Transcript: Everything We Didn’t Know About the World’s Buzziest Geothermal Startup

This transcript has been automatically generated.

Subscribe to “Shift Key” and find this episode on Apple Podcasts, Spotify, Amazon, or wherever you get your podcasts.

You can also add the show’s RSS feed to your podcast app to follow us directly.

Robinson Meyer:

[1:06] Hello, it is Thursday, April 23. One of the most interesting companies in clean energy is going public. For the past few years, if you asked anyone in climate or decarbonization what company they were excited about, they were pretty likely to say Fervo Energy. Fervo uses oil and gas extraction techniques to generate zero carbon, 24-7 geothermal power. And in theory, this electricity should even be dispatchable, meaning it can be flexed up or down like how natural gas plants that are used on the grid today.

Robinson Meyer:

[1:34] Fervo has the support of climate advocates, famously, but also in a quite interesting way, the current Secretary of Energy, Chris Wright, and I would say many Republicans

Robinson Meyer:

[1:42] in Congress and even the Trump administration at the most broad. Last week, Fervo Energy filed documents with the Securities and Exchange Commission for an initial public offering later this year. And those documents are our first real look inside the company’s finances and how it understands its future. They tell us a lot about what the liftoff path for advanced geothermal will look like through 2030 and 2032. And we’re here to talk about them today. So here to talk about the good, the bad, the worrying, the less worrying, the optimistic, the hopeful. We have two great guests. You know, both of them. First up, we’re talking with former shift key full time co host now occasional time guest co host Jesse Jenkins, professor of energy systems engineering at Princeton. And then we’ll be joined by Matthew Zeitlin, a Heatmap staff writer who’s been covering the S-1 for us. Before we fully get into it, I do need to disclose something for the first time ever, which is my brother recently began working at Fervo, but he hasn’t told me anything non-public about the company, so don’t get too excited. I’m Robinson Meyer, the founding executive editor of Heatmap News, and you are listening to Shift Key. Jesse and Matt, we are here. Welcome to Shift Key.

Jesse Jenkins:

[2:47] Hey, thanks for having us. Thank you.

Robinson Meyer:

[2:49] So Jesse, I just want to start by, you have done a lot of work with Fervo. Can you begin this conversation just by orienting us to how you think about their, how you think about advanced geothermal, how you think about kind of Fervo’s stack, and how you think about maybe the future of the company?

Jesse Jenkins:

[3:04] Yeah, it’s been really exciting to watch them go through the various stages. I think when we started working on research with Fervo in, I think, 2020, it was a small business innovation research grant, SBIR grant of like $65,000 or something like that from the DOE geothermal office to kind of help explore the potential for flexible operation of these hypothetical future power plants they were planning to build. And since then, we worked on multiple papers trying to understand the long-term potential of enhanced geothermal in the U.S. And watched as Fervo took that drawing on the back of a napkin concept into a commercial operation of their first pilot three megawatt scale plant in Nevada, and now on the cusp of an IPO. So exciting to watch that evolution. The deal with geothermal is that we have only three and a half gigawatts of geothermal in the United States operating today. That’s conventional, we call it hydrothermal power. And the reason it’s so limited is that in order to do geothermal the traditional way, you have to find a location where you have three key things all in the same place. You have to have hot enough rock conditions. So you need enough heat that you can make usable power out of that heat.

Jesse Jenkins:

[4:09] You need to have a natural fracture network or some kind of geology that allows for water to circulate through that hot rock. And then you need to have the water. So sometimes the most traditional ones, they actually are using water that’s been down there for a long time and naturally migrated into these fissures. And then they’re extracting either as dry steam in a few cases or as a brine that they then extract heat from to flash a working fluid into power. So the challenge is that finding all those three things naturally occurring in the same place is challenging. There are limited number of locations like that. And when you do find them, you tend to find 25 megawatts or 15 megawatts or fairly small scale production with only a handful of exceptions like the geysers field in Northern California, which is like a gigawatt scale type field. And so it’s just not a very exciting investment opportunity, right? Invest in exploring for drilling lots of potential dry holes in the ground. And then when you find one, you get a pretty small resource potential.

Robinson Meyer:

[5:03] And there’s basically also no replication, right? You can’t get really good at building geothermal wells because they’re all bespoke. They’re all in different places.

Jesse Jenkins:

[5:11] Yeah. And the geology is different. The chemical composition of the brine is different. So there’s all kinds of different challenges there. So what Fervo is trying to do, along with a few other advanced geothermal companies,

Jesse Jenkins:

[5:22] is try to solve that problem. And the way they do that is by saying, look, there’s hot rock all over the place. If you drill down deep enough, it’s hot everywhere. But even if you don’t drill that deep, say three, four kilometers, you access in many places is temperatures that are suitable for geothermal power generation. The problem is that you’re drilling into impermeable hard granite or other kind of crystalline basement rock for the most part at that depth. And so what they are doing is taking a page out of the shale gas and oil book, which is to drill down into those impermeable surfaces, you know, layers, find the hot enough rock, and then start drilling laterally for several kilometers usually. And drill a parallel well next to that and then use hydraulic fracturing to create the reservoir that you need to circulate water through. And then they will pump in water from an external source and circulate that in a closed loop with very little of that water hopefully leaking out into the pores of the largely impermeable rock. So that is an engineered solution, right? They sometimes call it engineered geothermal energy systems or enhanced geothermal systems. That’s a replicable strategy that if you find a big chunk of hot rock down there, you can go do this one after another, one set of production injection wells after another, and take geothermal to a gigawatt and maybe even terawatt scale in the long run.

Robinson Meyer:

[6:37] Okay, I have to say that I’m having an insight here that I never realized before, which is I had always assumed, you know, I know that technological name for what Fervo does is enhanced geothermal. But I thought it was enhanced because we were using enhanced drilling techniques from,

Robinson Meyer:

[6:51] the shale boom, oil and gas. But actually what’s enhanced is the rock itself. We’re basically enhancing the resource. It’s enhanced in the same way the pro-steroid enhanced games are enhanced. That’s right.

Jesse Jenkins:

[7:04] The performance, the permeability. Yeah. And in fact, I think the real historic reason is that when they first started doing this, they were trying to do it to stimulate additional production at conventional geothermal wells. So they were enhancing the productivity of a conventional well. It also could be able to engineer geothermal. But yeah, it’s unfortunate acronym. I think in general, people are talking about advanced geothermal or next generation geothermal. That’s probably a better way to put it. But again, the exciting part is like you’re engineering the resource space that you need through hydraulic fracturing and reservoir creation and engineering. And so it’s a technical engineered solution to the limited availability of naturally occurring hydrothermal resources. And it turns out the U.S. is a really great place to do that for a couple of reasons. We have a lot of areas with relatively hot rock closer to the surface due to the sort of natural geothermal gradient or heat gradient, how much hotter it gets the deeper you drill. And there’s one thing America is still good at, it’s drilling wells. So we have an enormous amount of technical know-how and workforce expertise and innovation coming from our massive oil and gas sector. That’s where Tim Latimer and Jack Norbeck, the founders of the company, they have backgrounds in that sector as well. And most of their leadership and on the ground employees do as well. So they’re pulling from an enormously talented workforce. It’s not a copy and paste application of the same exact techniques as in oil and gas, but it is learning an awful lot and creating a technical foundation to enable this next generation of geothermal power.

Robinson Meyer:

[8:31] They do have, I mean, you can see in the document that they do have challenges that don’t come from oil and gas. For instance, oil and gas, at the end of, while not a commodity, because as we know from previous shift key discussions and our energy expertise, like the mix of particular crudes that you pull out of one location are not the same as you might pull out in another location. But you are pulling out kind of a commodity antecedent while if you in while in for an enhanced geothermal system, you have to generate electricity when you get it to this, when you get hot liquid to the surface. Which means you need to stick a power plant there and have an interconnection and make sure.

Jesse Jenkins:

[9:09] And actually, and run a pumping. Yeah, and run a pump that is pumping and injecting that fluid through the subsurface. So some fraction, usually on the order of like 15% of the power produced by an enhanced geothermal power plant is actually used to run that injection pump and circulate the fluid. So kind of the net output is lower. That is one of the opportunities for flexible operation we can talk about later, but that is a key feature of these plants. They’re pumping water continuously through to circulate as a working fluid to extract that heat.

Robinson Meyer:

[9:38] Zero Lab, your lab at Princeton, has done some research for Fervo about the scale of the potential resource here. Can you just tell us how big could geothermal eventually be and why?

Jesse Jenkins:

[9:48] Yeah. So the reason I’ve gotten so excited about advanced geothermal is it is a potential terawatt scale resource. There are not a lot of those, right? Solar is, wind is, nuclear power is, fossil fuels are. There are just not a lot of resource options out there that you can actually scale to. A terawatt is like the whole production of the U.S. grid. It’s a thousand gigawatts. It’s like the entire production of the U.S. grid at the moment. So this is a large scale resource. Obviously that’s like a technical potential. It’ll take time to ramp up and get there. But the other thing that’s exciting about EGS is it is likely to experience pretty steady cost reductions as you deploy more and more of it at a dynamic we call experience curves or learning by doing, which is something we’ve seen in wind power and in solar power and in batteries. The mechanisms responsible for the tremendous cost declines we’ve seen in those technologies as we’ve built more and more of them. And there are a variety of mechanisms that you can anticipate with enhanced geothermal that as they get more experience and scale up are likely to lower the cost. That includes improvements in drilling. There’s very little limited experience actually in drilling in hot crystalline rock. That’s not what the oil and gas industry likes to do. And so they don’t spend a lot of effort trying to do that. So there’s some low-hanging fruit and some innovation improvements that could happen beyond just porting oil and gas technology over and starting with that.

Jesse Jenkins:

[11:06] There’s changes in the reservoir design itself. They can drill longer laterals. They can get better at generating longer fracture networks so they can space wells further apart and get more circulation per well. Various improvements in the reservoir design. And there’s potential for the surface plant itself to come down in cost. You know, we mentioned every conventional geothermal plant is sort of a bespoke design. And so they use these turbines that are kind of hand-built specifically for that power plant. Just recently, Fervo announced and confirmed in the S1 that they’ve procured 1.7 gigawatts of what they’re calling geoblocks or 50 megawatt standardized power units from Turboden, which is a leading producer of these geothermal Rankin turbines. So they’re trying to standardize the surface plant. And the reason you can do that is you can basically engineer the reservoir to produce the right increments of heat for a standard power plant. And so, you know, you can just copy and paste and build, boom, boom, boom, boom, a bunch of these 50 megawatt units. And that is also likely to experience learning curves and cost reduction because there’s a real substantial difference today between the cost of a geothermal Rankine turbine, which is just a steam turbine, and the kinds of steam turbines you would find on a coal plant where we built, or a gas plant where we built hundreds of them. And so there’s a big cost reduction that’s possible at the surface as well as below the surface. So what we looked at in our paper was, if you could get that experience curve going,

Jesse Jenkins:

[12:27] Would it make geothermal cheap enough that it would take over a large share of the U.S. market and under what conditions? And what we found is that it could easily reach hundreds of gigawatt scale by 2050 if it started around today at good sites with reasonable economics, supported by an investment tax credit, right, which we potentially have, or early willingness to pay from folks like Google and others that are procuring this power, and then kicked off that learning curve dynamic. Expanding to a few other sites that are kind of what we call near field geothermal sites or sites near traditional geothermal wells where we know it’s hot. And using those relatively high quality initial sites to kind of bootstrap that learning curve dynamic. And then once that fly wheel is going, you can expand to many other areas around the country and reach that large scale. So that’s the long-term potential. If you can kind of get on that learning curve trajectory, keep driving down costs into the $3,000 a kilowatt range, which is what Fervo is targeting in there and to the kind in the S-1. That’s a truly scalable resource that’s quite competitive and even could work in the eastern portion of the United States.

Jesse Jenkins:

[13:33] When I first started researching geothermal, I assumed it was a Western-only solution. That’s where all the traditional geothermal is. It’s where the best sites to launch enhanced geothermal are. And so I was like, great, we’ll solve that problem. That’ll solve our needs in the West, but what about the East? We’ll still need nuclear for the East or something. But it turns out there’s actually, if you do see these kinds of cost reductions in drilling and in surface plant, there are pockets in the East, in places like Mississippi and West Virginia and New York and Pennsylvania, where you could actually conceivably produce economically competitive power with EGS even in the Eastern portion of the U.S.,

Robinson Meyer:

[14:06] We might talk about this in a bit, but I think one thing I learned from the S-1 is that Fervo has acquired almost 600,000 acres of federal land where they believe there’s good geothermal resources. And they did it basically before 2020, before people started to get excited about enhanced geothermal.

Jesse Jenkins:

[14:21] Yeah, before anybody knew this was coming. Yeah, exactly.

Robinson Meyer:

[14:24] The story they tell is one that they basically no longer are interested in the leases that the federal government is offering for geothermal resources. And they were both able to buy better acreage with better resource at lower costs than in current acreage is going for now, even though they think their

Robinson Meyer:

[14:41] portfolio is has a better resource. Jesse, one of the things that people are most excited about with Fervo and one of the things, frankly, that you got me excited about with regard to Fervo and other enhanced geothermal companies is that this is dispatchable power. It’s not only that it’s 24 seven, but much like Like we currently flex gas plants up or down to meet demand on the grid. We might be able to flex geothermal plants up and down. Can you just describe like how that would work and why it’s important to kind of overall value of this energy technology?

Jesse Jenkins:

[15:12] Yeah, so most people think of geothermal as a kind of zero marginal cost resource. It has no fuel cost, right? It’s producing power that’s on the margin, basically free. And so it would make sense to operate it like a quote unquote baseload resource running 24 seven, because why would you ever turn off? The reality is that if you are deploying geothermal in a world with lots of cheap solar, for example, or wind and other parts of the West, there are many hours when power is literally worthless or very inexpensive, right? You’ve got wind and solar flooding. The market was also zero marginal cost. And so producing power in those hours, you can do it. But why would you? It’s not valuable. When it’s valuable is the times when the sun is setting and the wind is dying down and you would otherwise have to fire up gas power plants. So one of the cool things about enhanced geothermal is that you’re basically engineering a fracture network inside a very impermeable rock, right? You basically have a container around it of granite. And that means that very little fluid or pressure will leak out of the reservoir if you inject more fluid into it. And so you basically built yourself like a pumped hydrate reservoir underground for free, because that’s what you needed to create your heat exchanger to get the heat out for your power plant.

Jesse Jenkins:

[16:22] So Tim Latimer and Jack Norbeck, co-founders of Fervor, they came to us early on back in 2020 with this vision, having found a paper about a demonstration project that was done by DOE and others in the geysers in California in the early 1990s, where they practiced basically modulating the injection of fluid into the well, into the reservoir. So picture this, when power prices are really cheap, you turn off your production well, you throttle it back so that fluid is not coming out of the well or is coming out of the well at a much slower rate you now crank up your injection pumps because they’re consuming power but power is free so you’re buying it from the grid and you’re running your injection pumps harder than you normally would for steady state operation and you’re pumping fluid below the surface that fluid has nowhere to go because your production well is not letting it out at the same speed you’re pumping in and so that builds pressure and fluid in the

Jesse Jenkins:

[17:16] And you’re basically charging a battery. And then when power prices get high in the afternoon, you do two things. You stop pumping with your injection well. And that immediately boosts your power output by like 15% because you no longer have that parasitic draw of trying to operate your steady state injection. And you open up the throttle on your production well and you get a surge of geofluid coming out of brine because it’s pressurized. It’s under pressure now and it wants to come out. And so you get this sort of surge flow that will come out naturally without any injection right at the peak time.

Jesse Jenkins:

[17:49] So the only thing you have to do to take advantage of that is build a slightly bigger injection pump, which is pretty cheap.

Jesse Jenkins:

[17:55] And the more expensive part is size your surface plant to be able to accommodate that flow, that extra peak flow. So if you’re a 50-megawatt baseload operation, you might need to be able to accommodate 75 or 80 megawatts of peak flow. So that means you have to build a bigger surface plant to take advantage of that. That does add some cost, but it’s basically all in the power cost. The energy reservoir itself is free and it’s multi-day. As basically a long duration storage alternative to like Form Energy or others in that space. So that’s the kind of technical concept. It’s one that, again, has been piloted in a trial. To my understanding, Fervo has done a limited amount of testing with ARPA-E funding at their site in Utah as they’re drilling and doing initial flow tests. But they’re not planning to do this in commercial operation in the short term. But it is another source of value unlock that they could turn to. And what we found in our papers was that it was as important as drilling cost reductions to the long-term economics of geothermal energy, right? If you’re a technology, what you basically need to do is have cheaper costs than your value, right? That’s how you make money. You make money on the spread between value and cost. And so there are two ways to enhance that value. You can drive down the cost or you can deliver more value. And that’s what this sort of flexible operation allows you to do is shift production

Jesse Jenkins:

[19:12] out of hours when power is worthless and dump that energy into ours when power is valuable. And that makes EGS better than baseload. It’s a flexible, firm resource like a gas power plant.

Robinson Meyer:

[19:25] Let’s bring in Matt into this discussion. So, of course, one reason that these, it’s always a big deal when these, they’re called S-1 filings come out. It’s usually described in the press as like, this company is filed to go public because what it means is that a company that previously had private finances is now disclosing them for the first time. And we can kind of get a look inside its books in the same way that we do regularly on a quarterly basis with public companies.

Robinson Meyer:

[19:48] Matt, you’ve been writing about the Fervo S-1 filing for us here at Heatmap. What stood out to you about this filing and maybe just orient us to kind of where this company stands today and what it’s looking to do in the future.

Matthew Zeitlin:

[20:03] S-1 filings they’re the opportunity for companies to do two things i mean the beginning of it there’s a heavy like narrative component where they’re essentially in writing making their pitch to investors to kind of explain what the company is where they’re going how they plan to make money over time and then there is financial data, which is often what people are really interested in. Technology companies have started going public later and later, so the financial data is more interesting. But Fervo is definitely very much a company that is raising money for its future operations so they can earn money. Its revenue is token. It’s almost zero. But what the company is describing is that they have something like 100 and then another 400 megawatts, ideally coming online, starting at the end of this year, beginning of next year, and over the next few years. And they also need to raise a substantial amount of money, both from the IPO and then also from financing, project finance, which they also talk about a lot in this document, to get those megawatts online. And the other thing that’s really interesting about it is that they kind of describe their customer base and how they want to operate the business.

Matthew Zeitlin:

[21:11] On the revenue side. And this is very much a company that’s optimized for a world in which offtakers are buying PPAs, and they put some kind of reliability or clean premium on those PPAs. You know, like everything that’s been published since January 2025, there’s not a ton of talk about climate change and carbon emissions. But unlike some other documents we see, there’s more than zero. Like the carbon free nature of this is still a big part of the appeal. And they definitely envision a world in which they are selling PPAs or something like $100 to $130 a megawatt hour PPAs, which is kind of the going price for a clean firm bought by a big tech company. And in the case of Fervo, that big tech company is almost certainly going to be Google. Google is all over this document. I believe Google is an investor in Fervo, and Google is certainly a customer in Fervo. And they are going to be, if everything works out, their biggest customer for a long,

Robinson Meyer:

[22:04] Long time.

Matthew Zeitlin:

[22:05] They have an agreement that they would potentially sell up to 3 gigawatts of PPAs to Google although it’s a document notices this is not a contract they’re not

Robinson Meyer:

[22:15] Obligated to know it’s like a it’s Google.

Matthew Zeitlin:

[22:17] Obligated to pay for it

Robinson Meyer:

[22:18] Yeah exactly basically the way i mean this is one of i think the interesting things we get light we get light on inside the document is that the Google Fervo deal basically gives Google the option in the future to buy Fervo’s power if Google wants and it’s not and then to impose conditions on Fervo Yes, they have full audit rates on any Fervo project that they buy from.

Matthew Zeitlin:

[22:37] And not selling to their competitors. So this is just very much a creature of this world that’s developed, I guess, since the late 2000s and early 2010s, where technology companies are signing PPAs and they’re paying a premium for non-carbon and then more recently for reliability slash firmness. And so it’s kind of in the same context. The financial, at least on the revenue side, kind of look something like, I don’t know, the Three Mile Island restart had similar PPA numbers thrown around and the premium was considered similar. The reasoning for the premium was similar. You know, it’s reliability, it’s firmness, it’s non-carbon. It’s a little bit, I mean, the Three Mile Island is a good one for a price referent, but it’s a little bit more like the, you know, Google’s offtake of Kairos Power or Amazon’s investment in X-Energy. It’s a strategic investment, right? that these technologies will take off and then be a major source of competitive power for them to power their data center operations in the future. And I should say it has like in many ways, they’re playing the role that like the DOE or the government would normally play in driving technology demonstration and scale up that drives down the cost of these technologies over time. And so that willingness to take a bet is a really important role in the sort of long term evolution of these technologies.

Robinson Meyer:

[23:50] Let’s step back and put some numbers on all of this. Fervo’s revenue last year was $138,000. Their revenue in 2024 was $199,000. And their loss last year was almost $58 million. They have about $789 million of kind of construction that’s in process on their balance sheet at the moment. And I think the big, something that you called out in your coverage, Matt, that I think is maybe the most eyebrow raising aspect of this filing is that they have this pilot project or this initial deployment project in Utah called Cape Station. And they are wrapping up phase one of construction on this project.

Robinson Meyer:

[24:28] They think they need $125 million to finish phase one of Cape Station. They still have to build Cape Station Phase 2. And Cape Station Phase 2 is kind of where most of the megawatt hours are going to come from out of the project. And that’s a $940 million project of which the S1 says...

Matthew Zeitlin:

[24:46] I believe the term is majority unfunded or something.

Robinson Meyer:

[24:49] A majority of which remains unfunded, unquote. Exactly. And so part of the point now they are also going out into debt markets and it looks like they’re looking for project finance to finance this. But it does seem in some ways it’s kind of analogous to a biotech company, which goes public relatively early in its life with a kind of drug that’s in trials. And there’s a lot of excitement about the drug, but it still is going to have to invest a lot of money in the drug down the road. And what it’s doing is it’s kind of giving public markets a chance to be like, hey, do you want to bet on this drug? Because we think the drug is going to be good. But like, we’re going to equity finance basically the final trials on this drug and you’re going to have a piece of the action if you want it.

Jesse Jenkins:

[25:30] Yeah that’s a reasonable analogy although they actually have been successful in raising project level non-recourse finance which is really remarkable actually for a company of this scale and stage so they raised i think 421 million dollars in a debt facility with nine lenders for phase one at cape station and that is you know non-recourse loans mean like the recourse there is the asset in the project not the company itself so it’s not a loan to Fervo energy llc or whatever the corporate entity is it’s a loan to the you know the holding company for this project and that’s typically the kind of thing you would do for a mature technology like solar or wind batteries right you would finance those at the project level because you’re building an asset that has value and that asset can serve as collateral for the loans and you know banks know how that’s going to perform and they can underwrite it and they can appropriately price that

Jesse Jenkins:

[26:22] Raising $421 million for a technology that has so far been deployed at three megawatt scale and operated for about a year is quite remarkable. That’s the kind of role that the loan programs office at DOE and now the energy dominance office or whatever it’s called is sort of meant to play is, you know, offering this sort of debt backing for these first of a kind large scale deployments that wouldn’t otherwise be able to raise debt. But for whatever reason, LPO has largely spurned Fervo, or Fervo has chosen not to go down that route. But they were successful in raising this project-level financing, which is kind of like this bridge to bankability concept you’ll hear Jigar Shah talk about a lot. When he was running LPO, the whole goal was to help companies bridge to this level where projects are bankable, meaning financeable at the project level. Now, they may not be done, and we don’t know what the cost of capital was for that entity or that project finance. But the fact that they were able to raise it at all as a non-recourse loan is a very good sign that the economics look favorable for those projects.

Robinson Meyer:

[27:18] And I guess I should add that in some ways the company may be one LPO financing vehicle away from funding all of Cape Station Pays, too. I mean, we don’t really know. It would be a natural thing for LPO to come in on. We know Chris Wright, the current Secretary of Energy, is very supportive of Fervo and has had ties to the company in a formalized way that I’m not going to be able to remember on the fly on this podcast. But it would be a natural place for the Trump administration to intervene.

Robinson Meyer:

[27:43] You do get a sense of the kind of cast of characters around Fervo. I mean, Devon Energy, the drilling company is all they have someone on the board. They’ve invested in projects. They’ve invested in Fervo. John Arnold, the philanthropist in Houston and kind of energy czar, former Enron gas trader, kind of all over lots of interesting permitting and bipartisan energy and environmental causes is an investor in some of the projects. He actually, he gets a royalty fee, I think, on Cape Station Phase 1 on all the power that comes out of it. You just get some interesting, like you get an interesting kind of set.

Jesse Jenkins:

[28:15] Yeah, Google, obviously.

Robinson Meyer:

[28:16] Google, obviously.

Matthew Zeitlin:

[28:17] And then Bill Gates and DCVC.

Robinson Meyer:

[28:19] Yeah. Jesse, I don’t know if you had time to look at the S1, but did anything stick out to you about it?

Jesse Jenkins:

[28:26] Yeah, what I found really notable was the kind of initial project economics that they shared. They talked about the cost of Cape Station Phase 1 being about $7,000 per kilowatt. That’s, you know, high compared to a gas power plant, I would say. Like, you know, even with the increased costs of natural gas power plants these days, you might be able to get a combined cycle plant for $2,000 to $3,000 per kilowatt. That’s double or triple what it used to be. But the project doesn’t have any fuel costs. And so at $7,000 per kilowatt, you are more expensive up front, but then you’re producing, you know, zero fuel power over time. So that’s more expensive than kind of current market rates, but not that far out of the money for those kind of clean firm contract premiums that we are seeing in the market.

Jesse Jenkins:

[29:06] And it’s actually lines up very well with the early kind of baseline range in our learning curves paper. Now, that’s not too much of a surprise. We’ve had conversations with Fervo in the past and tried to benchmark our models. But it’s one thing for a company to tell us, hey, our costs are probably going to be this. And like, you have to take that with a grain of salt as a researcher that every startup is optimistic about their future costs. It’s another thing to put it in an SEC filing where you have potential like, you know, securities fraud implications if you dramatically misreport those kinds. So that was interesting to see. And it is a bit above the kind of initial costs that we were starting our learning curve at in our baseline. It’s a little closer to our higher cost trajectory. However, we were trying to model after Cape Station type costs because, you know, when you deploy the first of a kind project at pilot scale and then you scale that up by 10x right to your next project, like there are really dramatic cost reductions that tend to happen early on. And indeed, we have heard Fervo talk about they can drill 70% faster at 75% lower cost or something like that at Cape Station than they did at their initial demo at Project Red in Nevada. So when we start these learning curve estimates, we try to start from kind of a stable point where they’ve already done that initial commercial deployment and then see what the kind of sustained economies of unit scale and repeated learning by doing can do. And so we were modeling after they’d already deployed 500 megawatts of capacity, assuming that they would get down to about $5,000 per kilowatt in a baseline case. And then over time, they could get down to that $3,000 a kilowatt, a number that they have in their long-term low-cost trajectory.

Jesse Jenkins:

[30:35] So they’re kind of right on the midpoint. If they get, you know, it’s $7,000 a kilowatt for phase one, and they can further reduce those costs in phase two, they’ll basically be starting that learning curve right where our paper had them landing. And that’s exciting because what we found in that paper was that even if you don’t have some kind of long-term net zero carbon policy driving

Jesse Jenkins:

[30:56] Decarbonization, just having the investment tax credit in place for projects commencing through 2032, which is the current law. Is enough to potentially bootstrap, along with development at those near-field, high-quality initial sites, is enough to bootstrap the learning curve to a level that could take geothermal to be about 100 gigawatts or more, depending on natural gas prices, of U.S. power by 2050. That’s the size of the U.S. nuclear fleet.

Jesse Jenkins:

[31:20] So I was keen to see those numbers. They also talked about the length of the laterals that they’re drilling. Again, that’s kind of right in between where we saw things at Project Red, which we did have data on when we started our paper and where we are anticipating they would be at commercial scale. The one big unknown that is not in the filing and won’t be because we won’t know this until they’ve completed flow tests for some period of time is how much, what is the flow rate of circulation of the fluid through the wells? That’s the key determinant of basically how much energy you can extract per well you’ve drilled. So we know, you know, they’re drilling in, they’re reporting the temperature. We know how many wells they’re drilling. What we don’t know is how fast they’re going to be able to circulate water through those fracture networks, basically how much circulation porosity connection do you have across the fractures. And that plays a huge role in the effective output per well,

Jesse Jenkins:

[32:11] and therefore the unit economics. And so that’s the one key kind of big unknown right now is have they achieved the flow rates that they need to for commercial operation. And there’s also another wrinkle, you know, if you don’t kind of get natural flow rates that are there for a long, like 30 year operation, because you are extracting heat slowly from the rock around the wells, you could potentially pump up the injection pump and use higher pressures to force greater circulation through the rock. And that can get you to higher flow rates that would boost your near-term production, but the effect of that would be to extract heat faster and shorten the longevity of that reservoir.

Jesse Jenkins:

[32:49] So that, you know, the natural flow really does impact the unit economics. Either you get less heat for the well or you can pump more and get more heat out, but you have a shorter lived well and you’re going to have to drill more in the future to kind of top up the production of that facility. And so that’s still a big open question that we won’t really know until they’re operating at Cape Station for some period of time.

Robinson Meyer:

[33:09] And I would imagine that increasing your injection rate also increases the risk of something that they talk about in this report, which is induced seismicity, which we’re not going to have time to get to maybe in this show. But they let’s just say that they flag it as a risk in the report that doing fluid injection at depths could increase the seismic risk. And, you know, it’s potentially a difficult to ensure risk if that were to happen.

Robinson Meyer:

[33:32] I just want to flag a few more things in this report and then we’ll wrap up. I think the first is that we got a sense of what their portfolio looks like after Cape Station. So they think Cape Station is a 4.3-megawatt resource in Utah.

Jesse Jenkins:

[33:45] Which you should pause and say that’s more than all geothermal in the U.S. today at that one site.

Robinson Meyer:

[33:49] They think at this one site they can basically double U.S. geothermal production. But then in some ways it’s only an entree to what they claim is a ready to build site in Nevada. What they call the Corsac site, which is 8.1 gigawatts on 41,000 acres. And then after that, they actually have a, they say, now they don’t describe this as ready to build, but if they’re, as they talk about the acreage that they have under lease, they have a 10.8 gigawatt site in Utah, and then a series of sites between, you know, 1.4 and 7 gigawatts throughout Nevada and Idaho, actually a lot of sites in Nevada that they claim, you know, are explored resources, or at least lease acreage that they have under lease with a good resource. And it kind of gives us a sense of where they might expand, let’s say, through the early 2030s, if Cape Station is successful. Matt, is there anything else we should add? You know, if there was one more thing in this S1 that stood out to you, what might it be? I have some I have some suggestions, but I want to hear what you what you would pull out.

Matthew Zeitlin:

[34:54] I thought one thing that was interesting is that they’ve adopted a very tech industry-like thing in that the founders will be in control of the company, seemingly indefinitely, almost no matter what. They’ve adopted this dual class share structure, which should be familiar from, say, Google or Meta, where the founders, I think, own shares, I think, have 40 times the votes of the common stock that they’re selling. Yeah so this is a this is a little interesting because the people who run you know infrastructure companies control a lot of capital including a lot of capital it’s not really quote unquote the shareholders so giving them kind of this extra level you know because they’re raising all this project finance so giving them kind of this extra level of control i guess the idea is that you know maybe they don’t feel pressured to sell the company or to develop too quickly or it’s the type of thing that again is more familiar from the software and then also weirdly enough the media world. A lot of innovation, dual class share structures are created to keep the Murdochs and Salzburgers in charge of their various companies. But yeah, I mean, it’s not something you see a ton of in like publicly traded oil and gas companies. No, that’s right. I mean, it does think, I do think it sort of signals as it does in the Murdoch example or the Google example, like a long-term interest in control of this company, like they’re in it for the long term, which you can read in different ways, right? But yeah, that is a quite distinct feature of this filing.

Robinson Meyer:

[36:15] Well, there’s a lot more to talk about. It’s a big filing. Matt has a great story on Heatmap that we’ll link to in the show notes. I recommend that everyone reads it because there’s actually stuff in that story that we didn’t get to on this call. Until then, though, we’re going to have to leave it there. Jesse and Matt, thank you so much for joining us. It’s always great to have two friends.

Jesse Jenkins:

[36:32] It’s fun hanging out with you, Rob. Thanks. Always, always.

Robinson Meyer:

[36:40] We’ll leave it there. Stick around at the end of the show, by the way, for a message from our sponsor, Salesforce. So excited about that. We’ll be back next week at the usual time with a new episode of Shift Key. Until then, Shift Key is a production of P-Map News. Our editors are Jill Inkubman and Nico Lorichella. Multimedia editing and audio engineering is by Jacob Lambert and by Nick Woodbury. Our music’s by Adam Pramilow. Thanks so much for listening. We’ll see you next week.

Mike Munsell:

[37:08] Hi, my name is Mike Munsell, and I’m the Vice President of Partnerships with Heatmap. I recently spoke with Sunya Norman, the Senior Vice President of Impact at Salesforce. Over the next three episodes of Shift Key, we break down how Salesforce approaches impact, covering everything from its AI energy score to climate tech and resilience investments.

Sunya Norman:

[37:28] I’m Sunya Norman, SVP of Impact and Salesforce. I think I have the best job in the world. Essentially, my team of impact professionals helps to create Salesforce as a platform for change, focusing across a broad range of issues from environmental sustainability to philanthropy to supporting nonprofits with leading technology and also engaging our employees in volunteering and community work.

Mike Munsell:

[37:55] And given your work on impact, how does Salesforce think about sustainability, especially in regards to AI?

Sunya Norman:

[38:01] It’s a strategic focus for Salesforce. It’s really become a business imperative for large publicly traded companies like Salesforce. It’s also a core value. And the way that we think about it is operationalizing that core value and embedding sustainability across everything that Salesforce does, from our purchasing to how we manage our offices to even how we deploy our AI technology. As Salesforce is positioning ourselves to be a leader in agentic technology, of course, we need an accompanying sustainability strategy.

Sunya Norman:

[38:38] We’ve published something called our AI Sustainability Outlook, and essentially that shares our three pillars of AI sustainability. The first is smart demand. This one means using AI wisely. So for us, Salesforce Agent Force is built to deliver high performance while also minimizing environmental impact. And we’re helping our own customers understand the environmental impact of their agent force deployments so they can make informed choices. And that’s also where we see the AI energy score coming into play. The second pillar is efficiency. This is about the entire value chain of AI from the chip to the servers in the data centers to the data centers themselves. But it’s also where we’ve had the pleasure of collaborating with our AI research team. And that team is really inspiring, really innovative folks who specialize in developing domain-specific AI models. And these models are designed to excel at a really specific task. So that’s the domain-specific part, while consuming much less compute, in turn, much less energy than the large-scale frontier models.

Sunya Norman:

[39:53] The third pillar is what we call clean supply, and this is a continuation of a journey we’ve been on for a long time to support the world’s clean energy transition. For many years now, we’ve been really proud to source 100% renewable energy for Salesforce’s global operations. Now with AI on the scene, we’re thinking about how can we invest so that the data centers, the power AI infrastructure are sourcing clean energy, whether that’s low carbon energy, think wind, solar, newer technologies that hyperscalers are hoping to scale like geothermal or nuclear. It’s a really exciting space. And we’re hoping to bring strategic investment through our philanthropy and through our policy engagement to make sure that we’re on the right trajectory with our clean energy transition.

Mike Munsell:

[40:44] And can you give our listeners an overview of the AI energy score? Tell us more about that. And why is Salesforce the right company to create such a benchmark?

Sunya Norman:

[40:53] Let me maybe start with what is the AI energy score? It’s a collaborative effort, something that Salesforce launched with Hugging Face in partnership with a bunch of leading tech and AI companies, and the goal was to create a standardized way that we all evaluate AI energy use and something that we’re gifting to the industry. With the onset of AI, there’s a lot that hasn’t been standardized or developed yet, At its core, the AI energy score is essentially a benchmark. It measures the different models and energy consumption related to common tasks those models might perform. If you’ve ever bought an appliance like a dishwasher or a washing machine or a toaster, I was really inspired by something called the Energy Star. And that allows consumers to not have to nerd out and go super deep into how many kilowatt hours an appliance is going to use, but just have a very simple five-star system of what is good and what maybe has room for improvement. So the idea is that the AI energy score would enable technology leaders and decision makers sourcing AI models in a similar way, essentially giving them the data they need to make meaningful decisions.

Mike Munsell:

[42:12] And can you talk about what adoption looks like for the AI energy score today, what success looks like more broadly for Salesforce for that AI energy score?

Sunya Norman:

[42:21] Yeah, we launched our first version of the AI energy score back in 2025. And then we actually came out with a version two that builds on that foundation, has additional reasoning tasks that we introduced, but also expanded to additional models. What success has looked like for us at Salesforce is integrating that information into our own internal benchmarking. And we’re actually even working on incorporating energy data into our AI model cards. You can think of them as almost like the nutrition facts on the back of a food item so that we have more information internally and can help our customers have the data that they need to make decisions that are more sustainable.

Sunya Norman:

[43:05] Of course, we would hope for widespread adoption. Really, something doesn’t become a true standard in the industry without that adoption and scaled usage. Transparency, in my view, leads to greater trust, arming customers, technologists, stakeholders with the data they need to feel like these models and this information is credible. The data isn’t just for data’s sake. Again, it’s about making decisions so that energy efficiency and sustainability can be top of mind and can become a core design principle for AI systems and technologists. Today, sadly, it’s probably more of an afterthought, and we want to make it easier for this to be a consideration alongside things like performance and cost of use.