Speaker 1Hi, I'm Bruno Alves, and welcome to the Infrastructure Investor Podcast. In our last episode of 2025, I sit down with Doug Kimmelman, founder and executive chairman of Energy Capital Partners. Doug is a power markets veteran, so we delve into the U.S. electricity market, which is experiencing what he called a shining moment, driven by unprecedented demand growth after 30 years of stagnation. Key drivers include the massive power needs of artificial intelligence and data centers, along with LNG exports, the onshoring of manufacturing, and the expansion of electric vehicles and crypto mining. This surging demand, which would require as much as 400 gigawatts of new capacity when you factor in retiring coal and nuclear plants, is rapidly transforming the sector. We discuss how the market is prioritizing slowing the retirement of existing generation, brownfield expansions, and fast-to-deploy, solar and storage, while the reliance on natural gas increases despite turbine bottlenecks. We also explore ECP's stellar calpine exit and discuss how the presence of hyperscalers with strong balance sheets is reshaping the contracting environment, leading to longer tenure, highly financeable power purchase agreements. Hi, Doug. Welcome to the podcast. Thanks for having me. Looking forward to our chat. So you did an interview with Infrastructure Investor earlier this year, where you said something that I think really captured the zeitgeist. And you said, and I quote, electricity has never had a shining moment like this since it deregulated 30 years ago. And I think we have to start there. Can you unpack what you meant for the benefit of our audience, and particularly what this moment means for investors?
Speaker 2Absolutely. And I think that is a good place to start. Well, if we go back a little bit of a history lesson here that many know, but not all do. First of all, let's start with deregulation, and then maybe we'll go to demand for electricity. But we started in the mid-90s, spreading across the country, a wave of deregulation, airlines, banking, insurance, natural gas. And it was just a look at where consumers are spending their most money. And it didn't seem logical that the most, the most efficient model for a regulated monopoly utility was to buy power generation for itself, for its own customers. Wouldn't it be more efficient if the customer or if the transmission and distribution company, the T&D company, had the ability to shop the market for electricity to bring costs down for consumers? Now, the whole country didn't adopt it. I'll say maybe 60% of the country adopted it, in states like California and Texas and Ohio and Pennsylvania. But I think it worked. And now that brought on this whole independent power industry, deregulated power generation industry. And so that's one important part of history. But since that time, we basically have had little to no demand growth for the past 30 years. And what's upon us now is demand growth kicking in, kicking in at very high levels. And that demand growth coming at a time where we have utilities, those that are still regulated in some states, particularly the South, and the deregulated power industry, coming a little bit at who's going to win here and who's going to be building out our new capacity. So it really is a shining moment where AI, probably the key driver of GDP growth in America, no electricity, no AI. And I think it was a shock to people that you can't just ball up your utility and get electricity.
Speaker 1Yeah. And I just want to put a pin in AI for just a second, because I wanted to use your recent exit, of Calpine, to actually illustrate how much the market has changed in just a few years, right? And so, you know, just to frame this quickly for the audience, Calpine is one of the biggest exits in the US energy market of the last decades. You sold it to Constellation Energy for nearly $27 billion. It has 27 gigawatts of natural gas generation spread across the US. But what's really interesting, right, is you bought it in 2017, at a time when really long-term electricity growth forecasts were pretty sluggish. You placed it in a continuation vehicle in 2022. And I think the consensus was still around sluggish growth, maybe with a bit of post-COVID pickup in the mix. And then you sell it in early 2025 when the forecasts are booming, as you just rightly referenced. So talk me through this journey a bit. And specifically, what was your original plan for Calpine? And at which point did you realize you had this stellar asset to capitalize on this massive growth trend?
Speaker 2Well, we always viewed it as a stellar asset, which is why we bought it. Relatively new, well-located, efficient, largely natural gas-fired generation, concentrated in some of the best markets in the country, especially Texas, California, some in the Northeast, and a world-class renewable asset. I call it a baseload renewable asset, geothermal, called the geysers. I believe the largest geothermal asset in the world. But in 2017, this word "energy" was a dirty word, and the public markets, everything having anything to do with energy traded off. And I think the world was more focused on, we don't like oil and demand is going down, but we hardly use any oil in the United States for power generation. It's got nothing to do with electricity generation, but the sentiment wasn't there, and the stock traded off pretty significantly. But when we looked at the free cash flow yield, incredible amount of cash generation relative to very modest. Just maintenance cap X, that we saw an incredible return just by running this business as usual and throwing off an enormous yield. That was the base case, but the base case was way exceeded, and way exceeded before even AI came along, that we did a lot of things to dramatically increase the cash flow, increase the EBITDA of this business. A lot of capacity expansions, a lot of growth projects brought online, which I believe is certainly the largest. The largest in the US battery storage project, maybe the largest in the world, 2,700 megawatt hours, 680,000 homes. So that added a lot of EBITDA. We expanded the capacity at the Geysers, focused on a pipeline of carbon capture projects. We de-levered the business. We hedged it maybe more appropriately, added a lot of new contracts. So we did a lot to grow the cash flow of the business. And then along comes demand growth. And demand growth, everyone's focused on AI, but let's think about a whole bunch of other things that are part of this demand growth story. LNG exports, right? The US now exports 30% of its production. I never thought the day would come the US would be an exporter, energy always an importer. But these LNG export terminals, hugely electricity intensive, right? Because we have to freeze and liquefy that natural gas. So that's a big one. Didn't see this wave of onshoring of manufacturing coming to the US, I think there's something like 200 manufacturing facilities under construction right now in the US. Hugely electricity intensive. Electric vehicles certainly not going away. Cryptocurrency mining, the load in Texas for crypto mining, the electricity load, greater than the entire load for the city of Houston. So a lot of pockets of demand piling on and AI piling on. And if you're the low cost producer in well located, where we can do, let's just say relatively quick brownfield expansions. Like some of our capacity that fits a data center, obviously that interest, you know,
Speaker 1comes in. Right. And now we will get to the AI build out and the massive demand driver that it is. AI is driving this market, we all know it, and power is at the heart of it. Lots of numbers thrown around these days, I'll put some in the mix just to frame this again. The recent Brookfield report predicted data center capacity is projected to grow from circa seven gigawatts in 2024. That's more than the tenfold increase. A lot of this will be rolled out in the US. So as a power markets veteran, tell us what you think the power mix is going to look like to serve this massive need. Perhaps focusing on the power sources that you feel can be rolled out the quickest to meet this demand.
Speaker 2Good question. Well, let's triangulate the demand real quickly because there's forecasts all over the place. But if we start with right now, data centers. We are about 2% of electricity demand in the US and the US roughly has 1200 gigawatts of power generation capacity. So 2%, about 25 gigawatts going to that. The projections are anywhere from 10%, 12%, 15% of power demand going to that. So we need in the neighborhood of another 100 gigawatts for the AI, but don't forget that laundry list of everything else I just mentioned with anchoring and manufacturing and LNG. Potentially there's another 100 gigawatts as well. And then there's one, believe it or not, that's bigger. 70% of the country was serviced by nuclear and coal maybe 15 years ago. We're now 40% nuclear and coal. It's old, especially in the case of nuclear, a lot of it built in the 60s, 70s, and 80s. So they're shutting down. We still are the world's largest generator of nuclear power. We had 100 or so operating reactors. I think we're down in the low 90s. And then coal is part of the mix. Environmental challenges. We may have to run longer than we thought, but we might be losing another 100 to 150 gigawatts of coal and nuclear because of age and environmental. So as much as 400 gigawatts, that's a 33% increase in the total capacity. This is not going to happen overnight. There is no something like just, I use the word just, just build it. That doesn't work in this industry. So we've got a decade long process to get ourselves rebalanced. So your question, where's it going to come from? Well, the easiest one, I'll use the term net megawatts. Let's slow the shutdown of coal and nuclear. And I think we're starting to see that slow the shutdown of coal and nuclear because if that's the biggest piece that we might lose 100 to 150, let's slow down the loss of that. So that's top of the list. Second is brownfield expansion, which is a real benefit to an incumbent, someone like a Calpine, a Constellation. If you have hundreds of sites around the country, probably a lot easier because you already have a gas pipeline coming in. You could add some capacity to that. That can happen quicker because these hyperscalers, they want three things. Number one is they want speed. Give me that power in the next two to three years. Number two, they want scale, not 50 megawatts. We want, we want 500 or a thousand megawatts and location matters. It's, you know, nothing against I'll, I'll pick on what state should I, I won't pick on any states, but we can think of some states where there's not a lot of population, but you know, Northern Virginia has been a big hub, Texas and Dallas, a big hub. I don't know, I don't know if there's a big hub outside of Chicago, Ohio, Pennsylvania. So location matters as well. And so those brownfield expansions in those areas, number two, number three, believe it or not solar and storage. And we should talk a little bit about renewables. The quickest and cheapest power that we can bring online is solar can bring it online in his neighborhood of 12 months. And not withstanding, there are tariffs, not withstanding subsidies are fading away. It is still cheaper and it's cheaper than natural gas. The reason why it's cheaper than natural gas because natural gas of not the price of natural gas, but the price of building a natural gas fired combined cycle peaker has gone up. The equipment has gone up, the gas turbines and short supply have gone up. So it's that much more expensive. So on a relative basis, solar and battery storage looks pretty good. And that's going to be a piece of the equation. The fourth place is a big one. It's new natural gas build. But the problem with new natural gas build is going to take some time. It's going to happen in large scale, but it's going to take some time. And I put last, some people put it first, but I put it last as new nuclear. I think we're a little overbought in terms of the excitement, but either new SMR technologies or building, you know, large scale nuclear plants around the corner. I don't think any of that is around the corner. I love nuclear. I love its baseload. I love that it's green, but we got a good 10 years before we see any needle move on that one.
Speaker 1Yeah. Well, let's, let's quickly unpack, a few points actually there before we move on. One is really this combination of solar plus batteries and how that mashes or marries with what has been, you know, pretty strong aversion, I feel on the hyperscaler side from anything that sort of affects the reliability and continuity of the power flow that they need, right? Are we at a point where a combination of renewables and storage can really play a significant role here, or they always have to be, you know, paired with something else? Give me your thoughts on this.
Speaker 2Well, I think if we go back a year or two, the hyperscalers were of a mindset, we will not be interrupted. We must have firm power. And we use the term five nines, 99.999% reliability. One of my colleagues yesterday, we were laughing about it. It might now be one nine and four eights, 98.888% might be good enough because I think reality is sinking in. So there have, you know, there are two potential societal complaints to the AI build out. Number one, you're taking away my electricity and you're going to drive up rates for consumers. And number two, you're going to do this on the back of natural gas, fossil fuels, you're going to drive up carbon emissions. And I think the hyperscalers are obviously very attuned to those issues. They don't want there to be a backlash. If you look at the whole operating expense of a data center, you know, electricity isn't the main driver. Maybe it's 10% of the cost. And if they have to pay a little bit more, they will. So I start with consumers should relax. We all know this will not end well if we're taking power off of the grid and taking it away from the retail consumer. The regulators eyes are open to that. The politicians eyes are open to that. So there's going to be, you know, this back and forth of behind the meter versus in front of the meter, which we can talk about. And I think a lot of it, you know, is going to be behind the meter or at least a contract where the hyperscaler is going to be paying more than the retail's consumer for that power. So that's a piece of this. But I think that the hyperscalers on the question of the carbon emissions, I think they know they're going to have to buy a lot of natural gas fire generation. Most of those, the Microsoft's, Amazon, Google's, are social conscious. Most of them have committed to a net zero outcome for all of their business lines. And that means they're focused on things like carbon capture, things like investing in renewable projects and nuclear, things like that. And I think that's going to provide an offset. So I think Amazon's the largest purchaser of renewables in the country. Not so much to serve the data centers, but to earn an offset. We call it a virtual PPA, right? They'll invest in a wind project in Colorado and they'll agree to a fixed price, $70, $80 a megawatt hour. But that power is going back onto the grid. It's not going to them, but it gives the developer a nice 10 year contract, gets it built and Amazon earns an offset. So I think we'll see it playing there. But on that reliance, I think it's going to have to be a bit of a reliability question. If we're going to have some things that are behind the meter, well, there are days where we're going to have maintenance and it's going to have to be shut down and there's weather events and unexpected outages. And the hyperscalers do, maybe not five nines, but up there in the high 99% of reliability. So supplementing with solar and storage, supplementing with gas peakers, or maybe even cutting a deal, which we've done projects like this, cutting a deal with the. Independent system operators with the regulators that let's have a trade-off here. If there's an ice storm in Texas, Google, are you really going to turn your back on people dying in hospitals without electricity? You're going to be willing to be interrupted. So let's, instead of wait for that fateful day, let's cut a deal now. We're willing to be interrupted for certain emergency situations as defined for certain periods of time. But in return for that in the shoulder months, in the spring and the fall, when there's excess electricity on the grid, maybe we can take it back from the grid at that point to deal with our maintenance. So we don't need as much backup. And then me as an investor, I don't really want to own a power plant that has no grid connection because I'm no expert in technology. But in 10 years from now, is a data center obsolete? Well, there's a fair chance that it could be. And I don't want to be stranded like the data center and have a stranded power plant with no interconnection. And so perhaps it is a good idea to have a power plant that has no grid connection, which is a good idea, not to be drawing power from that grid every day, but to at least have that interconnection for the future and for emergency events.
Speaker 1I actually want to go a little bit deeper into, you know, how this AI demand is changing power markets and the presence of these large balance sheets. And I want to maybe dig a little bit deeper into contracts, into things like PPAs and offtake agreements. So tell me how this combination of power scarcity and demand from the hyperscalers is kind of changing the game. Is it contract 10 years getting longer? Is it long-term PPAs for types of assets that previously would be merchant only? Is it both? What are you seeing?
Speaker 2Yeah, we, you know, from investors, we sometimes get the question, why is this cycle different? And if we go back to previous cycles, we've had boom and bust in the world of deregulated power generation, that when prices got a little bit high, or capacity payments got a little bit high, a wave of developers came in and we're all going to build our 750 megawatt combined cycle. And an overbuild happened and it went the other way. And it was a merchant market. You didn't have a 10-year contract. Maybe you had a few-year hedge with a financial institution, but that was about it. But the big difference now is that we've got the best balance sheets in the world that are offering up 10-year plus contracts. And you should be able to earn a very reasonable return just over the life of that contract. Now, you heard me say earlier, I do care about my terminal value. And after 10 years, the data center happens to be, you know, technologically obsolete. I do want to have some chance of selling that power back onto the grid. But that contract is the pot of gold. We have the best balance sheets in the world offering that up. And surprisingly, there's been a lot of MOUs, letters of intents, but real contracts signed far fewer than you would think. And it's working through all of these details of pricing, of interruptibility, those type of things. And those incumbents that have existing sites where they have access to gas, access to transmission interconnection, have excess acreage, we call it powered land. That's the big thing that everyone wants. It's powered land. Those are the ones that are making progress with the hyperscalers and getting these contracts in place. They're highly financeable, these contracts. They take away the merchant risk until the end of that contract term. And that's really the rush of figuring that out.
Speaker 1The bottlenecks we are hearing, and this is when it comes to new build, gas generation, and the bottlenecks with turbines, et cetera, et cetera. How bad is this? Because when I think about, we're talking about AI a lot because they have a lot of muscle, so to speak, but you've pretty much highlighted all the other demand drivers that also want electricity and also probably want generation. So how can we marry this with a bottleneck or how bad is the bottleneck maybe is a better way of putting it?
Speaker 2Yeah, well, G, Vernova, Siemens, and Mitsubishi hold a lot of cards. I would say if we had to pick the number one choke point in moving this forward, it's gas turbine availability. We've been very fortunate. We, about a year ago, bought our own turbine manufacturer quietly, but I can't say it'll be our best deal ever because I've got Calpine sitting there, but it might be. Company's called ProEnergy and Missouri based, and they started with retrofitting decommissioned 747 jet engines, 48 megawatts, standard design, stackable in containers, deliverability in 12 to 18 months. So we're kind of cutting the queue with what we have. And it's just our own personal perspective. The demand is unbelievable and the pricing is unbelievable. So this shortage is going to continue. I also think the turbine manufacturers are looking very closely at the backlog and who's getting it. I think they're getting smarter. There are developers and then there are developers. And I think early on, a lot of developers maybe with not all that well-capitalized or not all that well-experienced, or maybe they didn't have the land, or maybe they didn't have the off-day contract, and maybe that's not going to get built. So they're now the Siemens and G and Mitsubishi in the luxury position of even high grading. If you miss a certain deliverable or timeline, you're out. And we're going to go to the higher likelihood gas developer. But they hold many of the cards. And I think that is going to continue for some time. And I think there was a little bit of naivete on part of some of the digital infrastructure data center developers who thought, well, I just call up the utility when I need 300 megawatts and they'll hook me right up. And we're finding that that power isn't available from the utility. An interconnection might be three years at best, more like five years. And so this powered land, brownfield expansion, get me a gas turbine as quickly as you can. And all of us in the power world are building our relationships greatly with the turbine manufacturers, in our case, buying our own and building out our own, which I think gives us a competitive advantage of having that availability. But it is the choke point. And the pricing, remember I told you the spread between renewables, solar and storage, and gas fire generation, that spread is widening. Even though maybe solar is getting, with tariffs, it gets a little more expensive, but the gas turbine costs have pushed up the price that gas fire develops. So we're talking, I don't know, maybe we were $1,000 per KW to build, bring a gas plant online five to seven years ago. I don't know if we're going to hit $2,500 per KW, but maybe. But the hyperscalers are, that's still a bit of a rounding error. And they're willing to make that happen.
Speaker 1I wanted to know what you think needs to be done in terms of regulatory policy to bring new generation to the grid and address both carbon emissions, but also the rising cost of electricity, which is becoming a pain point with consumers, or at least they associate a lot of it with AI now.
Speaker 2Yeah, well, maybe just let's talk about the rising price of electricity, because I think there are some in the regulated utility world, maybe find it convenient to shift that blame over to the hyperscalers. Most power generators, we're not making money yet from AI and hyperscalers, but everyone's saying, wow, the price of electricity has gone up. Well, what has caused the price of electricity to have gone up? If we look at a retail bill over the last 10 to 15 years, there has been massive transmission and distribution spending. I pick on the state that I'm from, New Jersey, which was a big contentious part of the governor's race of whose fault is it of high electricity prices. They had something called Superstorm Sandy, maybe that was 12 years ago, and huge damage in the state. It was very easy for the largest utility in the state to go to the regulator and say, we need to harden the grid. We need to underground lines. We need to take the wood poles away and spend tens of billions and still tens of billions more. If we go back industry-wide, we're into the hundreds of billions of spend on transmission and distribution, some of it hooking up remote renewables, some of it just hardening the grid, but they got those spending through. If we look at T&D, which is now way more than half of the bill, if I look at that bill in New Jersey, maybe you got 40% of it, T&D, 20% of it is things like taxes and policy in there, and a smaller piece is the power generation. We're up maybe threefold in T&D, where power generation actually has trended down. We got to start with the facts, because it's just very easy to blame. It was the power generators and the hyperscalers that caused your electricity rates to go up. No, the utilities, regulated utilities, read any earnings report. They're in the business of rate-based growth. Rate-based growth is spend capital. They did a great job convincing regulators we need to spend capital on T&D. That's what drove up prices, not the power generation side of the equation. Now, going forward, it's rate design. The large customers, the hyperscalers, are going to have to pay their fair share. They can't do it on the back of consumers. They realize it. The regulators realize it. We're seeing many utilities adjust the rate design so that there is that fair and equitable trade-off, because this will come to a screeching halt in an inflationary time that we're in if power prices just run out of control.
Speaker 1Sure. I just want to bring us back full circle, because it feels to me that ECP is now in this privileged position to literally serve the zeitgeist, given what's happening in the power markets. You've been here for a long time doing this, right? But this is a special moment, as we highlighted at the beginning of the podcast. I'm wondering if and how this is potentially changing your strategy and how you do business, and if you could talk us through that bit, and if you could talk us through how you do business, and if you could talk us through how you do business.
Speaker 2Well, believe it or not, this is my 43rd year in electricity. My joke is maybe I started when I was two. I'm hopefully not as old as that would portray. And I've seen incredible cycles in 43 years. Regulation has changed. I've seen high gas price environments, $14 natural gas. I've seen $1.50 natural gas. I've seen high interest rates. I've seen low interest rates. I've seen environmental policies come and go. We've had periods of a boom in co-gen, which, by the way, when we get behind the meter, not all that different from co-gen projects for a big refinery or a LNG facility. So we've had that period of growth. We've had flat demand. Now we have rising demand. We've had renewables. Are they be all and end all? And then we had renewables are expensive and, let's say, subsidies we don't need from some of the current administration. So we've seen it all for a very long time period, but we've been consistent investors. So you have to be patient. Every asset is different. Every market is different. It's one of the beauties of investing in power generation in the U.S. The PJM bears no resemblance to California, to Texas, or to Florida, or to anywhere else. So capital is mobile. It can move around. And I don't know if there's 5,000 power plants in the country, and I don't know if there's 3,000 utilities. Most countries of the world have one or two utilities. We have the most power plants in the world. We have the most fragmented, inefficient utility system. We don't want to go back to the 1935 Public Utility Holding Company Act and how we got to where we are. But it just creates a lot of opportunities and now with growth. So electricity is our bread and butter. We're seeing deep value in renewables. We think renewables are way oversold, especially those in the public markets. Renewable demand is not something that comes from the federal administration. 30 states have renewable portfolios. We don't have renewable portfolios. Not any one of them have changed. We've got huge corporate buyers. I said Amazon's the biggest purchaser of renewables. They want the offsets. We have retail choice where a lot of people can choose the green option. So as we look at our investment strategy, you look for where there's value. And so we see value there. And then when we look where there's customers, and on the hyperscalers, we have these great balance sheets, and we think we're going to be able to source some of those contracts. So we look to that as well. And then we have maybe 30% of what we're going to What we do, we put under the umbrella of sustainability or environmental infrastructure, carbon capture, and renewable fuels, recycling businesses, battery recycling business. There's a lot of tangential businesses around this as well. And so you're patient. Portfolio construction matters. You don't want 15 investments that all look alike. So we find our way, and having done this for a very long time, maybe you get better over time with how to put together a portfolio, and the macro dynamic that you're facing. And this is a very unique macro that we're facing these days.
Speaker 1Thanks, Doug. I think that's a great place to leave this on. Thank you very much for coming on the podcast. I appreciate your taking the time.
Speaker 2I enjoyed the discussion. I hope we can do it again.
Speaker 1That, again, was Doug Kimmelman, founder and executive chairman of Energy Capital Partners. And I'm Bruno Alves, editor-in-chief of Infrastructure Investor. To hear more of our episodes, head over to infrastructureinvestor.com forward slash podcast, or you can search and subscribe to the Infrastructure Investor podcast wherever you like to listen.
Speaker 3Registration is now open for the Infrastructure Investor Global Summit 2027, the definitive gathering for infrastructure capital, taking place February 22nd to 25th at Station Berlin. Join more than 1,200 limited partners, 400 speakers, and 3,500 industry leaders from more than 50 countries, all converging in one location for four days of access, insight, and dealmaking. Where capital, opportunity, and strategy converge, secure your place today. Early bird registration is live now at peievents.com. That's peievents.com.
PodcastTranscript AI