Joe Coletti
Welcome to Powering Sustainable Ideas, a new podcast series from RBC capital markets where we interview the leaders and companies powering the sustainable future. This week, you’ll hear a conversation that took place at the Energy Disruptors UNITE Summit in October.
RBC’s Rob Nicholson, Head of Renewable and Energy Transition, Canadian Power, and Utilities and Infrastructure sat down with John Kousinioris, CEO of Transalta to discuss Data Centers and more specifically how The AI revolution is driving material incremental power requirements. Let’s dive into the conversation.
Rob Nicholson
Good afternoon. Everyone. Welcome. My name's Rob Nicholson. I'm a Managing Director with RBC Capital Markets in the Power, Utility Infrastructure, Group, and I've had the fine pleasure of working with Mr. Kousinioris for many, many years and many of the folks in his organization. And if you don't know, John, John is an accomplished lawyer and now is the CEO of TransAlta, which is has been one of the largest coal fired generators in Canada, but now no longer, no longer. Congratulations. That's fantastic. So our topic today is regarding really power demand, and specifically, we've heard some conversations this morning about AI and data centers, so that's going to be our focus. But John, before we go there, your organization has operations in Australia, in the United States and in Canada, and level set today, if you could, those different power markets, what's, what's the status, before we load on data centers? Where are we in, say, Australia and the United States and Canada, in terms of the grid and how it works and those kinds of things.
John Kousinioris
So, thank you for the introduction. They're very, very different. The three markets, I would say, you know, the US is such a large market that it's almost impossible to talk about in a sort of a generalized way. But we have significant operations in the Pacific Northwest. It is growing rapidly, population growth, industrial growth, discussions around data centers, a very strong push towards renewables in that part of the world. And increasingly they're having reliability problems. And I'll come back to that sort of thematically, because although each of the three jurisdictions are different, the challenges that they're facing are in many respects the same. Western Australia is a relatively small market. Strong mining, a lot of behind the fence, remote generation. They have more of a regulated market, so they have capacity pricing, energy pricing, and they too are struggling with an increase in renewables at the same time, trying to make sure that their grid is reliable. And maybe I'll just talk briefly about Alberta and Ontario. We tend to think of Ontario as sort of a class high regulated market and a bit of a contracted but an interesting market, and also looking to procure quite a bit of generation right now, I think that's something that's taking place there. And interestingly, increasingly, at least, the theme seems to be sort of technology agnostic, in terms of where they're coming in. And then Alberta, which is our principal market, our core market, kind of a unique energy only market where there isn't sort of a separate capacity price that is paid to generators. So everything comes out from a compensation perspective in the hourly price that takes place and a market that's undergoing quite a bit of regulatory change, I would say, probably more regulatory change than any of the other markets that we're in. We had a renewables pause that is now over, with a little bit more scrutiny around everything from reclamation to participation and regulatory processes to certain parts of the province being sort of offside in terms of renewables development, and are now in the middle of a market renewal process, and that'll take a little bit of time, I think, for us to work through. So quite a bit of quite different but thematically, all three jurisdictions are broadly struggling with how they're going to balance three components. And that's affordability, reliability and clean. Things have to be decarbonized, and all three markets, to a greater or lesser extent, are struggling with meeting the needs of those in a balanced way, and I don't think anybody has a good solution for it, and I think there's a lot of work to happen, and unfortunately, I think we're going to sort of stumble through the evolution, I think, or the energy transition that's taken place in all three markets. Are struggling to meet the needs of affordability, reliability and decarbonization. A lot of emphasis on decarbonization over the last number of years, at times, I think Rob, if we're honest, at the expense of affordability and reliability. And we see reliability problems in Alberta. We see reliability problems in the Pacific Northwest, and we see reliability problems in Western Australia, and I think they're going to probably become, I hope I'm wrong, but I suspect they're going to become more acute over time, rather until
Rob Nicholson
Maybe just summarizing, I guess what I've heard from you is they're working, but maybe they're at a level of stress that you could put them on a tipping point. Is that fair?
John Kousinioris
I think it's probably fair.
Rob Nicholson
Yeah, okay. So the increase in AI between 2015 and 2019 more than doubled, yet the power demand did not go up at all. Primarily that was driven by data centers actually becoming more efficient and the use of electricity. But the belief is that that curve is sort of declining in their ability to increase efficiency. So the US in 2022, 3% of the electricity in the United States was dedicated for data centers and AI uses. And in 2030 which is only eight years later, it's going to be 8%
John Kousinioris
I just saw 12.
Rob Nicholson
Okay, well, there we go. Nobody really knows, but it's a big it's a big number. Right? To go from three, let's say eight to 12. We'll go 10, something like that. So when we think about that, what like what does that mean? You own a power generation company? Generation doesn't necessarily directly plug into an AI data center. So paint a picture for me about, if we have an increase like that, what does that mean in practical sense.
John Kousinioris
So look, I think the numbers that you talked about are actually right. So I think if you were to look at sort of mid, the mid 2000s so from about 2005 frankly, till the late teens, load was pretty flat. And in fact, in some jurisdictions had actually declined, which is something we saw happen in a bunch of places all over the world. Today, we're hearing people talk about, you know, three, three and a half percent per year increase in load and to your point, on data centers. The last bit of information that I heard is, if you go sort of from now to 2030 you're looking at somewhere between 20 to 25% a compound annual growth rate in the demand that data centers are going to have. So it's, it's sort of staggering, and I think what that translates, and I think you can work by analogy to Canada, if that is right, from a prediction perspective, you're going to need somewhere like 50 to 60 gigawatts of new generation installed in the US, kind of conservatively, which just to give people a bit of perspective, is sort of two and a half times to three times the amount of installed capacity that a province like Alberta has, like it's not an insignificant sum. And the latest numbers I saw on that Rob were, I think that's about a $500 billion spend, roughly to get done. And that excludes the data centers, like the equipment that go in a data center, and it excludes all wires, so no transmission, no distribution, literally, just generation and kind of the blocks that you would put the data centers in. I don't know how we do that in that time period, candidly, and I don't know how we do it, in a way that continues to ensure that kind of our broad decarbonization goals continue to evolve. And I think you're seeing it. I think you're seeing it all over the US. It hasn't become as acute in Canada, but I think if you look at what's happened in the US to the extent data centers come to Canada, I think you could see sort of analogies taking place. But it's not just North America. I mean, if you were to look at places globally, Singapore and Amsterdam, kind of that area in Central Hall, when we're also big hubs for data, they're also kind of going like, we can't, we can't deal with the demand and kind of tie it in, you know, and deal with it. So it's, you know, it's super challenging, you know, if it takes, I don't know, a year to permit something, if it takes two years or three years to get the hardware you need, from a supply chain perspective, to actually build it out, then you need to build it. There's a shortage of labor, candidly, all over North America to get it done. So just from a path of just being honest about it, getting from you know where we are today, to where, in theory, we need to be by 2030 I don't know how you do it. I know people don't want to hear that, but I'm just telling you, I don't. I just don't know that it can be done without breaking things or resulting in astronomical costs. Like it takes, it takes three years to get a turbine today for a gas plant, it can take sort of two years to get, you know, wind turbines, you know, sorted out, and then you've got to get them delivered, and you got to build it. And I think permitting takes longer than ever before. It's like seven or eight. So I just don't know how you go from, you know, where we are to where we kind of need to be.
Rob Nicholson
So we're going to need 50 to 60, at least new facilities in a five year or seven year time frame, is the prediction. So I guess the question is, it doesn't sound like that's possible from your perspective, but what about if we combine resources from a gas facility, some wind, some solar. Like, how do we how do we put the puzzle together? Do we need to go all in on one? Or do we try everything? Or what do we do?
John Kousinioris
I think you're going to need all of the above to kind of meet it. Because I don't think it makes sense to sort of say, ‘Oh, we're going to ignore kind of our renewables targets, and we're just going to go with thermal to make sure that we can provide kind of a stable grid to permit all of that to take place.’ So I think it will require balance, whether it's on the same project or not, or whether it's, you know, more broadly speaking, in the grid. So I think it requires, you know, all the comments on the investment that's required, and kind of the attributes or characteristics of it are going to continue. And it's not just data centers, right? Like, generally, we've seen a ton of digitization take place. We're in the very like, we're in the infancy of EVS, like, you got to be, you know, just beginning in terms of demand. You've got Ais that are coming. Cloud computing is also significant. And what's interesting is, the more we do, the more it actually creates demand. If you see what I mean, it isn't sort of linear. But the demand is actually difficult, and the efficiency that comes from chips and whatnot is actually slower. In other words, the time it takes for them to improve the efficiency is taking longer. So there was a time where, every year.
Rob Nicholson
Moore's law.
John Kousinioris
There you go, see, I knew you knew all this stuff. So all of this is making it harder to basically do, but I just think it's hard to do, like I you know, we have two, you know, basically three in the Western world, wind generation OEMs. Three. You know, I mean, it's Vestas, it's Siemens, it's GE. That's it. They're busy, like they're operating flat out, you know, in terms of turbines. I just don't know, you know, solar is, I think a little bit you have a more flexible, I think supply chain and I think the costs have come down, but it's challenging. And just to give you a bit of a sense, and you and I talk about, and I talk about this all the time, you know, I remember three or four years ago, four years five years ago, we used to think about, you know, how much it would cost to build a wind farm, for example, and it was about a million and a half dollars a megawatt, roughly speaking, that's like 2.2 million today. So, you know, that's, in part, supply chain challenges, but like the inflationary impact, Do I think it's going to come down? Sure, but it's, you know, it's popped up, right? Solar has come down. Gas is pricey, so I don't know, Rob, how you do it. I wish I did
Rob Nicholson
You're not painting the rosiest picture I've ever seen. John.
John Kousinioris
No, look. I let me, let me. I think you need to have a clear eyed I think there's a difference. So it's easy to kind of say we're entering into this unprecedented and unbelievably sort of exciting period, and we're going to be able to decarbonize and be able to meet all of the electricity demand that we need, and I think it's going to be bloody tough, like I hate to say it. I think it's going to be hard to do it. Think it's gonna be hard to do it in an affordable way. And I think, you know, creating kind of the policy environment that gives you the kind of certainty that you need to be able to make those kinds of investments is also challenging. So I like, I so am I excited as somebody that leads an independent power producer, hugely excited. I think the opportunity set is huge, huge, like unprecedented, kind of a, you know, a once in a generational kind of opportunity. But I think it's going to be challenged to kind of hit all of the holistic targets that we have.
Rob Nicholson
So when an Amazon shows up at your doorstep or Microsoft, what are they asking of you as a power generator? Would they come to the door and they say, ‘We want to build a 200 megawatt load next door to this location, and can you supply us power?’
John Kousinioris
Yeah, it's, it's changed. And look at Amazon, you know, Microsoft. These are customers that we have in our company. And I would say it's evolving right now. And maybe I won't speak about them. I'll talk about sort of the more generic kind of discussion that we're having. I would say two and three years ago, it was more about, you know, can you build us a wind farm? Can you do it on time? What's the time frame going to take? We've got targets broad. It was more macro, decarbonization. Suppose I've got a specific data center I need you to sort of power. So they were more concerned about cost, and they were very much concerned about, will it actually be built? Because there were so many examples of people agreeing to build out projects, and they didn't actually materialize over the course of time. So they were quite picky, I would say, about who the counterparty was, and did they have a track record of actually delivering the power that they promised to deliver at the appropriate time. It's much more nuanced today, I would say, certainly from a data center perspective, at least the kinds of discussions we're having, and our discussions are very much Alberta focused discussions. They're not in Western Australia, and they're not really yet in the Pacific Northwest. And there's some challenges in the pack Northwest that probably make that, at least for us, a little bit more challenging. But when I think about Alberta, it's just speed primarily. Like, I have to tell you, the decarbonization discussion isn't driving the decision making. It's sort of time to power, candidly, would be the number one thing because, because the reality is today, the development cycle, the time cycle for development, is longer than the time cycle for the development of the data center, if you see what I mean. So they're challenged from, from a timing perspective. And I think the other thing that's interesting about data centers is sort of the economic model has kind of changed. So, you know, they're less price sensitive about the power. So, you know, it's not like it's sort of an industrial concern where power might be 60% of the process. Let's say you're making hydrogen, and power really matters, because it's, you know, 80% of the cost, or whatever right is power. With a typical data center, the cost of electricity in that sort of construct is like 20%. In other words, it's not an irrelevant it is very much relevant. But your but your project won't, you know, the price of power isn't going to make or break right kind of, you know, making your investment decisions, so they're willing to pay a bit more candidly, in exchange for just having access to the power in a kind of a real time sort of way. That may change over time. But I'm just so that's so the green part, at least. You know, speaking about our perspective, it's not, it hasn't been, to date anyways, in the context of Alberta, a significant driver. And look, we're uniquely able to say, look, we've got wind, we've got hydro. So there's ways that we can use, kind of our portfolio to kind of decarbonize the offering. But they want, but, like, they want 99.9% reliability, and they want speed and and I don't know how you at least in Alberta, we can't provide that in a green way right now.
Rob Nicholson
So there has been some resistance. We're hearing more about local resistance to building out these big data centers. People are starting to realize that it actually has an impact on their retail price of electricity at home. It has an impact on water supply. It has multiple impacts. And if you think about, there was a discussion earlier this morning, a little bit about how latency isn't necessarily all that important anymore, or not as important. So if you think about again, the big constructors of these data centers, where do they go? Do we see more here in Canada? Do we see it coming across the border in Mexico?
John Kousinioris
A regulated kind of construct, I think is becoming harder. I think, you know, when you look at the transmission build out, when I think you're looking at water use, I think you touched on all the kinds of things. People are very sensitive to pricing. They're also pretty sensitive to the time frames. I mean, there's an active discussion right now in parts of the US where some of the utilities are saying, Look, I need you to give me some sense of certainty so I can build out this transmission line. I'm going to need, like, 10 years or 15 years of certainty to actually make that investment. And there's a reluctance to kind of give that level of assurance. Well, I don't know how they're going to make the investment, because what they're worried about is people leave, and then you've got retail consumers holding the bag to pay, you know, all of the infrastructure that's basically built out. So it's challenging. Like, people are concerned. They're concerned about what it means to their power bills five years from now. They're concerned about like, does it actually create any jobs or not? Not, not really, once it's built, even though they're at least in the States, they're kind of estimating it's sort of like up to half a million people short to build out the demand. It's but once the construction is done, it's not like you're gonna employ like 500 people at the at the facility, like it kind of runs, you know what I mean, at the plant. So I think it's gonna require more engagement and as things. Get more congested. If I can put it that way, I think it's going to become harder still.
Rob Nicholson
So your organization has been building siting infrastructure for many, many years. Does the model change here then where maybe the data center guys who are, frankly tech operator runners need to partner with you more as a sighting location. Do you see a almost a change in the business model where you were, frankly, sort of a supplier of electricity wholesale into the grid to now we're more connected, but maybe even another step connected, where you're planning together, simply because that's what has to happen?
John Kousinioris
I think the answer is yes. I think it's still evolving. You know, I think one of the things that's been interesting for our own company is the value, I would say in the legacy assets that, you know, five years ago, people would say, ah, you know, they're a little bit older, and I don't know how long they're going to run. Are they really needed in the market? I think that dialog has changed. Given the challenge with building new the existing steel on the ground has a lot of value, I would say. I think the kind of relationships are changing. I mean, look, through three mile island is they need power, and they needed to run reliably. And, you know, that's power that doesn't emit too like, you know. So there's a lot of pluses to basically getting that done. I think with some of the larger companies, you might see kind of a backward integration of the supply. So it wouldn't surprise me if some of them buy power generation, or they, you know, we hear examples of people buying an industrial facility that and they kind of decommission the industrial facility because it's set up to the grid. And rather than having the power drive a steel plant, for example, it'll end up plugging in and, you know, driving their data center, just to kind of circumvent the time frames that it costs to get stuff done. So I think all of that is sort of nascent. And, you know, it'll take some time for us to see it kind of, kind of play out.
Rob Nicholson
So stepping back from AI for just a moment, we've seen Hydro Quebec announce, they need 30 gigawatts at least over the next 20, years, or something like that. They're not talking about AI necessarily. They're just talking about things like EVs, things like taking your home heat and electrifying it, some of those kinds of things. And you look at Ontario, we had a premier who was, no way are we building any more wind farms in this province, and all of a sudden they're back on the agenda again. We've got OPG building a new nuclear power plant. I guess where I'm going is, you've pitch painted a picture, which means we need every resource possible. You've got TMI coming back online. I don't know you. Tell me, do the western provinces build a nuclear power plant anytime?
John Kousinioris
BC is short. I think you know, Manitoba, for all of the hydro that it has is talking about it being potentially short. Think Ontario is short. And these are all run by Crown corporations, right? So, so when you look at that, and you look at where sort of new load can go in, like if you're a data center or whatever, Alberta becomes, frankly, one of the best jurisdictions to be in, just because it's sort of an open model. It's hard to have, I think, it's hard to have kind of public utilities, you know, building the infrastructure for these kinds of things, when really they should be worried about, yes, it could be part of industrial policy, but they're fundamentally worried about making sure that retail consumers, people that are in this room have reliable and affordable power, like that's their number one mandate, as opposed to kind of while they're already short going in and, Candidly, in Alberta, we have a bit of a supply and demand imbalance. So as things go by, I think, you know, we've got more supply relative to the actual demand or load that we see in the province, but the reliability of the jurisdiction, I think, isn't quite where we'll need to be. Now that we've had a ton of renewables come in, which is great. They've helped decarbonize, but it has, to a certain extent, pushed out gas, in our view, because when it's cold here, like there's no wind,
Rob Nicholson
You've got electrons, but no capacity.
John Kousinioris
Zero capacity, and the days are short, so solar is going to help, but it's not going to be the answer to the question. And so what fills in the gap? You can't just say, oh, you know, sorry, we're gonna shut it down for like six hours, and you'll be blacked out when it's cold. So there still continues to be a role. So on nuclear to get it's hard, like, I gotta say, you know, I see what SaskPower is kind of doing and kind of moving it along. You know, when I think of Alberta, which doesn't really have a government driven entity, which is, which is drawing the development of these kinds of things, like for a company like ours to get into a nuclear development piece, it's super difficult. I mean, it's, it's probably 10 years, maybe 15 years, to get it permitted and built. The costs are super, super high.
Rob Nicholson
And the risks are significant.
John Kousinioris
So you need to make sure that the revenue stream for that setting aside all the technological, regulatory permitting needs that, that kind of the revenue stream to kind of cover that four or five times more investment, you know, roughly speaking, is enough to kind of justify doing it. It needs to be certain. So, so how do we do that in Alberta with, price clearing every hour? Like you can't make that investment here, like you'd need to, we would need to redo the I personally think you would need to redo the entire market construct to permit something like that to take place. Really, really challenging,
Rob Nicholson
Right? So from a structural, market structure perspective
John Kousinioris
And technological risk just and writing the check, it's big. It's big, right?
Rob Nicholson
Right, one question, though, but we didn't dig into it too much, but we did talk a little bit about policy. And what are, if I could ask you, what are two or three things that either our government, our regulators, or even just the public should be thinking about as far as actionable items to enable this?
John Kousinioris
So look, one I would say I think we need to do a better job of feathering in the renewables with kind of generation. And attributes that provide reliability. Like, I think it's a bit it's a bit of a free for all. I can put it that way, and I think the market is wonderful, but I think sometimes we have an overly large belief in the market, and there's so many distortions coming in that I'm not sure the market actually works as well as people think it does in meeting all of those solutions. So I think, setting reasonable decarbonization targets with an eye to reliability and affordability for consumers is critical. So maybe a little bit more of a planned or kind of, this is kind of a pathway that we want from decarbonization. It's a realistic one, and this is kind of what we're seeing being permitted. Do you know what I mean? We're going to make sure that there's, you know, these assets are here to provide affordable reliability, to make sure the system works well. So I think that would be one. Two, I think I think Alberta is not bad, but I think, generally speaking, permitting interconnections, all of that takes a hell of a lot of time, a lot longer today than it did before. So can we streamline that without reducing its effectiveness? If you see what I'm saying, the other thing that is really important from my own perspective, is we need to kind of have a stable regime. So being in a place where it's changing all the time where there's, you know, and don't get me wrong, I think the restructuring stuff that's going on in Alberta is stuff that we need to do. Like, I'm not but having a stable regulatory just stability is so important. Like, if I knew what the carbon price was going to be in five years, as opposed to guessing if there's going to be an election and it'd be undone if I knew what kind of the interfaces between the federal government the provincial government, or I knew what the route, you know, what the just, what the market construct is going to be, and had a sense of what pricing was going to be, we can make investment bets then, because we kind of have a sense a pathway, sort of stability, but the more turmoil there is in that you sort of step back and your shareholders rightly say, like, ‘What are you guys doing?’ You know, I mean, it's hard to make those kinds of investments. They're risky to do, or the return you need to make those investments is disproportionately high, which means consumers have to pay more to get it, to get it done, not because there's a windfall that the companies are making, because it's the risk, and because he refuses to give me loans on good terms. And that's really what it what it all comes down to it.
Joe Coletti
Thanks again for listening to powering sustainable ideas. Brought to you by RBC Capital Markets. Please remember to subscribe to get more great content and be alerted about future episodes. This episode was recorded on October1, 2024, If you'd like to learn more or continue the conversation, please visit rbccm.com/energytransition. See you all next time.