Listen: Data demand: US power sector responds to AI-driven growth

View Full Transcript

Hill Vaden:

All right, Sam, it's April 16th and we have just recorded a fascinating podcast on data centers and power demand with Dan Thompson and Adam Wilson. What'd you think? It kind of blew my mind.

Sam Humphreys:

Yeah, I think that's a good way of putting it. It's a subject, I think you and I were talking about this the other day, that it's suddenly been brought to everyone's attention. It's been there for a while, but we haven't really talked about it. Now it is this big thing and it involves everyone. How our lives are so entwined with it and it's this big mystery that we don't really think about. I mean, the guys did a great job of explaining for starters what a data center is, which I didn't know necessarily going into this, and it was a great conversation, which I think we're going to have many more of, hopefully, as this thing grows. What about you?

Hill Vaden:

Yeah. I think we're going to have to bring them back and get some more of our colleagues to talk about data centers. I know we've got plans to do something in the upcoming weeks on AI, but this is a huge theme. We talk about the intersection of investment and energy. This is bringing a whole other intersection to it. There's a whole lot to cover. Yeah, we can hand it off to the podcast now, but it's a really enjoyable episode.

Sam Humphreys:

Yeah, really good. I hope you all enjoy it.

Speaker 3:

This episode of EnergyCents is brought to you by our Financial and Capital Markets Energy Advisory Group, part of S&P Global Commodity Insights. Our team of experts provides the investment community with actionable insight and integrated thought leadership that identify the trends and trend makers of global energy markets. Solutions cover the full energy and natural resources sector from traditional fossil fuels to emerging clean tech ideas and supply chains, and are available via recurring reports, webinars, robust data sets, and personalized engagements with experts.

Hill Vaden:

All right, welcome back to EnergyCents, an S&P Global podcast covering all topics on the intersection of energy and finance. This is your host, Hill Vaden, and I'm here as usual with Sam Humphreys. Sam, it's about, I think three days before Taylor Swift releases her new album, four days. Are you excited about it as are so many Americans?

Sam Humphreys:

Yes, I am. I am still really miffed that I didn't get tickets when she's over here in a couple of months, but yeah, I can't wait. I'm a right nerd. We were watching her documentary on Disney+ the other day, and it's phenomenal. I want all of her shoes. What about you?

Hill Vaden:

Yeah, and my daughter was telling me today that she's hopeful that it releases timed with the UK so that when she wakes up, it will have already been released, that she wants it released early in the European day rather than waiting for the start of the 19th. It's the 19th, right, in the US?

Sam Humphreys:

It's the 19th. Go with that.

Hill Vaden:

It will have been released by the time this podcast goes live. I'm going to use this as a segue, but we have the Taylor Swift of energy conversations. As our guests today, so Dan Thompson and Adam Wilson are here today to talk to us about data centers and power demand, which was... Well first, Dan, Adam, how are both of you? Thank you for joining.

Dan Thompson:

Oh yeah, thank you. It was quite comical. My wife actually accuses me of being a Swiftie, but it was funny on the lead in, I think Adam and I were staring at each other like, are you a Swiftie? Who's a Swiftie?

Hill Vaden:

Oh, no. Aren't we all?

Adam Wilson:

Yeah, it's unavoidable at this point. She's everywhere. You just have to come to accept it. Hey, I have no aversion to Taylor Swift. She makes some catchy music.

Hill Vaden:

I have several of her records, so I suppose I would be a Swiftie. We are going to talk about data centers today. Data centers, this is a topic that we've been, I think we had this on the calendar like a month ago, maybe a month and a half ago. You guys are so busy working with customers and working with others about power demand from data center installations that we've had to delay it time and again, so I'm grateful that we're finally able to get on your calendar, and I'm sure our listeners are anxious to hear about data centers and the overlap there with the energy sector, because there's a lot going on. There's a lot going on as it relates to clean tech. There's a lot going on as it relates to potential gas demand. It's North America centric, global perhaps in nature.

A lot to discuss. We'll probably run out of time, but Dan, maybe if you could just kind of set us up with one, a brief description of what a data center is and what a data center does, and then we'll start thinking about how all of this growth is impacting North American and global power markets.

Dan Thompson:

Yeah, sure. It's kind of funny. Data centers are a thing that we all use, but you're probably not knowing or understanding that you're using it. Meaning that if you have a smartphone and you use Facebook or you use Instagram or TikTok, if you manage your finances online, if you send emails, all of those things are flowing through a data center at some point along its path from me to you or from you to Facebook or from wherever to wherever.

Essentially what a data center is, is a large building, or it can be a smaller building even, that are filled with these sort of rows and rows of racks filled with computer systems at a very basic level. These buildings can consume tons of power. Each rack that you see in a data center can be equivalent to the energy consumption of a US home. When you see a room full of this, you're literally seeing neighborhood scale power being deployed in a single room. A lot of these data halls are about 10,000 square feet or roughly the size of, I think now three US homes. All this is being crammed. I'm just trying to get at the scale of it all.

As a pop culture reference quickly, if you've kept up with the Marvel films and you saw the original Ant-Man, there was actually a battle scene that happens inside of a data center. One of them was shrunk down and was kind of flying through and they end up getting inside a server and they're trying to blow it up or whatever. Anyway, if you remember that sequence, if that happens to ring a bell, so that's what the inside of a data center looks like. Just imagine in your head these sort of tall, black colored racks that are filled with computer systems and there's literally just rows and rows of these things. That's essentially what a data center is. Companies deploy their infrastructure inside those, and so your digital assets as a company, so file servers, mail servers, all that lives essentially in a data center.

Hill Vaden:

Perfect.

Sam Humphreys:

Okay, so I forget, Hill, if you had asked this second part for that question, but I'm going to jump in. You've explained what a data center is, which is brilliant. I think part of the reason they are such a hot topic at the moment is this talk about AI, and now let me get my technical term right, crypto mining. Can you just sort of expand on what relevance those two key developments in technology are playing on data centers?

Dan Thompson:

AI you can think of as yet another use case for data centers. AI is being deployed inside the cloud, and I'm using air quote if you can't see me. It's being deployed largely inside the cloud, and the cloud is largely deployed inside data centers or is 100% deployed in data centers. As we think about AI and its potential as potential growth, what that means is growth and growth potential for more data centers.

There's also an interesting thing in AI, if you look at how it's deployed from an infrastructure perspective, it uses computer systems which are very power hungry. If we go back to that rack analogy, and it's basically equivalent to one US home, the future of this looks like neighborhood scale power in a single rack, not a whole room of racks. We talk about this in terms of increasing power density, meaning more power is being shoveled into the same square footage. That's what has everybody's attention.

If you pay attention here in the US at the power markets, there's been a number of power companies who have missed their forecasts and are reevaluating those forecasts for how much new energy they will need in their given markets. A lot of them have pointed at data centers as why they missed their forecasts or why they were off or why they need to re-evaluate. I think that sort of launched them forward, and then AI has everybody saying, "Oh, okay, well not only is this problem a problem, it could be more of a problem."

With crypto folks, so they are using very similar infrastructure that the AI folks use, meaning it's power hungry. And with the skyrocketing value of Bitcoin, for example, it has everybody looking at this again. We have been here before though, let me just throw this out. With the crypto folks, there was a number of years ago where everybody was saying, "Oh, these are going to wreck everybody's power markets because everybody's going to deploy these things en masse." Then suddenly GPUs got really hard to find. GPUs, by the way, are the underlying computer system. This is a very generic way of saying it, but they're sort of the underlying computer systems that make all this work and they are very power hungry. Anyway, we've kind of been here before and it didn't quite launch as crazy as everybody thought it might with crypto. However, there has been a resurgence there and it's become a meaningful demand source of power for power companies. That's why, yet again we're talking about it.

Hill Vaden:

Let's bring Adam in on this. If we're thinking about North America power demand, if I'm correct that load growth in the US has been fairly flat forever, for the past several decades, and all of a sudden data centers are coming in and adding growth needs to the North America power system that really hasn't been asked to grow in terms of capacity for a very long time. Can you help us put this into context from the North America power side of things?

Adam Wilson:

Yeah, absolutely. Thanks for having me on. Yeah, as you said, Hill, power demand, at least within the US has been relatively stagnant for the last 20 years or so, give or take. This has even been in the face of rising demand from data centers. Data center load has increased about fivefold over the past 10 years or so, but this has been countered by energy efficiency measures and other things that have reduced the overall impact on the grid.

Now the question is with the manifestation of AI, as that becomes a bigger role in the overall tech world, will that start to buck that trend? Just to put some context around this, and Dan, you can please feel free to correct me on some of these figures. Some of the projections vary, but just if you just take a simple Google search and how much power that requires compared to what a ChatGPT or some other AI request is, it's anywhere between 10 to 30 times as much power to do one of those AI searches compared to a basic Google search. And we're talking about millions or billions of these happening annually. This is what is starting to drive this rapid development interest, or I should say continued increasing development interest for data centers and how we expect this to increase the load on the grid.

What is likely to happen is this load will not be uniform across the US grid. It's going to be concentrated in a handful of select markets, at least to start, and that's where a lot of companies, a lot of utilities, Dominion Energy in Virginia has been a name that gets thrown around a lot because one of the biggest data center markets in the world is Northern Virginia. It's either them or Beijing. Dan, I think you can confirm that. That's where we're seeing a lot of this development.

The power demand increases that we're expecting to see from this surge in AI use, and it's honestly not just AI. I mean, we're still seeing power demand just from all other basic data center tech uses. It's going to be concentrated in a handful of markets like Northern Virginia is the main one, but there's several other markets, Phoenix, Atlanta. Really the conversation comes down to how are these local grid operators, how are these local utilities, what their kind of forecast projections are, and how are they going to handle this power demand increase in the face of all these kind of crazy projections that we're seeing in terms of power demand due to AI?

Dan Thompson:

Adam got it right there. The largest US market is Northern Virginia, which is north of DC, so Ashburn is the town that that's in. Then there's sort of surrounding counties around Ashburn are now seeing a lot of data center development. It is second only to Beijing. Beijing is the largest one in the world. Oftentimes people, if you're familiar with Ashburn and sort of where it is, people are usually shocked to hear that that is nearly the same size. It and Beijing are very close in size, but Beijing just barely edges it out. People are usually shocked to hear Ashburn is the same size as Beijing in terms of data center, that's crazy. It is a little crazy, but yeah, that's the truth of it.

Sam Humphreys:

As someone who has no idea what that is, because my geography is really poor, I apologize, but can you just explain for the folks that are outside of the US, what is driving the growth in those certain geographies within the US? Why is it in, is it Ashburton, did you say?

Dan Thompson:

Ashburn. Yeah. Ashburn got its start, it's actually kind of a funny story. Ashburn is sort of north of Washington DC, and a lot of people, your knee-jerk reaction would be, "Oh, okay, so they established near Washington DC, so this must be a lot of government workloads or maybe the Washington DC Metro is large enough to support a lot of development." Actually the truth of it is Ashburn has, it starts way back to the beginnings of the internet. There was this thing called the DARPANET, and there were a number of different connection points across the country. One was called MAE East, M-A-E East, and that was in Ashburn. Then there was MAE West, which is what we know now as Silicon Valley, and so all of these, the beginnings of the internet sort of spurred growth. In terms of data centers, growth sort of begets growth.

We often joke that data centers are a bit like sheep and they like to be close to each other. There's kind of the economies of scale both from, if you imagine that all these data centers have to be connected by a lot of fiber connectivity and also power connectivity, if you're close by, you kind of have economies of scale because we can run fiber in terms of finding fiber routed ways, power routed ways, things like this. It really has its beginnings, like I said, all the way back to the beginnings of the internet, and it just continues to be interesting.

Just as real as that though is also this constant hand-wringing over are Microsoft, Google, Amazon, who are the main builders and leasers of data center space in the Ashburn area, is there going to be a day when they say Enough is enough and they start spreading out all over the place? I've been here at S&P through the 4551 acquisition nearly for 10 years now, and I've been hearing that for 10 years, and so far enough has not been enough.

Anyway, so they do kind of continue to congregate. I will say though, the circle that we draw around Ashburn and call that the Northern Virginia market continues to get larger and larger and larger. It's just because land, as you imagine, availability, land prices dictate to an extent where these things go. And so as that market has gotten bigger and bigger and bigger, it has necessitated us to sort of expand the circle, but it's all to service that same purpose, which is the Northern Virginia market.

Hill Vaden:

Adam, thinking about the power demand or the power side of this, as Dan was saying, you want clusters of data centers for all the reasons Dan just described. Are you looking for things also in terms of power? I understand a lot of the data-hungry companies want clean power and they want reliable power. I'm sure that they want low-cost power. Northern Virginia has its power grid as it is today. Is this a chance for Dominion or whomever to find new customers, or are they going to get squeezed out as these companies look for on-site generation on their own?

Then a tangential question that I find really interesting is hydro. If I'm one of these companies building one of these data centers and want to put out a press release that says it's renewable, can I use a dam to do that? Are you seeing that grow? Is that becoming a bigger player than it was 5 or 10 years ago?

Adam Wilson:

Sure.

Hill Vaden:

A lot in there.

Adam Wilson:

Yeah. We'll break that into two parts because the second part of that question is a big topic right now. To speak to your first question in terms of how these utilities are handling this demand, what's interesting is, as you alluded to, a lot of these data center companies are hyperscalers, Amazon, Meta, Google, these are all known as hyperscalers that leverage cloud computing and data centers as part of their core business model. They do have these self-imposed clean energy commitments to go 100% carbon free, 100% renewable by 2030 or by 2025 or whatever. The common way that they do that is by contracting with wind and solar plants via power purchase agreements. Most commonly they use virtual power purchase agreements where they contract with a facility in a separate location, buy that power, and then that power kind of offsets the power that they consume in another location, if that makes sense.

That kind of hedges against the inherent variability, uncertainty of the energy mix, of the grid that they're actually consuming power from. There's a lot of power in Texas, for instance, that's kind of the hot corporate renewable market in the US at least, where Amazon, Google, honestly, dozens of companies have contracted with wind and solar farms to offset power consumption in Virginia.

These companies are claiming 100% renewable. We've bought all this renewable energy. At the same time, it's putting this increased pressure on the local grids. We will sit with the Virginia example. As their demand increases, they're trying to keep up with clean energy commitments with the data center companies that they have. At the same time, the concern is that it's almost coming at the expense of the broader customer base of Virginia, and we're starting to see more of these utilities and some of these high growth markets for data centers add a little bit more natural gas into the mix that they were expecting to build 2, 3, 4 or 5 years ago. We're starting to see a little bit more talk about delaying the retirement of legacy fossil fuel and coal generation in some of these markets to make sure that they maintain reliability in the face of this growing energy demand from these data centers.

It's a bit of a tug of war between these companies trying to keep their own clean energy commitments while also kind of forcing utilities to reckon with the fact that the utility's number one job is to maintain reliability, maintain power 24/7. If that means that they have to bring more natural gas generation online, if that means they have to keep a coal plant online three to five years longer than they had originally intended, that's what they're going to have to do. That's a conversation that's been going on right now is yeah, these companies are fostering the procurement of additional renewal sources, but at the same time, we're seeing a lot of these concentration of data centers and the development, we may actually be delaying the clean energy transition a little bit by forcing more non-intermittent generation, such as natural gas and coal to stay on the grid.

Yeah, it's kind of a long-winded answer to the first part of your question. I guess I'll pause there and see how you want to follow.

Hill Vaden:

No, it was a great answer. Sam, any comments or follow-ups?

Sam Humphreys:

Just a short one. It might not be within the realm of this, but we've done an episode before on transmission networks to get this power to homes and businesses. With this increased demand, whether it's from solar, we've talked about the issues with trying to get solar power to the end user. Is that another hurdle that companies are going to need to overcome in the near term just to get the power to the data centers themselves, or is that kind of tied up with what you were saying about delaying the targets of the green energy transition?

Adam Wilson:

Absolutely. Transmission and distribution. Honestly, a couple years ago, transmission and distribution infrastructure was kind of considered one of the biggest hurdles to the clean energy transition and then arguably still might be. Now data centers and AI demand has kind of become the new hot topic for good reason, but it's something that these data center companies are continuing with themselves. They have development plans to build a new data center, but the utility might be like, "Hey, okay, you can build that data center, but I know you want to bring it online in 2025. But honestly, due to transmission constraints and our own development plans, you actually can't connect that until 2026, 2027." It's something that is impacting the data center industry as well as the broader energy industry as a whole.

It's a big topic of interest. There's a lot of investment being put towards transmission and distribution upgrades to try and stay ahead of this, but it's having a really hard time keeping pace with the amount of wind and solar, this intermittent generation that's coming online. It's having trouble even keeping up, let alone making up the ground that it needs to make up to kind of accommodate all this energy demand that we're expected to see.

Dan Thompson:

If I could just add onto that, just a slight maybe retelling of what Adam just said. Because when I first started digging into the power stuff, there was a few things that didn't click right off the bat, and so just so it's said. What Adam was saying a minute ago is the hyperscalers, the Microsoft, Google, Amazon, Meta and so on, they are paying to have renewable energy created and put on the grid, but it's this rising tides lifts all ships kind of mentality. There's no expectation today that the renewable energy that Microsoft are bringing onto the grid, those electrons are flowing into their data centers. There's no expectation of that today. They're just dumping clean energy onto the grid and then using that on paper to say, "Okay, well, I put this many megawatts on the grid. I took this many megawatts off, so I'm good."

There is though still this issue of transmission distribution, even just getting any power, clean or otherwise, to data centers, which is what Adam was just describing. There are a couple of problems playing out here. There is a trend though that we've been seeing where the hyperscalers want that renewable energy to be closer and closer together to the data centers with the future goal that they can reasonably say that the electrons they generated are the ones they're consuming. They're trying to get from this sort of paper-based accounting, if you will, to actually, the energy that I'm consuming is clean and renewable. It's not just sort of a carbon accounting, if you will, but that's still a long ways off, I think. In terms of the Paris Agreement and other UN initiatives, the way they're doing it today is good. It's just their own goals of trying to get closer and closer to those generation sources.

Hill Vaden:

All right. I've got a follow-up question for you, Dan, on this, but I don't want to let Adam off the hook on the hydro.

Adam Wilson:

Yeah. Circling back to hydro, there is this trend that we're seeing, and it's not just within the data center industry, this is energy industry-wide. 10, 15 years ago, there was this big talk about transitioning towards more renewable resources. There's things called renewable portfolio standards that states have where certain percentage of the energy mix has to come from renewable sources, and that's helped foster the development of wind and solar generation across the US, and more recently energy storage.

But we're starting to come up to that issue where wind and solar, due to their intermittent nature, can only get you so far. There's been this transition away from renewable energy to clean energy, just what power generation sources do we have that are carbon free, and obviously hydro is one, and it's honestly not something we really expect to make a lot of waves within the US, at least within the corporate world.

Hill Vaden:

Make a lot of waves.

Adam Wilson:

Yeah, no pun intended, but it is something that we're starting to see. Microsoft has contracted with several hydro plants in Central and South America, and then we're starting to see a little bit more in Africa. A lot of these companies are starting to look outside of wind and solar generation towards a little bit more reliable, consistent forms of generation, with hydro being one.

The technology that we're kind of starting to see a resurgence of, particularly within the data center world, is nuclear power actually, because nuclear is carbon free and there's honestly a lot of legacy generation available. One of the deals that happened in the last month or so that really caught a lot of people's attention was Amazon contracting about 960 megawatts from a nuclear facility in Pennsylvania, I believe. The question is, is that going to start a new trend of these companies looking to contract directly with nuclear facilities, merchant facilities that have potentially generation that they could kind of siphon off and dedicate directly to a data center campus? JP Morgan put out a study that they tagged as much as 15 to 16 gigawatts of nuclear energy within the US that could be available for presumably any corporation, but data center companies more specifically to contract with directly. That's something that we might expect to see a little bit more of in the future with this deal between Amazon and Talen Energy maybe kind of starting that trend.

Dan Thompson:

Microsoft too are also investing in both fusion and fission. They're trying to play both sides of the new coin. Yeah, it seems all the hyperscalers, well, let me not say it so broadly, Microsoft and Amazon, and I believe Google also are investing in nuclear technologies and nuclear plays. It's just a question of who can get their hands on power really, and hopefully it's clean, hopefully it's renewable.

Adam Wilson:

The attractiveness of what I was speaking of with these 15 to 16 gigawatts of nuclear generation is that they're online and running right now. There is a lot of investment, as Dan said, between Microsoft, Amazon towards next generation nuclear reactors. More specifically, a lot of these small modular nuclear reactors that are more in the 50 to 100 to 200 megawatt range rather than the grid scale nuclear reactors that are 2,500 megawatts. There's a lot of development with companies like NuScale that are developing these reactors, but a lot of these projects are still 5, 6, 7 years away at minimum, and that's if they can get the funding and all the infrastructure needed to get these projects up and running. A lot of investment towards next generation technologies, as Dan said, but where based on this deal that Amazon made, a lot of attention has shifted back towards legacy nuclear generation and then how data centers can kind of tap into some of that clean energy.

Hill Vaden:

Yeah, we've done a couple podcasts on nuclear, and I think you're going to do more. You're going to have to another one on data centers. Sam, I want to save the last question to come from you. Before I give you that, Dan, quickly overbuild. I mean, this sounds like all the recipes, everything is there to get ahead of our skis. Is there a chance that we end up with a bunch of empty data centers or a bunch of generation that the data center demand doesn't show up? I mean, is that on anyone's radar?

Dan Thompson:

Oh, yeah. This question gets asked all the time, and what I would say is Nvidia are saying they can't make GPUs fast enough. They have already basically, I don't know if they put it in terms of pre-selling, but it sounds like when you listen to their leadership talk, they're in some sense, presale-ing, but they definitely will admit they can't build them fast enough. They can't keep up with demand. There are also plenty of buildings coming out of the ground that are pre-contracted. I would say it's actually in terms of data centers, it's few and far between that are built speculatively. They have some sort of either I've got a signed contract or a very warm or hot sales cycle going on this building. From today's view, it does not feel like we're getting out in front of ourselves.

We have seen a situation where that happened, though. In China, if you look at their data center development, they were developing data centers like crazy, and then it has slowed dramatically over the last four or five years. Part of that though is a bit artificial. What I mean by that is, so in China, the major hyperscalers there are Baidu, Tencent, and Alibaba, or at least I should say used to be. To keep things as low politics as possible, all three of those sort of got on the wrong side of Chinese policy and the companies were either carved up or they're still sort of in litigation to figure out how that should look. That really slowed development from them because suddenly Baidu, Alibaba, Tencent stopped leasing space, and so suddenly data center providers who had built for them or were building in hopes that they would lease, they were kind of left holding the bag.

The difference though is what I was just describing a second ago. In China, they were building speculatively. Here, we're not seeing speculative builds nearly as often. That's a bit of a good thing, I think, in terms of this question. Also, US government tends to be a lot less progressive in their antitrust type. I mean, you can kind of look at what happened in China as almost like an antitrust type motion that we would see here in the United States. It's a bit different, but if you'll kind of give me the grace of a broad analogy there. In any case, so we tend to be much slower with that type of movement here in the United States, and so that would be repeated here as a very low risk, I think.

The other thing that I would point to is that we hear often comparisons between the .com bubble. People saying, "Yeah, you guys said all this was going to happen back in the .com days. That bubble burst and it never happened." The difference between now and the .com days were back then it was a bunch of venture capitalists essentially investing in things they didn't quite understand. You had a bunch of startups doing interesting things, making these cool looking webpages. Nobody had ever seen a webpage before, and everybody just got really excited about something new, and you had people investing that didn't quite understand what they were investing in.

If you look at what's happening today, the companies who are investing are strategically doing so, and it's companies by the name of Microsoft, Amazon, Google, Meta. These people are not, they're not going to invest literally at this point, hundreds of billions of dollars in a mistake, or at least I can't see that happening. All of those companies, by the way, are using the technologies themselves and also selling it to startups who are using it, which they also are investing in. There's this huge investment cycle that would suggest that, or is designed to, kind of prop up this development that they're doing.

To me, if you look at those indicators, I mean, could you call that a bubble? I'm not an economist, so I wouldn't be able to categorize that perhaps completely, but I can say with surety, what's happening now is definitely different than what happened back in the late '90s, early 2000s. If this is a bubble, it looks dramatically different than that bubble. I think I can say that with confidence.

Sam Humphreys:

Awesome. Well, I think it's clear that we need to have more conversations on this, Hill. There's an awful lot to dive into on this subject, but for the interest of time, we are going to wrap up. One thing to leave everyone with is just some projects or key things to look out for in the next 6 or 12 months. Adam, I'll throw this over to you to start with. What are your thoughts on that?

Adam Wilson:

Yeah, sure. In terms of the power side and what these data center companies and utilities will do to address this problem, I'm looking to see how creative some of these companies get in terms of procuring power to power these data centers. I mentioned the deal with the nuclear facility that Amazon made with Talen Energy. I think we could see more of those.

One thing we're expecting to see is data center development begin to shift away from these kind of traditional facilities where there's a high concentration of networks, the Northern Virginia area that Dan was talking about. They may go to where the power availability is rather than where the network availability is, to put it simply. The issue that these companies might run into building a data center is like, North Dakota, I honestly off the top of my head don't know if this is true, but they may have a very reliable, abundant power grid that a data center could build and easily tap into. The problem is infrastructure. They may not have the transmission lines, the substations, everything that you need to power a 100 megawatt, gigawatt scale data center campus.

What we might see is a data center seeing a retired coal plant that already has this infrastructure already in place, they already have the transmission lines, they already have the substations going into it. The data center kind of takes advantage of that as well as, I mean, it's a potential tax credit by repurposing a legacy coal plant, thanks to tax credits from the Inflation Reduction Act, to repurpose some of these retired units into a data center campus. It'll be interesting to see if we see more development towards those types of deals, and if we start to see data center companies, data center plans shift outside our kind of traditional markets for data centers currently and where those might be in the next 5 to 10 years.

Sam Humphreys:

Dan, any parting thoughts from you as well?

Dan Thompson:

The thing that has my curiosity for 2024 is we've spoken in the past about data center developments, it's been in the hundreds of megawatts. Suddenly we're talking gigawatt scale. My curiosity for 2024, is that actually going to come to fruition, or is this just talk? And at what timescale is that going to happen? That's kind of the thing that I'm keeping my eye on is how quickly some of this comes about, which is, there's a lot of things, I think kind of holding that up. We could talk an entire hour about supply chain issues. We could talk about renewable and sustainability issues. There's a number of things that could slow some of that. But for me, what I've got my eye on is the first gigawatt scale campus.

Sam Humphreys:

Awesome. Now we've lost Hill. It's all right. I'll wrap up. Thank you guys. That was such a good talk. There was some really cool things for people to take away. Thank you for your time, and hopefully see you again soon.

Dan Thompson:

Yeah, thanks for having us on.

Adam Wilson:

Yeah, thank you.

Sam Humphreys:

Thanks, bye.

Speaker 3:

This podcast contains insights and data copyrighted by S&P Global. To learn more about our solutions or read additional market research, visit us at spglobal.com.