It’s a new year, which means the veteran energy analyst Nat Bullard has dropped another annual, data-rich presentation on the state of energy and decarbonization.
And per what has become tradition, Nat is back on Catalyst — for the fourth time — to discuss some of Shayle’s favorite slides, cherry-picked from the 200-page deck.
In part one of their two-part conversation, they cover topics like:
- The significance of China’s rapid electrification
- Why the proportion of GDP spent on electricity has remained flat while oil has proven volatile
- The massive backlog and rising capital costs for gas turbines
- How current tech capex compares to past large-scale endeavors like the Manhattan Project and broadband build-out
- The extraordinary explosion of large load interconnection requests in Texas
- The divergence in load forecasting between grid operators and transmission providers
- Global drivers of electricity demand growth beyond data centers
Resources
- Nat Bullard’s 2026 Presentation
- Catalyst: 2025 trends: aerosols, oil demand, and carbon removal
- Catalyst: 2024 trends: batteries, transferable tax credits, and the cost of capital
Credits: Hosted by Shayle Kann. Produced and edited by Max Savage Levenson. Original music and engineering by Sean Marquand. Stephen Lacey is our executive editor.
Catalyst is brought to you by Uplight. Uplight activates energy customers and their connected devices to generate, shift, and save energy—improving grid resilience and energy affordability while accelerating decarbonization. Learn how Uplight is helping utilities unlock flexible load at scale at uplight.com.
Catalyst is brought to you by Antenna Group, the public relations and strategic marketing agency of choice for climate, energy, and infrastructure leaders. If you’re a startup, investor, or global corporation that’s looking to tell your climate story, demonstrate your impact, or accelerate your growth, Antenna Group’s team of industry insiders is ready to help. Learn more at antennagroup.com.
Transcript
Shayle Kann: I’m Shayle Kann. I lead the early-stage venture strategy at Energy Impact Partners. Welcome. We’re back. Longtime listeners will be familiar with my favorite time of the year: the beginning of the year when my friend Nat Bullard—a longtime analyst and researcher in the energy and climate space, but also now the co-founder of Halcyon—puts together his annual opus of hundreds of slides on the state of energy and decarbonization. It’s chockfull of fun data that you don’t get to see elsewhere. As always, I picked my favorites and Nat and I talked through them. We talked data centers, obviously, but not just data centers; oil markets, solar batteries, all sorts of things. As usual, this was too long a conversation to fit into one podcast. So this is part one. We’re going to cover a whole bunch of interesting things now and then come back next week when we will cover part two. With no further ado, here is Nat. Nat, welcome back.
Nat Bullard: Shayle, great to be back. As always, happy 2026.
Shayle Kann: Is this year four that we’ve been doing this? Something like that.
Nat Bullard: It is indeed year four. We’ve been doing this recording as long as I’ve been doing a big presentation.
Shayle Kann: I do love a big deck. 200 slides this year. So, wait, I have an important question for you. There’s no way it just happened to be 200 slides. You made it an even number. What was the one that got cut out?
Nat Bullard: The one that got cut out is nothing I can tell you. I had about 365 slides as of the start of November. The better way of thinking about it is that I essentially cut one slide for every slide that’s in there. In fact, I usually cut more than that. I start off with 300 plus and cut it down to about 150 or 160 and then build back up. So it’s first an exercise in addition, then it’s an exercise in subtraction to get to the magic number. As you remember way back when I had 141, which is a sort of arbitrary prime number count—or maybe it’s not even a prime number of slides. I always felt like there was stuff that I had left, so I made it bigger and kept it there. I think it’s quite possible that if you do more than that, you’re going to lose the edit capability that gives it some of the strength over the course of 12 months.
Shayle Kann: I think you need to start making the number some kind of a meaningful number. Our mutual friend Andrew Bebe, at his firm Obvious Ventures, their first couple of funds sizes were always very fun. The first one was $123,456,789, and then the second one was like $313,131,313. It was palindrome. Anyway, you have to come up with something better than 200.
Nat Bullard: I was thinking about natural log pie. It’s a rich tapestry of options of numbers multiplied by something that I should be able to get in there. But I do like 200 for now. It’s easier for people to wrap their heads around and to benchmark a little bit when they’re paging through it too.
Shayle Kann: Alright, enough naval gazing about numbers of slides. Let’s get into some slides. As usual, I’ve picked a subset of my favorite slides that I found interesting, and we’re just going to run through them. We’re going to start on slide 15. Given that this trend has been ongoing for quite some time, and in fact the lines crossed a decade ago, I’m surprised I didn’t already know that China is significantly more electrified—at least as measured by the share of final energy that comes from electricity—than the United States. Substantially more so, and I had somehow missed that trend.
Nat Bullard: This is a great one. It’s some work from Ember that it’s been doing for quite some time. What it tracks is, as you say, the share of final energy that comes from electricity. Another way to think about it is how electrified is an economy. There are lots of different ways that you can get to a high number. One of them would be that you have very little primary industry. Another would be that you have plenty of applied primary industry, but you apply a great deal of electrification to processes that otherwise would be driven with some kind of thermal energy.
Shayle Kann: Or you have a tiny bit of primary industry and it’s all aluminum smelting or something.
Nat Bullard: Exactly. An example of a country like that would be Norway. Which is both an advanced economy, has some industry, and is highly electricity rather than primary energy. But China is none of those things. It’s a huge industrial economy. It’s a huge user of primary energy, but it’s also a consistent user of electricity for its final energy source. It’s also moving at a much more rapid pace than either North America or Europe, which has slowly picked up over the course of five decades from 10% to a little bit north of 20%. China meanwhile has gone since 1970 from 3% to 30%.
Shayle Kann: That’s where I find the chart looks really interesting. In 1990, North America is already at roughly 20% electricity, which is still the truth today—22 or 23%. That’s the number that I always use. I always tell people if they over-index on electricity or say solar is going to be the whatever, it’s worth remembering in the US that electricity is 20% of final energy consumption. That’s been true since 1990. Whereas China in 1990 is down at 7% electricity and then jumped to 30% today. So it’s a very different trajectory.
Nat Bullard: It is a totally different trajectory at a totally different scale too. Everything in China is bigger when it comes to energy and in particular when it comes to the primary inputs. I think it’s a really important measure as you talk about the electrified future or electro-tech or the electric tech stack, that China is grasping this as opportunity that’s also being done at scale. It’s one thing to say Norway can do this. It’s another thing to say that China’s doing this.
Shayle Kann: Not that you necessarily can speak for the Chinese central government, but you’re certainly closer to it than I am. I’ve heard that one of the reasons China is focusing so much on electrification is that they wish to control their own destiny. They don’t have massive domestic reserves of hydrocarbons, but they can produce their own electricity. That is why they’re investing in solar and the battery supply chain and nuclear for that matter. Do you think that explains this? Like China is just saying, we can’t rely on energy imports long term, so we’re going to electrify?
Nat Bullard: There’s a bit of a nuance to that, which is that the primary imports that you would have of primary energy in China are going to be oil, of which it is still a major importer, and natural gas, for which it is still a major importer. It does import coal for energy balance reasons, but it has an absolutely enormous indigenous coal supply that will last for centuries. One thing to remember about this primary energy from electricity is that that doesn’t mean that it’s entirely coming from hydropower or solar wind. It can be coming from thermally generated sources. But it is definitely within the realm of one’s own destiny. Wherein the electricity is generated within boundaries, within a nation state, it is effectively sovereign. In that sense, yes, it does provide a lot more control over destiny. Less exposure to market forces, to geopolitics, to everything else. If you are firmly in control of that element of energy and wherein electricity is almost entirely within the national purview, then you would want to spend more and more energy getting that electricity share of energy up as high as you can.
Shayle Kann: Okay, so let’s move on to slide 17, which I think is an interesting coda to slide 15. Slide 15 is about how electrified an economy is. Slide 17 is super interesting and I had never seen this data put together. It’s about what share of GDP is spent on electricity versus spent on oil specifically. And the shapes of those two curves are very different from each other in a way that I guess I might have predicted, but is stark when you look at it. Describe the difference between the two.
Nat Bullard: Absolutely. We’re five and a half decades into an era of thinking about energy shocks. When we talk about those, we make it seem system-wide, but it’s really about a shock in liquid hydrocarbon prices, specifically oil. If you look back at the data, you can see just how indexed the global economy was to oil in terms of how many units of oil input it took to get a unit of GDP, and the spend within different economies on oil specifically. In 1980, the year after the second oil shock coming from the Iranian revolution, just under 9% of per capita GDP expenditures globally were going to oil. Imagine $1 out of every 11 being spent on oil of per capita GDP expenditure. That’s pretty extraordinary. At the same time, the share for electricity was a little over 3%. If you carry this trend across the entirety of the last 45 years, what you see is that the oil share goes down significantly by the late 1990s. It’s less than 5%, but it also bounces around quite a bit. Right now the share is in the range of about 5%, but it’s been as high as 6.5 or 7%. In 2020, it was below 4%. Electricity, meanwhile, is essentially completely range bound. The highest it’s ever gotten is close to 4%, and the lowest it’s ever gotten is 3%. For all the talk that we have about spiking prices for electricity and the share of GDP that might come from electricity expenditures, it’s fascinating how range bound it is. We basically spend between 3% and 4% of GDP on electricity, and that is that essentially.
Shayle Kann: That’s the question. The reason this is interesting is because of the future, not necessarily the past. Just to reiterate, the past electricity looks like a flat line for 50 years. Rising spend on electricity overall matches GDP growth essentially; it has to, because it’s a flat line. Whereas oil is super spiky. It’s gone down since the eighties, but it moves around a lot because oil prices move around a lot. So that’s how it’s gone historically. What happens now is a super interesting question. We’re in this moment where oil prices are pretty low. President Trump is trying to do everything he can to get oil prices even lower. He’s got this stated goal of $50 a barrel if we start exporting a ton of Venezuelan oil, et cetera. So he’s trying to get oil prices low. Meanwhile, electricity prices are under upward pressure. I don’t think anybody would debate that. Do we see electricity escape its collar and spike? Could we see the lines cross? Which, by the way, in this chart, they never have; we’ve never spent more of GDP on electricity than oil historically. It’s interesting to see whether this dynamic of one super volatile thing, which is oil, and one super stable thing, which is electricity, whether that’s going to hold.
Nat Bullard: Let’s do this as a thought experiment to see what would it take to make those lines cross. First of all, it would take much lower cost for oil. Much lower oil price. Two, a much lower reliance upon oil as an input to GDP or to economic growth. We already get more units of economic activity out of a barrel of oil effectively every year. But you’d need to rapidly increase or enhance that. Secondly, you’d need to both spend a lot more on electricity and get less from it. You would need to have it be less of a contribution to GDP growth. If you had both of those things happen, you’d be spending a lot more, but you’d be getting less GDP out of it and therefore, GDP is not going up as much. The expenditure is going up. That’s how you would do it. You’d have to have like $10 a barrel and people using three times as much electricity.
Shayle Kann: This is what’s going to be interesting. Let’s just take electricity on its own. Forget the comparison for a second. I think most people would bet that we’re going to spend more on electricity overall over the next few years. The question is, will GDP keep up? They’re tied to each other because the primary reason we’re going to spend so much more on electricity is AI. There are a bunch of people betting AI is going to help GDP go to the moon. Other people saying it’s going to hurt GDP. That question underlies whether we break this 50-year trend of basically spending the same portion of our GDP on electricity.
Nat Bullard: Right. Remember that it is also global. There are global, not just US or Western European questions within there. What happens when places that have a limited but non-zero reliance on oil rapidly electrify and electricity becomes more of GDP? You’re in turn electrifying more of everything and more people have access to electricity. We’ve somewhat plateaued on global access to electricity. There’s a million different ways that we can think about cutting this up to make it look possible, but it’s the right kind of question to ask without a clear answer. My not very satisfying response.
Shayle Kann: Good. Not a clear answer. Let’s move on. Slide 28. I want to talk about gas turbines. We’ve talked about this a bunch on this pod, and many of our listeners are going to be well familiar with this. I hadn’t actually seen the data laid out though. It’s measuring the order book for gas turbines that we have already seen relative to current production capacity. Basically how under supplied are we on gas turbines? What do you see there?
Nat Bullard: It’s important to start this at the front of the series, which is 2001. There were more than 80, in fact, closer to 90 gigawatts of gas plant orders in 2001. Which is an awful lot. We have to remember the “dash for gas” that you and I started hearing about from industry veterans is now quite some time ago, two and a half decades ago. But there was a time when the world was ordering quite a lot of gas turbines, and manufacturing obviously was of the mood to meet that demand with new supply. Only to find order books that collapsed from 80-something to well under 40 the next year, and then staying steadily below production capability for pretty much the entire time, with the exception of a few years, all the way up until right now. The current production limit—and there are not that many companies that make gas turbines—is somewhere in the range of about 60 gigawatts a year. We’re likely last year to be past that by about 20 gigawatts and to be past that by about 30 gigawatts this year. Who knows, based on current orders, 40 gigawatts above a 60-gigawatt production limit. There’s a lot of reasons for this, but first and foremost is if you are in the process and have the priority to manufacture gas turbines, what you really don’t want to do is be oversupplied. It’s not really a great tenable market position. Being undersupplied has probably a net positive on your ability to book contracts and to secure durable orders from customers you want. It has pricing benefits. People are going to howl at you to do as many as you can, build as much as you can. But if you are in charge of building gas turbines now, you probably have the institutional memory of the early two thousands.
Shayle Kann: We’ve talked about this before with regard to electric transformers. It’s a similar situation where folks who’ve been in the industry a long time do remember a historic period wherein there was this huge order book boom, and then the market fell out from under them and they ended up oversupplied. So there’s been reticence to expand capacity too much for that reason. They are expanding capacity, but maybe not fast enough. What’s interesting is that even with that history, we are the most undersupplied we have ever been, or at least since the data starts at the beginning of the century. Even today, for 2028, the order book is over a hundred gigawatts relative to about 60 gigawatts of production capacity, which helps to explain why—you know the expression, “everything is computer”? I like “everything is turbine.” Because now we’re seeing if you’re a jet engine company, you are pivoting to provide turbines for the grid. This boom supersonic and all the aero-derivatives—everybody who’s got a turbine is trying to turn it into an AI data center power supply.
Nat Bullard: Right. Not just that, companies are turning things that are a usable, if frankly somewhat imperfect solution for large scale, always-on grid connected power into power. Aero-derivatives are traditionally used for very specific applications, and they’re not being used necessarily to power things all the time, constantly for a decade straight. They can, but that’s just not typically within the design spec. The design spec would be for combined cycle turbines that are grid integrated and that are part of a big liquid, well-supplied power market in which they play the role that they’ve historically played. It is interesting times for all of these things. In terms of what this shortage for now, what this order book mismatch brings to the market. Who it brings to the market, the kind of approaches that people then take in terms of how they buy and sell power and everything. It’s different times. It’s nothing like we’ve experienced in our career, but for those who’ve got a little bit more tenure than us, it is achingly familiar.
Shayle Kann: Relatedly, of course, we are undersupplied. So what do you expect? Prices are going to go up. You’ve got some good data on that from you folks at Halcyon. We talked about this the last time you were on the pod a little bit, but I want to run through it again because it is interesting. You’ve got good data on the average capital cost of various types of natural gas turbine power plants as they are planned. These are ones that are not operating yet, but the expected capital cost, and how that trends into the future. Walk me through that and particularly the breakdown of the different types.
Nat Bullard: One of the exercises that we do that manifests itself as a data series people can buy is going through the regulatory corpora in the US and pulling all of the data from the CPCN (Certificate of Public Convenience and Necessity) or an equivalent. That is basically the utility going to the state and saying, “We need to build this X,” and in this case we’re looking at gas plants and all of the data that gives you an idea of what these things are supposed to cost is within there. It’s not broken out in a neat and tidy fashion; it tends to be hidden away in proceedings and responses and rebuttals. The upshot is that we can map out more than 160 active plants, close to about 80 gigawatts worth of actual capacity, in which the cost of a combined cycle has doubled. 2026 deliveries, things that are going to come online this year, are in the range of let’s say $1,200 per kilowatt. The projects that are looking to come online in 2030, 2031 are a little bit shy of $2,500, so close to doubling in that time period. The reason I find this data useful is this isn’t based on the announced capital cost for right now, it’s based on the price as it moves into the future. It’s an updated live number that reflects the actual market conditions underneath these things once they’ve been announced. Not just whatever deposit you put down with your turbine supplier, but the actual, ongoing cost to make this thing into a reality.
Shayle Kann: By the way, I also wonder, does this include EPC (Engineering, Procurement, and Construction) cost for example?
Nat Bullard: This is the delivered cost. The allowance for work during construction, all the sorts of things that flow into it, minus the things that we know are discrete and separate. Like if you needed to build 80 miles of feeder for it, we stripped that out.
Shayle Kann: We should come back and look at this data set again in like 2030 because I do wonder whether they’re actually underestimating the total costs. Because EPC and stuff like that in particular is super inflationary and it’s really hard to find EPCs right now because they’re all booked out. I wonder whether it’s going to end up being even more expensive than they think.
Nat Bullard: That’s why we revisit this every month because it moves. It’s interesting to add new assets. It’s almost maybe even more interesting to watch movement within existing assets based exactly on that. Being like the EPC target went up because we did.
Shayle Kann: It’s the same, any number of reasons, but part of it is the same reason why gas turbines themselves are; it’s an undersupply problem.
Nat Bullard: Exactly. You can see that coming through and yes, we should be revisiting this essentially constantly. We’re already seeing people getting verbal quotes that are higher. I’ve had actual developers come up to me and say like, “Your numbers are low.” I was like, “Well, show me yours and I’ll show you mine.” And then they don’t. So we haven’t actually gotten anything more concrete than this, but this is what’s written. This is what is essentially disclosed by law, and it’s a pretty fertile ground. Briefly on the other types of turbines. Simple cycle, not quite up so much, but up by about 50% from like a thousand to about $1,500 a kilowatt. Then we start to see reciprocating internal combustion engines as well, or RICE turbines. Those are really expensive. Those are like $2,500 to $3,000 a kilowatt already. Interestingly, we don’t see a long delivery pipeline for those. The delivery pipeline for those only runs out a couple of years. We don’t see anybody planning RICE installations in the 2030s.
Shayle Kann: Is that because I think they’re mostly used for either bridge power or backup? Replacement for the diesel gensets.
Nat Bullard: In theory. But they’re increasingly being deployed at a scale that suggests that they’re being used for something closer to bulk power.
Shayle Kann: Right. Maybe that is, like I said, everything is turbine.
Nat Bullard: Everything is turbine.
Shayle Kann: Okay. Let’s stay on the theme of all this power build-out stuff. I really like this next one, slide 32. “So much CapEx,” that’s my version of the slide title. You’re comparing the total amount of CapEx spending on this—which is going to be predominantly data centers so it just says Tech CapEx in 2025—to other historic booms in CapEx spending in the economy, which is a good way to compare. What do you find?
Nat Bullard: I’m going to give credit first of all to Michael Cembalest and his team at JP Morgan Asset and Wealth Management. They built this slide first, not me. In the past, I’d done some examples of interstate highway and broadband CapEx as a comp, but I’d not done this full suite that they’ve got here, which goes from all the public works in the 1930s like the Hoover Dam through the Manhattan Project. And we could call our wave of electrification in the US in the late forties the Apollo project, the highways, broadband build out, and then the tech CapEx. Things like the Manhattan Project, electricity, the Apollo project—these are less than, or barely above, one-half a percent of US GDP at their peak. Even the Apollo project, even the interstate highway project is like six-tenths of a percent. Building out broadband CapEx in the year 2000 at its peak was like 1.2% of US GDP and Tech CapEx right now is just under 2%. Basically higher than anything else. To your point, this is the capital expenditure to build compute essentially. This is CapEx for building just the actual computational elements as well as the buildings that contain them and the power stuff that’s within the fence of the company’s capital expenditures. It is not power and transmission and water CapEx.
Shayle Kann: Right.
Nat Bullard: So it’s a pretty fascinatingly big number relative to everything.
Shayle Kann: I will add all these other data points, as you’ve said, you can go into history and you can figure out what year was the peak. The peak of broadband CapEx was the year 2000 when it was just over 1% of GDP. That’s comparison against 2025 actual CapEx of tech, which may or may not—probably isn’t—the peak.
Nat Bullard: Right. In fact, the 2000 example for broadband is instructive because the NASDAQ bubble burst in March of 2000 and CapEx kept going. Michael Burry made this observation recently on Michael Lewis’ podcast that the capital expenditure actually lags what might be happening in the purely financial market. So this could keep going for a while. The CapEx is committed. Sometimes it’s already underway and a lot of it will keep going. This is probably not the peak. Most of the estimates, based on what companies themselves are saying for their estimated CapEx, have a higher number for next year. And then again, you attach the relevant quantum of investment in the electricity sector to it and it’s a lot more money too. It’s hard to say in many cases specifically, is this new CapEx specific for energizing this data center? But certainly the prime mover of demand growth and of building new infrastructure in the US is for energizing data centers. The utility CapEx that goes with this is also in the tens if not hundreds of billions of dollars.
Shayle Kann: Good segue. Let’s get back into energy then. Let’s go to Texas. We’re going to ERCOT. Slide 91 is on the queue, the Texas interconnection queue. This is the large load interconnection queue, not the generation queue, although the generation queue looks kind of similar. Everybody knows this, right? Lots of people are trying to build data centers in Texas. No big surprise there. The queue has gone up a lot. It is pretty astounding how quickly it has gone up. The data suggests that the pipeline of large load interconnection requests in ERCOT in Texas was 40-ish, 42 gigawatts as of January of 2024. So two years ago it went from 40-ish gigawatts to 226 gigawatts as of November of 2025. So I presume now it’s even a little bit higher. Those are stupid high numbers. Just as a reminder, I just want to frame this up a little bit. As of maybe two years ago, there were about 30 gigawatts of data centers in the US in total.
Nat Bullard: Yep.
Shayle Kann: So this is going from 41 to 226 in Texas alone in two years in the queue. Now, that’s not all going to happen, obviously, but nonetheless.
Nat Bullard: Right. This is a great one. ERCOT kindly publishes this every month in a somewhat unstructured format, but high enough frequency that it’s worth extracting and putting into this fashion. This is an awful lot. 226 gigawatts. This current state peak load is in the range of about 85 gigawatts, so that’s like two and a half x-ing the existing state peak load if all of this were to happen at once. It’s gone really rapidly. It’s increasingly co-located, like a couple of tens of gigawatts of that are actually co-located in large interconnection load, which is interesting. That keeps ticking up. But sounding like financial disclosures here, not all of these assets will eventuate. I don’t think that Texas is actually going to be building 226 gigawatts of just large load in the coming seven to eight years.
Shayle Kann: It is the nature of interconnection queues. A lot of it is speculative and it’s especially the nature of bubbly interconnection queues. Like clearly most of this won’t get built out. I think it is indicative though of one thing that is definitely happening, which is just people have the perception: Texas, you can build stuff. Especially big stuff. A lot of data centers want to be big. So there is a mad rush of developers, hyperscalers, REITs, Rick Perry, basically everybody trying to lock up sites in Texas where they think they can go interconnect gigawatts. And that adds up to hundreds of gigawatts in total.
Nat Bullard: There’s something else here that I think you and I, and many of your listeners will be very familiar with, which is a highly speculative supply side queue. We’re very comfortable with the fact that, of course, wind and solar developers plan for 10 and they’re going to build two. Or that ratio might even be too high. You’ve got 10 assets that you’re planning and you’re going to build one of them. That you’re highly speculative in terms of where you’re going to go, what you’re planning to do. The size of any asset itself is also fairly prospective. But it depends on what you’re going to be able to get. You’d be silly not to max out the possible interconnect on the site. And you’d be silly to not try to do as much as you can for optionality sake. What I think we’re not used to is a demand side interconnection queue that has some of those same speculative elements. Like back in the day, if you’re building a hospital in suburban Atlanta, you’re not going out and picking seven or 10 possible sites for that hospital, and you’re definitely not picking seven to 10 sites scattered across four or five different states. If you are building a hospital, it’s because you have a human need for medical services in a particular place. You’re not viewing it as completely fungible between, maybe we’ll go to Tennessee and build this same hospital, or maybe we’ll go to Texas.
Shayle Kann: I think it’s a different thing though actually. I don’t think what’s happening is the developers are saying, “I need a data center and I’m going to pick seven to 10 sites in whichever one wins wins and I’ll build it.” It’s actually that a lot of developers, speculators, et cetera, are saying, “If I can develop this site, I can monetize it.” “I could sell this thing or maybe I can lease it if I’m a colo.” So I’ll do as many of those as I can do that I think are good because right now there’s a really valuable market on the other side. And so everybody’s doing that in Texas.
Nat Bullard: Yeah, you’re right. With my one other wrinkle being that if you’re the pure speculative element of this—let’s say this in the good way—you’re in the land business side of this. You’re in the site control part of this business. That’s true. If you are then building the data center on one of those sites, though, if you’re building the compute, you could be more fungible between exactly where you’re going to go depending on other factors. To some extent. There’s more spread across different places based on what you’re planning to do. “I need to build this compute in this period of time and I’ll talk to whoever has site control that will help me do that.”
Shayle Kann: All right, so then the direct result of this is the next slide, slide 92, which is the “no one knows anything” slide. We’re staying in Texas. We just talked about this crazy big load interconnection queue. So of course the question is then how much new electricity demand is there going to be in Texas as a result of that? That is the operative question, whether you are a grid operator or the market itself. And you have these great contrasting data sets of the load forecast from two parties who you would think would be pretty aligned because they’re both trying to answer exactly the same question and they work hand in glove with each other. Walk me through this data.
Nat Bullard: Sure. This is one of the “no one knows anything” slides. My perennial favorite before this was “markets respond to incentives.” This is the new one for our current age: “no one knows anything.” The transmission service providers who are responsible for building the grid and integrating the energy required to energize and electrify what happens in Texas are fundamentally serving the same market that ERCOT, the grid operator, is operating. However, ERCOT says, you know what, we could go from under 500 terawatt hours in 2024 to like a thousand by 2030. Let’s call it up 110% in that time period. The transmission service providers, on the other hand, are like, “Sorry. We expect to go all the way up to about 1600 terawatt hours.” We’re going to go 240% up from where electricity demand was in the state in 2024. Part of the reason is that they’re looking at different information. The TSPs are looking at everything that people are asking them to build. ERCOT is looking at everything that it thinks will actually happen, but also the incentives are there. ERCOT’s incentive is to keep things operational as to the highest degree possible, and at cost distributed across all of the people who receive service in Texas. The transmission service providers are paid for building assets and will happily, if possible, build whatever asset base they’re being asked to build. So the true number is either somewhere in between or much closer to ERCOT’s figure. ERCOT has the reason for demand-supply balancing and upkeep and everything else to get it really accurate in energy terms. But the transmission service providers have every incentive to go as big as possible because that’s how they get paid. They get paid to build assets.
Shayle Kann: It’s less like they just have different views of the future and their views vary so substantially from each other, and more just like the transmission service providers have an incentive to maximize the number and it’s not a real forecast.
Nat Bullard: It is a forecast. It’s driven by what people are asking them to do. The question is how much discounting are they doing? And that I think is where there’s a significant difference between ERCOT and the TSPs.
Shayle Kann: This is the forecast out to 2030. It’s four years away. It’s not far in electricity supply terms. It’s like the blink of an eye. There’s a difference between these two forecasts of about 500 terawatt hours. Contextually, the US total electricity demand is in the range of 4,000 terawatt hours annually, maybe closer to 4,500. So call that more than 10% of all US electricity demand as just the delta between these two forecasts in Texas alone.
Nat Bullard: Yeah. Or put it this way, do we think that in 2030, Texas is going to consume as much electricity as a third of the United States consumes right now? It’s a potential.
Shayle Kann: Roughly 11 or 12% last year.
Nat Bullard: That’s right. Sometimes these things are more helpful when we ask them in this comparative fashion: what would need to be true for all of that to happen? It’s a huge number, but again, it doesn’t exist in a vacuum. This is back to my earlier point about how developers work: there are other grids that are similarly aggressive in their expectation of what demand might look like based on requests that they’re getting without any ability to zero that out against similar or identical potential demand that might be built somewhere else and not happen in wherever it is. There’s no way to cross reference all of this stuff yet because of the nature of the way they’re regulated state by state.
Shayle Kann: Okay, so the next one I want to jump back actually to slide 35. Because this one, I think it’s a reflection of a US-centric mindset that I have that I’m very surprised by this. I don’t believe this data, but the data is from the IEA and it’s a projection of how much of the electricity demand growth through 2030 is going to come from various different sectors. We just talked about in the case of Texas, but it’s also true the case of the US, like this insane boom in electricity demand coming from data centers. What’s surprising about this other data set is that data centers are ranked fifth in terms of the source of new electricity demand. I presume that’s because this is a global perspective, not a US perspective.
Nat Bullard: That’s right. This is a global perspective. Globally, electrification of industry is going to be like 30% of the demand growth for electricity between 2024 and 2030. Even electrified transport, which we in the US are sort of being trained away from thinking about as a big driver of demand, is a bigger driver of demand around the world than data centers is or would be. But even like appliances, there’s a lot of the world that needs to add its first dishwasher. Even its first refrigeration space cooling. Just air con and buildings is going to be like 10% of the total growth and data centers are right around 8%. What I would say is consider this very much a moving target. I’ll be very interested to see what this print looks like a year from now, two years from now, three years from now. The other thing to be considered is, is there a trade-off? If there’s a finite quantum of new electrons that are going to be consumed between now and 2030, is it going to come to the point where more is being consumed by data centers and less in absolute terms by space cooling? That would be complex; that would be something for the rest of the world that would be akin to a trade-off that we really haven’t had to do in the US in quite some time. At least not at a national level.
Shayle Kann: Yeah, it’s a moving target. Data centers are going to move up this ranking before too long, even on a global basis.
Nat Bullard: I would agree with that. Certainly they’re going to move up, but where they land is a really big question. And how they interact with the rest of these different places that electricity will be consumed is going to be really interesting to watch.
Shayle Kann: Yeah. And the fact that in some markets it’s to some extent a near zero sum game in the sense that there’s only so much supply and we’re building out as much supply as we possibly can. So every new data center is a new electrified industry facility that isn’t going to happen probably. Nat Bullard is a long time climate tech analyst and writer. He’s a co-founder of Halcyon, which is an AI assisted research and information platform. This whole conversation was based on a 200 slide deck that is chockfull of other interesting information. You should go read the whole thing. It’s at nathanielbullard.com.
This show is a production of Latitude Media. You can head over to latitudemedia.com for links to today’s topics. Latitude is supported by Prelude Ventures. This episode was produced by Max Savage Levenson, mixing and theme song by Sean Marquand. Stephen Lacey is our executive editor. I’m Shayle Kann and this is Catalyst.
