Ep362: Load Growth and the Future of Capacity Markets

>> Welcome to Currents, a Norton Rose Fulbright podcast. I'm your host, Jim Berger. Today, we are joined by Brent Nelson, Senior Managing Director, Market Intelligence from Ascend Analytics. Welcome, Brent. >> Pleasure to be here. Thanks, Jim. >> So, Brent, you just tell us a little bit about what Ascend Analytics does and what your role is there just at the stage. >> Yeah, So Ascend is a software and consulting firm. We support financial decision-making across the power sector, various timescales, all stakeholders. So, we do everything from immediate term operational decisions, like how do you bid within the market to maximize revenues, to how do you manage risk, plan, maintenance, things like that, to long-term investment decisions and strategy. And that's where I sit, I sit on our long-term market view team. I lead that team and we work stakeholders across the board. So, utilities, product developers, IPPs, retailers, large corporates, financial institutions, anybody that's putting capital to risk in the power sector, we support them in one way or another. >> Okay, that's interesting. I imagine you're quite busy right now with the continuing changes in the electricity markets and the new demand. So, can you set the scene for us a little bit? Why are affordability capacity markets and data centers dominating the conversation right now? >> We've had pretty flat or declining load growth in most parts of the country for the last 20 years or so. And we're suddenly switching into a growth phase and not just a growth phase, but a very rapid growth phase. And we're seeing some of the growing pains of trying to adjust and pivot on and down, and as we're trying to manage this load growth, I think it's really important for folks to remember that we don't have an energy shortage. We've got plenty of energy most of the time. What we are really crunched on is capacity during very specific times of the year, and not even every year, just some years. So, what we're trying to figure out is how do we get the resources that we need to serve the conditions that really matter. And if you have an overnight problem, solar generation doesn't help you, right? So, it's not useful to just look at the gigawatts of resources that are coming online, because those gigawatts might not be solving the problem that's being created by adding new load. And as we're trying to figure this out and get the new resources online, I think one of the biggest things that we've started to discover is that capacity markets in particular are just not well suited to incentivize new entry. And we deregulated 20 years ago and turns out the competitive markets are not as well suited for getting new generation online, particularly capacity resources as we had hoped. And we haven't really seen this problem for 20 years because we haven't had load growth like we're seeing now. >> Okay, so you talked about capacity, and I know you've written about it. Just for the listeners' sake, can you just kind of describe at a basic level what a capacity market is, what it does, what it's supposed to incentivize? >> Yeah, so a contrast capacity market with the energy market. So the energy market is the energy that we need today and tomorrow, and we're to serve load today and tomorrow, right? So that's really looking at operational matching supply and demand in the moment. The capacity market is really a planning tool. It's saying how do we make sure that we have enough capacity available a year from now, three years from now to meet our peak demand a year from now or three years from now. And the reason that you need a capacity market, or kind of twofold, one is that you have this sort of planning mechanism that you need to make sure that you have enough stuff online to meet demand. But the other is that there's a revenue gap that exists in the market, right? So if you imagine a resource that has fixed costs, maybe it has CAPEX that it needs to recover, or maybe it has capital upgrades, or major maintenance, and it needs to recover. If it's that sort of last unit online, or maybe not even comes online, maybe it just sits in the reserve stack during those peak critical conditions, it's only going to recover its costs in the energy market, it's variable costs. And so there's this fixed cost at CAPEX recovery that needs to come from somewhere. And even in the place without a capacity market, which is ERCOT, there still is this revenue recovery mechanism that needs to exist. ERCOT does it differently with scarcity pricing, but you have to fill this revenue gap somehow. And that's what the capacity market is for. That's what it does. The big problem that we've run into and that I think is what folks in PGM in particular, experiencing right now, is that if you need to support new entry, so if you have demand growth, you need new entry. And if you need to support new entry, then you need to pay the cost of new entry. But what breaks down is that if you have a competitive capacity market where that cost of new entry becomes a market clearing price that gets paid to everybody, you end up with an affordability problem. You can't afford to pay the net cost of new entry, which is the technical term. You can't pay that to everybody, or you incur massive costs, and then everybody freaks out, and then you have this massive political backlash. And so what we're finding, and we're seeing this manifest in real-time, particularly in PGM, is that if you live in a world where those high costs are paid to all of the generation stack, that's a political non-starter. The pushback is to immense, and we're seeing this manifest, the state's threatening to leave the market, governance problems, price collars, all these sort of mechanisms that are in place to try to prevent this from happening. And one of the things that we've been writing a lot about is that it's basically politically untenable to have a single clearing market for both new entry and the existing generation. There will have to be a bifurcation somehow, because the political forces are too strong otherwise. And I just want to expand on that a little, make sure I understand it. That's because if you're spreading those costs across all generation, the retail rates for all the consumers go up too much, is that the conclusion? Okay. Yeah, it's for an entry. And you know, data centers are taking the blame. I think it's a little bit unfair. It's a structural aspect of how the market was designed. And yes, data centers are what are causing this right now in PGM, but it's not necessarily the fault of data centers. Any load growth from any source would have caused the same problem. Okay. And as you mentioned, certain markets like ERCOT don't have a capacity market. Others like PGM are having big issues with their capacity market. So can you talk a little bit about how you think this will play out differently in a market like ERCOT compared to a market like PGM? Yeah, so ERCOT and PGM are often the major contrast, because for ERCOT has this energy-only market. They talk about being an energy-only market, but they do have capacity revenues. And the structure that ERCOT has for capacity revenues is actually quite similar to PGM in that there is a capacity revenue incentive that gets created when things run tight. It's really a just-in-time capacity incentive, but it's still a capacity incentive that gets paid to the entire available supply stack. Everybody that's generating is generating into those scarcity prices. And some of the proposals that have showed up in ERCOT to kind of try to stabilize things and provide more stable revenue to generators and stuff like that are also quite similar. And that they would pay everything to the entire supply stack. And so I don't think that ERCOT really appreciates the affordability challenge that would manifest if they kind of continue down that path. There also has actually been some proposals in ERCOT. I was at a conference a while back where kind of one of the BP's market operations at ERCOT was talking about bringing your own new generation requirement for new load, which would basically be a structure that incentivizes new generation outside of the power market. Because then new load would have to pay for a new generation and that would happen separately from the rest of the market. And that's actually quite similar to what's being proposed in PGM right now. When you look at some of the other markets, it's a little bit different. Each market has its own kind of quirks. KISO is, that's California, is largely a regulated market. So if you want to get new capacity online in KISO, there's a central planning that comes down from the regulatory commission. They put procurement mandates onto the utilities and then they have to contract directly with new resources. SPP has a mixture of rural co-ops and regulated utilities and has a requirement that in order to participate in SPP, you have to have an F generation under contractor owned. So again, that new generation is being incentivized outside of the market. New York is directly subsidizing new entry, basically, you know, all the new entry in New York is going to be storage because of the clean energy mandates. And that's being incentivized through the index storage credit. MISO is, that's the Midwest ISO, that's all regulated utilities. They're mostly just rate-based in new generation. So that's how you get new generation online in MISO for all intents and purposes. And then in New England, right now, New England is leaning on state-level subsidies and direct procurements. But if they don't continue that sort of direct procurement and subsidy route, they're going to run this into some of these same problems as well. At its core, you have to find a way to pay for new generation in a way that's separate from the way that you pay the rest of the generation stack. Okay, so given this backdrop, you know, what do you think are the implications for AI infrastructure and other large loads that are coming online? Yeah, you know, as I said before, data centers are getting a lot of the blame right now because they're the big driver of load growth, but, you know, during the Biden administration, all the talk was about electrification, and electrification was going to cause the same problem, right? And so if you have electrified transportation, electrified heating, more electrification in industry, if you were to reshore more industry and cause more load growth, you would cause the same structural shortage that then requires paying for new generation and then you still have to figure out how you pay for new generation. So, as I said, I think there's a little bit of unfair blame that data centers get. They're bringing the market problem to the surface, but the market problem is already there. You know, I think what we're expecting to see is, you know, blue states are largely going to incentivize and subsidize or do direct procurements for their preferred energy resources, which are clean energy resources. So expect a lot of stores to be subsidized in blue states, expect a lot of [BLANK_AUDIO] elsewhere. We're expecting a lot of proliferation of curtailable interconnections. So, you know, that was what PJM just proposed. It was part of what was in SB6 in Texas, SPP and Mysore, both put out kind of curtailable interconnection, pro ideas or proposals. So, you know, we're expecting a lot of, we'll connect you, but if you don't bring your own supply, then we're just going to cut you off when things get tight. So, we're seeing a lot of those kinds of programs, a lot of bring your own new generation type programs so you can sleeve through that new generation with load. A lot of kind of retail rate design structures that are trying to ring fence costs to protect consumers from the cost of new new infrastructure that's required to serve large loads. So, we're seeing a lot of that kind of play out. One of the things that we think is underappreciated is that if you were to take a curtailable interconnection, it's not like you get curtailed that often, right? We're talking about a couple hours a year and when you look at a movement towards winter reliability being a bigger problem, it's not even every year. It's every few years that you get a big enough winter storm to really cause reliability problems. And so, you know, I think if you're a data center looking at a long lead time for a gas turbine, a long lead time for a grid interconnection, maybe you start to look at taking a curtailable interconnection, throwing some behind the meter storage and ride that out through most of the conditions that matter. And I think that's that's going to look better and better. But, you know, where you don't do that, if you're not bringing your own new supply, I think it's going to be hard to get new generation or new demand online. Okay. You talked about the bring your own new generation concept. And is that in front of the meter behind the meter, both can you explain a little bit how that would work? Could be both, right? So, you know, if you bring your own behind the meter generation, then you can take a curtailable interconnection and it doesn't matter because you'll just run your behind the meter when you get curtailed. If you contract with it front of the meter, then you can take a grid interconnection and show whatever market you're participating in that you're also bringing the supply to meet your demand. And that's going to be the path to getting your accelerated grid interconnection, getting any interconnection at all. The PGM, most recent PGM proposal is basically that we're going to curtail you unless you show us that you brought generation along for the ride. So, we're expecting a lot of that to continue to play out. I think the interesting market implication of that is that if you have bring your own new generation requirements more or less everywhere, then you start incentivizing all the new generation outside of the markets. And what's left in the markets is no longer a new generation. It's all the existing generation. And you end up with a very, very different price outcome, which I think is particularly important to folks who are owners of existing generation assets, you know, who are hoping for really high pricing in the capacity markets. And it's just not going to materialize if all that new generation gets compensated outside of those markets. Okay. And you mentioned how you think AIs may be getting a little, getting blamed maybe a little unfairly. And it is really bringing to surface other, the overall issue of just load growth. Have you been seeing load growth from these other sources like electrification, you know, whether vehicles or homes, things like that, or without AI, would we, you know, how much load growth would we be having? Depends on the location, right? So, if you look in Texas, there are locations where, you know, you've got new LNG, you've got industrial facilities going in, new manufacturing facilities. I was at a conference where Center Point, who's a Team D utility in Texas, was talking about the diversified load that they're seeing in their region. And so, you know, in some places, you'll see something like that. In other places, you still see structural declines in load. In other places, you'll still see very mild load growth through the electrification of heat. In some places where you have resistance heating in place, moving to heat pumps actually gives you a reduction in heating demand. So, I don't think there's a single clear story. But one of the, one of the points that we've been trying to make is that even if you have a bring your own new generation requirement for large load, you still have a structural market problem for organic load growth and smaller loads. So, you still need to figure out how you're going to incentivize new generation even separate from bringing your own new generation requirements. But, you know, each region has its own story. Okay. And how durable do you think the load growth from data centers is the load? Yeah, I think this is a much bigger uncertainty than most people realize. You know, right now people talk about this complete, pay any cost world of the training models where, you know, it's hyper scalars will pay anything in the, and the value of compute is so high that the cost of energy is just not relevant. But that might be true today, but you know, you look at, you know, I remember seeing a news article a couple months ago about Microsoft running through their entire token budget in the first quarter of the year. You know, our own company, we do a lot of AI based software work and, you know, we're conscious of the cost and the token consumption when we're doing things. And so, you know, I don't think this pay any cost world is, is going to be forever, right? At some point, you know, these hyper scalars are going to have to compete on cost. And it's not, it's not like, you know, if you, if you're a Facebook user on Instagram user, it's free to you and you are the product being sold to, to the advertisers. But when you get into like AI compute consumption, that's not free. It's not free to the end user. And so there's going to be a lot of room for, um, hyper scalars, data centers, AI companies to really truly compete on cost. And when they start having to compete on cost, then maybe you start to look at different trade offs around latency. Maybe you start to look at models where, you know, you're just not going to get access to AI usage in the middle of the night during winter storm. And maybe a lot of applications for AI don't require being used during the middle of the night during a winter storm, right? And so I think, you know, when we look at the growth of AI as a whole, the ability to time shift, I think is going to be much more important in sort of the next phase. And the value of time shifting is going to be much higher. And that has very different implications for the growth of demand, which is those peak conditions versus the growth of consumption, which is the rest of the time. And I think there will be a lot more room to improve utilization of the grid and shift, shift timing around. You also look at latency. I don't know if you've ever used AI for a complicated task, but, you know, an extra second or two of latency doesn't matter when you're doing a task that requires, you know, a minute, 10 minutes for the AI to complete. So the ability to shift compute around geographically becomes more valuable as well. So all that to say, you know, I, while we're looking at a future with a lot more AI than today, I don't know that the rate of demand growth is going to be the same as we move into a more cost competitive era. Okay, that makes sense. As forecasting and strategy shop, how do you think about the future of power markets in the face of this kind of uncertainty? Yeah. And just as an example, I want to throw out there, I've seen a lot of financing that contemplated capacity revenues in PJM, you know, that was a few years ago. Now it's, it's, I think, two and certain. Yeah. Yeah. And we, we talk a lot about the importance of not just turning the crank on a model, right? You could have a perfect model of the wrong system and you get a bad answer, right? You get put garbage in, you get garbage out. And the capacity markets is a perfect example, right? You can, you can run a model. And even if you model everything perfectly, which you can't 20 years into the future, right? But even if you did model everything perfectly, you can look at the missing money for a new generation. You can calculate what that missing money is and you can use that missing money to determine what the capacity price should be. But that ignores whether or not people are willing to pay that or ignores what the political reality is. And so, you know, we, we talk a lot about the importance of really thinking through what things can you rely on and what things can you not. And what things really do you have to get right? And so, you know, we think about the list of things that you really have to get right in order to have a meaningful view on the future of the market. You really have to make sure that you have price dynamics that look like what the market does first off, right? And there are some locations in particular where models don't reflect the market very well. But you have to make sure that you're reflecting what the market actually does. As we move into this era of load growth, where systems are going to be tighter and we don't have, you know, tons of extra reserve capacity anymore. You have to make sure that you really get your modeling right when you're sitting at the supply edge, right? So, what happens when you have load that's flexing? What happens when you have storage that's setting prices? Making sure that you get those dynamics and that behavior correct really becomes important when that starts to be where you live more often. You have to make sure that you think about how you get new capacity online, right? We've talked about this in MISO for years, right? So, MISO has a capacity market, but MISO also has a lot of regulated utilities. And we have always argued that those regulated utilities aren't incentivized to build new capacity and to keep themselves long and try to convince their state regulatory commissions if they need to briefcase more generation. And as a result, what you should expect to happen is that the capacity market in MISO should be oversupplied and the price should be low. Even though the cost of new capacity is still there, it's still real. It's just being paid for somewhere else. And that that reality is important. And, you know, always has to be thought about. And that's, you know, I think our thinking has evolved on that a lot over the last couple of years. But we've always said that you need to understand how new capacity gets paid for because that's going to determine what what gets built and how it gets built, okay. And talk a little bit about how you approach thinking about market evolution. I mean, obviously you're you're looking at political implications. You're looking at the the cost implications, what incentivizes new generation. So just talk a little bit about how you think about the market evolution. Yeah, you know, we think you have to really be thoughtful about how you separate near term from long term, right? Over the near term, you have pretty good information. You know what policy landscape is the cost landscape is, you have pretty decent information around load and load growth, around transmission lines, where where projects are, and where they're not, and what's in development, and what might come online. But as you march forward in time, your information goes down, right? Your quality of information goes down. And we think it's a little bit silly to be running a high fidelity model 20 years, 25, 30 years into the future, right? Like, why am I simulating at an hourly level the entire power system for 2050, when I don't know where the new projects are going to be, I don't know what load is going to be, I don't know what policy is going to be, I don't know where the new transmission line is going to be. And so while that high fidelity modeling, the high precision modeling is valuable in the near term, its value goes down as you move forward in time. And so we still run an hourly dispatch model 25 years out, but the way that we do it and the way that we think about what it means is different than the way that we think about it for the next five years. And so we lean a lot on what we think are durable truths. So when we think about locational dynamics, I don't know exactly what the grid in California is going to look like 20 years from now, but I'm pretty confident that it's still going to be more expensive to live in San Diego than in the middle of the Mojave Desert. And that has very real implications for where projects are going to get built. I'm very confident that if I have a competitive environment where lots of people are developing projects, then there's going to be a downward price pressure and we're going to see people go to where projects are the cheapest, where they're the most valuable, high value locations become less valuable over time as people discover that input their projects there. So these are the things that I think are more important to rely on as you move further in the future. Long run economic equilibrium is one of our core talking points. We see all the time where people have these forecasts that tell people that they're going to print money on their projects and they're going to earn 20% returns. And if that were true, why are other people not going in and building those same projects and earning those 20% returns? You have to think about what the competitive pressures are going to do. And long run equilibrium is kind of a core backbone of a competitive environment, right? And it's not like there's one developer who can find the valuable spot and then milk all the money out of it. I mean, there's lots of people that see all the same information that are all competing for the same returns. So we try, as we think about market evolution, we try to forecast using a mix of tools. In the near term, we're doing precision modeling. In the longer term, we're still doing that precision modeling. But with a big layer of economic equilibrium, what are the things that we can rely on? How do we make sure that we don't let the policy and conditions of today weigh too much of the future? Today's tariffs are not 20 years from now tariffs. Today's subsidies are not 20 years from now subsidies. We have to take a realistic view on all those things. And in some cases, we're going to be wrong. We don't have a perfect crystal ball for every policy change. But we think you have to take a stance on how policy might evolve, because if you don't take a stance and you say the status quo is the future, you know you're wrong. Why take a stance that you know is wrong? At least take a chance at something that might be right. A lot of that makes sense. And thinking about, you can predict the near future much better accuracy than in 20 or 30 years. And so just I want to ask for your kind of a practical lesson for a lot of the folks who I work with. For example, who are modeling how a project can be financed. What is a good kind of metric for determining like, OK, we can look out maybe over three years, but we're not going to look out over 15 years. What is a good standard for people to think about how they could model capacity revenue or other things like that? Yeah, so I think anybody going to a project finance process, you're going to need 20 plus years, because your project life is going to be 20 plus years. And so you need merchant sale and you need all those pieces. So we have to go that far on time because the exercise mandates it. But we have to think carefully about what we know when we move into that long-term range. And I think in the first three years, you have a pretty good model. As you move into like the five to 10-year range, your fidelity is going down, but modeling still has a lot of value. As you think, 20 plus years, you really have to transition and say, OK, what do I think is a realistic world that I might live in 20 years from now? And if I can build solar for, say, $60 megawatt hour, I think it's pretty unlikely that I can justify power prices, like solar capture prices that are $100 megawatt hour. It's unlikely that they would see that kind of a spread. If wind prices are $50 megawatt hour, it's unlikely that I'm going to see wind capture prices above $50 megawatt hour. So when you think about that long-term, I think there are a lot of things that could be right, but there are some things that definitely cannot be right. And we have to think about what are the things that can strain where prices might go. And so the cost of new generation creates an upper limit on how high power prices can go. I would recommend everybody seek out information that they don't want to hear. The developers in particular are always looking for the highest numbers and the highest value. And sometimes that's us, and sometimes it's not us, right? And so some people come to us because they like our high numbers, and sometimes people don't come to us because they don't like our low numbers. And if you're someone who's got capital at risk, and if it's your equity returns or your debt service coverage, you should really look for the information that might make you worry about whether the numbers are going to be high enough. I know a lot of people have been burned. So I think we just all have to take a critical eye and think about what are the things we don't want to hear. - Okay, great, that's good advice. - All right, Brent, I think we'll leave it there. Appreciate your time. - All right, thank you, Jim. (upbeat music) - You can find us online at www.projectfinance.law or send us an email at [email protected]. Please rate, review, and subscribe on Apple podcasts, Spotify, or your preferred podcast that today was produced by Emily Rogers. Stay ahead of the currents. (upbeat music)