we we recently looked at um you know these kind of community moratoriums or push or like kind of bands that people have had and and you look at the reasons why certain counties um or local townships or things like that put in place.

41:07And it honestly, a lot of it is just that there just isn't a zoning standard. And so there's just there's no, there's the lack of standard. And so that's the real reason. And so people want to put put pause on there, because they're, you know, they're just aren't sure what the impact is, you know, yes, there, there are some that highlight, you know, water supply impact, some that highlight land use incompatibility, some that highlight, you know, electricity demand and cost allocation, right. So there are some communities that do focus on those particular topics. But I think a big piece of this is just around kind of education mainly of what it is.

41:39And I think, you know, being able to address, you know, definitely the concerns that people have, the challenges that come in with new infrastructure that's being built out, but then also really understanding the benefits as well. So thinking about, okay, yeah, this new development may actually bring in billions of dollars of tax benefit to a county that, you know, has never had that before. but to get that there's a lot of rhetoric concern around kind of what does that mean if I if that's being built in in my community there's a someone had done a recent study where it was like a data center came into you know local town or a county or whatever it was and then you can draw a radius around the whole place and then you can see all the other counties around it that start putting in moratoriums or so there's other types of just like delays to development mainly just because they saw something else getting built out in a neighboring in a neighboring town interesting but i guess do you see then a scenario where we would actually start to see more tax incentives and otherwise where people actually want these data centers to come in just because of some of these there's i mean there's pluses and minuses of a lot of things right and we talked up front about the fact that there's a need a critical need for these or we're not at that point we We just need to get over the moratoriums and otherwise.

43:02No, I mean, I think there's a big part of it was maybe a year or two ago, there's a lot of concerns around, you know, would a state have a sales and use tax incentive that was really around drawing in the type of investment that they, that they wanted, especially for these data centers to come in. So yeah, I do think that there is, you know, that, that type of benefit. And, you know, I think there was a whole host of very competitive governor races that we just went through in the midterms. And so, you know, I think that's where a lot of it will get solved. It'll get solved kind of at that governor state level.

43:35And so it'll be interesting to see how that continues to evolve. But I do think, you know, I think if one state says, okay, we're closed for business, then I think other states will potentially say, yeah, we are open for business because we see the demand that's out there. We see the need for, you know, AI competitiveness, right? you know, we have the energy security within our state, right? So I think people will, will come back to that. Yeah, I think the answer to that is very localized, like it could be state or even county specific. And, you know, take the example of Northern Virginia, where it which has most densely populated data center, you know, set of data centers in the country, at least, if not the world.

44:22And, and a lot of that started out because of all the tax incentives, And there's less tax incentives now, I think, in that region because now it's the opposite problem. They don't have room for anything. Yeah, they're like saturated. Yeah. So, yes, I think there are incentives out there, and I think it's very localized as to where those incentives are. And then we started this by talking about just the price tag and the huge amount of investment that's needed. And so where are we seeing that investment come from and how much risk is there that the money won't be there? Yeah, I think that's where we're seeing some of the concerns of that in the market right now.

45:00I think it will be interesting to see how that continues to evolve. But I mean, obviously, as Matt was mentioning before, you have the hyperscalers themselves that are investing a tremendous amount. I mean, you have obviously concerns around circular financing with the chip providers and things like that that are kind of funding some of this infrastructure. but also just a whole host of private capital, but also sovereign wealth capital that also comes in. So there's a whole host of, I guess, funding sources that are contributing to this. And I think it's really coming down to a bit around who can really execute, who can really deliver these projects.

45:48Because I think that's where maybe there's a bit of a, I don't know familiarity bias if like you can if you can execute and you can deliver then you're going they're gonna get the fun yeah and then you keep you keep going back to the the people that can get it done yeah so if you guys had to rate the various constraints because we talked about money we talked about workforce we talked about water we talked about power supply chain like there's a lot of constraints which ones do you think are the most potentially problematic when we look ahead to next year and then beyond you willing to look into your crystal ball yeah I mean I don't think it'll be financing at least at the top of the list I think them there's a lot of capital out there that seems to be hungry for this space so I would say still power I think power is going to be continue to be a big constraint or a major constraint I think permitting and navigating and ask about that navigating all of the you know regulatory regulatory community stuff is going to be, I think, one of the constraints and maybe surpass power at some point.

46:57Because I do think people are figuring out the power solutions through these different models that we talked about. Yeah, you talked about a lot of different. Although power is still a big constraint, I think. So those two are probably the biggest. I think supply chain constraints, and that comes in various different forms. So when one aspect is solve for, then another one pops up and there's a number of different supply chain constraints that I think will be in the mix as well. So those three probably are the top three. Is that from your list? How about your list? That's one of mine. Yeah, I think Matt probably took most of the ones I was thinking about.

47:32We track what we were calling kind of like black swan type of signals or whatever in the market. And so, yeah, Matt hit on a number of big ones. I mean, the financing, I agree. I think there's a lot of capital out there. I think if interest rates continue to go up or things like that, I do think there will be maybe people get spooked or things like that. So that's one of the – At some point, that could become a constraint. Correct. Yeah, exactly. So that's one of the black swan kind of events or whatever that we're kind of tracking. But, I mean, Matt, you hit on some of them, right? Either it's an economic crisis or supply chain crisis, geopolitical, right, in terms of where our chips come from.

48:10So those are the things. I think maybe one other that maybe throw out there that we didn't touch on, which is, you know, again, a black swan is just maybe either a call it like a trust or innovation type of crisis. So we didn't even touch on that. So if all of a sudden, you know, there's a major failure, like failure, or, you know, all of a sudden, you know, AI, like there's concerns around, you know, you know, the trustworthiness of that. Again, not. Or more of some of the events we saw like over the summer, otherwise. Correct. Yeah. So, I mean, those could be kind of other things that are out there.

48:43But I think, like, the reality, though, in terms of, like, if we were to look at, you know, what's actually going to go to operations in 2027, probably some of the things that Matt was saying, really, it's, you know, it's the power, it's the workforce, it's the community pushback and permitting. Like, I think those are the things that, you know, you can't solve those things. You're not turning anything on in 2027. Yeah. Yeah. So how is you guys are looking then? I know you're tracking this carefully. If you're looking that it's progressing, I'll use sort of as needed, sort of to meet all the demand.

49:15What are some of the key things you'll be looking at next year to say like, okay, yes, we're moving in the direction that's needed or it's not moving fast enough? Yeah. I mean, maybe one of the things I would start with is, you know, the hyperscalers, CapEx commitments. I think if they are still elevated, that signals, obviously, they think there's demand in the market. they're willing to spend to build that capacity, which means if they're building capacity, they're also seeking capacity from others likely. So I think that's an interesting signal to look at. I think on the energy side, I know we talked about behind the meter, but the large majority is still connected to the grid.

49:53So I think looking at the interconnection queue times is going to be interesting and seeing if that comes down at all and if the government can help sort of grease the skids on that and getting more connected to the grid. And then the other one you mentioned to you is just delivery. Is the capacity coming online and who in the market is proving that they can deliver repeatedly on time and on budget? Because I think the market's going to respond really favorably to those people. Yeah, yeah. I think you and I were talking before we hopped on here and you were talking about how many are there going to be ready for service, right?

50:28Like that's the kind of indication, I guess, of data centers going into operations. And so being able to kind of track that, right? We track those that are in development versus construction, but really it's the how do you get over that last mile into operations. So there'll be something we'll be interested in to see what's actually up and running going into next year. But yeah, I think agreed with all the items that you mentioned, Matt. Maybe the other one to add to that would just be maybe a little bit around just where the demand signals continue to go. So as we think about how agentic workflows continue to embed themselves within our daily lives, right, in terms of how we interact with it on our phones or other devices or things like that.

51:16But then ultimately, where does that also go within enterprises as well? So a lot of work that we see in terms of the work we as PwC do with our clients in terms of how do these agentic workflows get embedded within a company's business. I think that will also be an education of just where we continue to go and signal to the hyperscalers to continue with their CapEx spend. So if I take a step back, for most of the listeners to the podcast, they would be on the end user side. So either as consumers or for the business perspective. And so probably the biggest question they would have is like, okay, what does this all mean to me?

51:53Like, what if it doesn't get built out fast enough? Or what if all of a sudden you can build these data centers everywhere? Like, I think there's a few different scenarios. So when you're advising businesses that are not directly involved but are more on the consumption side, what is some key piece of advice you would give? I guess for the call the everyday kind of consumer of this in terms of the implications for them, I think in the end it kind of comes down to there's maybe, I guess, going back to just where the overall market is, I think, in terms of kind of a reliance on maybe over reliance perhaps on this getting right for the broader market.

52:32So I think people might be feeling that in terms of their own investment portfolio. But I think also ultimately in the end, like just like as you and I, as like just everyday people, I think it'll be a little bit around like, well, how will, you know, this potentially continue to benefit our lives? Right. And I think there'll be a bit of a bit of it around like, OK, like how how am I finding new ways to solve problems that I don't like doing? Yeah, so you're saying basically, though, you would be less focused on, oh, there's going to be so many constraints that this won't keep moving forward. And more so, how can we use, like, if I'm a company, how can I harness this technology?

53:10Don't worry so much about the back end and is the power going to be there, power being power of AI or otherwise, but more so, how can I use it? It's probably what you would say to focus on. Well, yeah. Is that fair? It is. But one thing that may play out is if the capacity can't get built at the rate that we're thinking, it may drive the price of using AI tools up, right? And the cost per token that companies are using. And then that may constrain how you use it in your company or increase the amount you're spending on AI. So I think that could be a supply constraint. Yeah, in a different way.

53:51In a different way that as an end user, you might be thinking about. Yeah. And I guess because you focus on the energy power side, do you see that the power is going to be able to be built fast enough? Like you had, again, look, I love crystal ball questions. So if you had to look ahead, do you think it will keep up or you think we're going to hit a roadblock? I think there are definitely a lot of challenges ahead. Such a diplomatic answer. Yes. I think a lot of the clients that I work with are doing a lot of work right now to try to meet that. So, I mean, as it stands right now with some of the demand forecasts that we have going out to 2036 and we're seeing anywhere of, call it 215 gigawatts of demand at that time, And like the amount of power that needs to get built out in order to do that would be a huge effort to do.

54:49And so do I think our power providers can get there? I'm hopeful. All right. We'll have to come back next year and ask the question again. So, gentlemen, this has been a fascinating conversation. I guess there are other points that when you're thinking about these issues that you would touch on that we didn't talk about today? I would say maybe as I think about the work we're doing in the market right now and what companies are asking of us, I think there's three themes that we're hearing a lot of. One is around readiness to scale. And scaling may mean we're spending more money and we need to deliver more throughput of capacity.

55:29So how do we do that? And then another is around understanding all these risks. What are the risks and how do we mitigate them? And then a lot of that leads to then discussions around, well, what is our strategy or how are we playing in the market? And what are some of these themes we should be getting into? And what is our operating model need to look like in order to meet that scale and mitigate those risks we're talking about? So I think those themes are all sort of coming together and what we're hearing a lot from our clients that are trying to build this capacity. All right. Very interesting.

56:01Any other thoughts from you? Yeah, I thought Matt, some did really well. All right. Well, excellent. What a great note to end on then. So gentlemen, it was such a pleasure to talk to you. Thank you so much for joining me today. Great. Thank you.

56:16This podcast is brought to you by PWC, All Rights Reserved. PWC refers to the US member firm or one of its subsidiaries or affiliates, and they sometimes refer to the PWC network. Each member firm is a separate legal entity. Please see www.pwc.com slash structure for further details. This podcast is for general information purposes only and should not be used as a substitute for consultation with professional advisors, including accountants and lawyers.