Climate Rising

Using AI to Unlock Geothermal Energy at Scale: Zanskar

Climate Rising podcast avatar

Geothermal energy has long been seen as a promising, but challenging, clean energy resource. Unlike wind or solar, geothermal requires finding heat hidden deep underground, often with little surface indication of where to look. Joel Edwards, co-founder and Chief Technology Officer of Zanskar, joins Climate Rising to explain how advances in geoscience, data science, and machine learning are transforming geothermal exploration. By combining publicly available geologic data with modern modeling techniques, Zanskar is working to reduce the risk and cost of finding new geothermal resources. The conversation explores how geothermal systems work, why exploration has historically been challenging, and how Zanskar’s approach is enabling “blind discoveries”—finding viable geothermal resources that lack surface expressions such as hot springs or geysers. Joel also discusses the economics of geothermal, the role of data centers as a potential catalyst for growth, and what it will take to scale the industry.

Resources

  • Zanskar Geothermal: Startup focused on geothermal exploration using data science and machine learning

  • US Department of Energy – Geothermal Research: Public R&D efforts advancing geothermal exploration and drilling technologies

  • Gabb’s Valley Discovery (DOE project): Example of a “blind” geothermal discovery using modern geoscience + data science

  • Geothermal Steam Act (1970): US regulatory framework enabling geothermal leasing on federal land

Transcript

Editor's Note: The following was prepared by a machine algorithm, and may not perfectly reflect the audio file of the interview.

Mike Toffel:

Joel, thank you so much for joining us here on Climate Rising.

Joel Edwards:

Thanks for having me, guys.

Mike Toffel:

So, Joel, why don't we begin, if we could, with a brief overview of your current role and how you got there?

Joel Edwards:

Yeah, my name is Joel Edwards, a geologist by training, and I grew up in Iowa, but I moved west for school. And one of the things I learned when I moved west is that there are such things as hot springs. And I visited my first hot spring when I was an undergrad in school, and it blew my mind. Couldn't believe that these things exist. They don't make sense. How is there hot water at the surface? Nothing is heating it. That was my first interaction with geothermal. And then I had the good fortune of getting a small internship at a startup, a geothermal startup in Utah, and fell in love with the concept. I got to see projects in development at the time in Utah and New Mexico. And the concept seemed like a no-brainer. And maybe we can get into the concept later. And, from there, I went west to do more geothermal projects in Nevada. And then in the industry tanked, and I took a hiatus and went to more grad school just to get better in the geosciences.

I studied seismology. And then when I was done, I reconnected with an old undergrad friend named Carl. And we started exploring this concept of improving exploration outcomes for geothermal projects by bringing in modern geoscience and data science tools into the problem.

Mike Toffel:

Great. So, let's do a 101 on this topic of geothermal. Now, my own background on geothermal is pretty limited. I know that geothermal involves drilling into the ground to access that warm water, which some folks may think of as a source of power for a heat pump, like a ground source heat pump. Others may have heard that there are these circulation systems that take water from underground and use it for ambient heating or maybe some industrial processes. Take us through, like, what's the geothermal 101? What is it? What's it used for? What's hard about it?

Joel Edwards:

Yeah. The 101 is that as you go deeper into the earth, it gets hotter. It turns out that the center of the Earth is very hot. Some people think it's the temperature of the surface of the sun. We don't know. We've never been there, but we at least know it gets hotter as you go into the earth. But what's also true 101 is that if you go into the earth at different locations, it gets hotter at different rates. Meaning if you drill a well,

In Boston, versus if you drill a well in Salt Lake City, the temperature at the bottom of the well will be different. So that means there's this horizontal heterogeneity in the Earth. So those two facts are true. The deeper you go in the earth, the hotter it gets, but also it gets hotter at different rates in different locations. And so that second aspect means that there is an exploration element to finding the hottest and most accessible resources at the lowest cost.

And that's the space where Zanskar lives is we spend our time, and that's the space where a lot of geothermal companies live historically, which is where we can go to get the highest temperature, the largest quantity of heat at a location with a well field.

Mike Toffel:

Got it.

John Mulliken:

And so just for our listeners, what does it look like when you succeed, when you find a great well that's got a high temperature at a reasonable depth? What do you do with it? How do you put that to use?

Joel Edwards:

Yeah, you already named one of the factors. There are two factors to success for a geothermal well. It needs to be hot. That's one. And then the second factor is that the well needs to flow because the vehicle that carries the heat is brine or steam. It's a fluid. And if the well doesn't flow, then you can't get the heat out of the rocks. So, you need those two conditions, temperature and flow. And so, a successful geothermal well is one that produces higher than generally 280 degrees Fahrenheit, which is a little bit on the low side, and at that well flows hundreds to thousands of gallons per minute. So, a garden hose at your house, you plug that into your house, that garden hose, if you crank it all the way up and it's full flow, that's about 20 gallons per minute. So, a geothermal well needs to flow at a minimum and order of magnitude more than your garden hose, and ideally, more than two orders of magnitude more than your garden hose. And that will get you to the scale of megawatts of thermal energy coming out of that well that you plug into a power plant that converts it to megawatts of electricity.

Mike Toffel:

So, so far, you've highlighted three things: the temperature of the well, the pressure of the flow, and also the depth. The pressure and temperature, think both are about the value of the asset. And I'm guessing the depth has to do with the cost of accessing it. Is that right?

Joel Edwards:

That's exactly right. Yeah, as you drill deeper into the earth, it gets more expensive because it's not just creating a hole. You have to create a hole, and then you also have to construct a well. And that means you have to put steel into the ground and you have to put cement so that the well lasts for long time frames and so that the well can support a pump.

Mike Toffel:

Got it. And the output of this is used for electricity, for ambient heat, and for process heat? Are those the three outputs of that?

Joel Edwards:

Yeah, exactly. That's, that's 99 % of the use cases are some combination of power and heat. And that heat is sometimes pushed into greenhouses. It's sometimes pushed into buildings, et cetera, in the case of Iceland and others. And then there's one facility in the world, one geothermal facility in New Zealand, where they make the power, and then they also extract colloidal silica from the brine. So, they produce a mineral product that has commercial value.

But there's only one facility in the world that I know about that's doing that right now. That's obvious, there are a lot of folks really excited about lithium, pulling lithium out of geothermal brines. And there are some projects that are in development in Southern California, trying to do this. But yes, 99 % of the use cases for geothermal well fields are power and heat.

Mike Toffel:

Got it. So, my last 101 question. So, unlike an oil well, which has a fixed amount of material that you can extract from it, are we worried about the heat ever running out in a well setting, so that it becomes diluted over time? Certainly, the minerals you talked about create a question about sort of, it be there forever. But the heat, is it so plentiful that we don't really have to worry about that, or is it something that is actually of concern?

Joel Edwards:

Yeah, it depends. There is a natural recharge, a heat recharge in the earth. So, there's a natural amount of heat that's flowing through rocks. That's called the recharge rate. And generally speaking, commercial well fields exceed that recharge rate. And so, you are pulling heat faster out of the system than it is recharging. What you try to do, though, in these geothermal systems, is you set up production wells that are pulling heat out of the system, and you set up injection wells that are putting that mass back into the system. And if you set that up correctly, then the volume over which you extract heat is so big that you will not have any temperature decline. Does that make sense?

Mike Toffel:

Let me just pause for a second. So, the volume that you extract is so big. Did you mean so big or so small?

Joel Edwards:

So, you space your wells out so far apart that between your production and injection wells, you're producing the same mass as you're injecting. And if you spread those out far enough apart, the volume is so big that by the time the brines get back to your production wells, they're heated up, they haven't cooled down at all. That's a sustainable commercial development, and that is the gold standard. And there are well fields that meet that gold standard. There's no temperature decline in them.

John Mulliken:

So, this is pretty interesting in the sense that we've been drilling holes in the ground in order to extract energy for a long time, but they've primarily been oil and gas. And I'm guessing that a lot of the techniques for exploring came initially from those oil and gas techniques. Mean, tell me a little bit about traditionally how you went exploring for a while. Were you just looking for steam coming out of the ground, or was there more to it?

Joel Edwards:

Mm-hmm.

John Mulliken:

We'll talk more about Zanskar and what you're doing, but lay the groundwork for us. How did you find one of these in the first place?

Joel Edwards:

Yeah, the original geothermal boom was from the 1960s to the early 1980s. And in those days, all of those geothermal explorers went to either locations that had hot springs or geysers at the surface. So, there was natural evidence of a system there, or they went to locations where we accidentally found a hidden system. And these are cases where you had homesteaders, right, who were drilling water wells, and they accidentally hit steam.

You had mineral exploration companies where they were looking for some mineral product, gold or whatever, and they accidentally hit steam. And then there are some cases, as well, where oil and gas explorers accidentally hit steam. Those are the discoveries in Southern California. And so, in the 1960s and 1970s, the majority of the exploration and development efforts went to those two locations, natural evidence of a system or accidental discoveries. So, there wasn't the exploration problem was really a data curation problem. It wasn't really boots on the ground out exploring for new areas. There were a few companies doing that, but the majority of the dollars went towards known sites where there's probably something here.

John Mulliken:

And so, I mean, you might call this the wildcatting era. This is the time to let's do our best efforts to try and find some geothermal features. How did it work? Mean, was this a boom industry? Was this really something that was going well economically? And if so, I mean, it sounds like that ended. How come the majors began to walk away?

Joel Edwards:

Yeah, there were two things that happened in that era. One, the price environment, the power price environment was great, right? Oil and gas products were relatively expensive at the time, so the power prices were up. And so, these folks were able to go out and contract. At the time, there was a hundred-dollar-per-megawatt-hour tariff, a feed-in tariff in California. You would get a hundred dollars per megawatt hour for 10 years. Inflation-adjusted. I don't know what that number is today, but it's a big number. Power prices were excellent at the time. And then two, the United States got its act together, and it passed the Geothermal Steam Act in 1970. And it set up a regulatory framework for companies to lease geothermal rights from the federal government. And out West, where these resources live, the federal government is the biggest landowner. You need to work with the federal government to get access to these resources. So, between the price environment and the regulatory environment, that was the right mixture that sprang billions of dollars, inflation-adjusted, to come into geothermal and develop all these projects. And then in the 1980s, the oil and gas prices came back down to earth, and the power prices came back down to earth. And basically, the oil and gas majors got out of the game.

John Mulliken:

So, it sounds like there's a period of quiescence, right? Where there just wasn't much activity for a while. And into this comes Zanskar. Here you've gotten into the industry. Tell me a bit about what Zanskar's approach is. Mean, if it was so hard to find these features, how are you doing it?

Joel Edwards:

Yeah. So, this quiescence period wasn't on the industry side; there wasn't a lot of improvement in how we find systems. But during that quiescence period, the Department of Energy continued to invest in geothermal R and D, particularly in the exploration bucket. How do we go about finding systems from scratch? And then in the drilling bucket, how do we get more heat out of these systems? And so, they did some really interesting things during that quiescent period. They proved at the pilot scale that you could, from scratch, go find a blind system. They actually discovered, I think it was 2017 or 2018, in Western Nevada, called Gab's Valley. And they made that discovery using a sort of geo-science data science blended approach. And that discovery was really helpful for us when we emerged in the 2020 era. And when we started incubating this concept, are there new geoscience and data science tools that we could commercialize that could drive down exploration risks? We were able to go to the market and point to that Gabs Valley discovery, which was funded by the Department of Energy, and say, look at what they did on a very tiny budget, very small resources on a very short time scale. Wow, if you applied commercial resources to this, this looks really exciting.

John Mulliken:

So, tell me a bit more about data science, or what some of our listeners might think of as AI. But let's talk about it as data science. Where does it fit in? Is it doing something that a geologist couldn't?

Joel Edwards:

Yeah, there's this issue in geothermal. So, in oil and gas, there's a type of data in oil and gas that does a really good job at finding hidden reservoirs. And that type of data is called seismic reflection imaging. It's seismic waves and how they bounce and interact with reservoirs. And you could build images, and it will show you where reservoirs are. In geothermal, that doesn't work.

John Mulliken:

So, like what an earthquake would look like, except you're creating these waves under the ground.

Joel Edwards:

Exactly. You're creating the waves, and then you have your receivers. You listen for the reflections, and then you build images. And that's a really good toolkit for finding these systems. And geothermal, unfortunately, it's not that simple. There's not a single type of data that you can collect to tell you if a system is there or not, besides just drilling a well and measuring heat flow. And so, you end up with a multi-dimensional problem where you're looking at many different types of data sets and the combination of those data sets to try to infer where a system is.

That's the data science problem. You look at a series of independent variables to try to predict a dependent variable. And then, the data science tools that emerged in the 2010s in the commercial space, like coming out of Silicon Valley, right? Those, those were applied to geothermal by the DOE in 2010, 2015, arrow to some success, right? The Gaps Valley's success. So, when we started talking about this, we were thinking about data science.

But also at that time, machine learning tools were emerging from the commercial space, right? The transformer paper was published in 2017 by Google, right? Which solved the, uh, the parallelization problem associated with large language models. And that created this, and then GBT models were published in 2019, 20. So when we were starting to think about how to explore for geothermal, machine learning was, was blossoming or early days in the commercial sector, and machine learning is just data science on steroids, right? You're able to look at way more dimensions of data to try to predict outcomes.

Mike Toffel:

So, can you talk us through some of those data elements? Like, what are the XX’s in you?

John Mulliken:

Interesting.

Joel Edwards:

The Xs are simply, I think, intuitively, you could think of some like the rock types, right? The physical properties of those rocks. And then, you know, for us, we're really interested in fracture zones because these systems live in fracture zones. So, all the properties associated with fracture zones, how old are the fracture zones? How fast are they dilating all these things? There are different types of data that can be pulled from satellites. They could tell you, like how the surface of the earth is moving over time or how it's moving in the horizontal dimension over time, things like that. We look at basically suites of geologic data that are associated with the rocks or the deformation of the rocks, things like that. And then we make predictions on whether there excess heat flow at that location, meaning whether there is a geothermal system.

Mike Toffel:

And so, these are coming from satellites and also from on-the-ground measures or drone measures. Where, besides satellites, are these X's being obtained?

Joel Edwards:

Yeah, really three buckets. Have satellites, you have airborne, and then you have boots on the ground. And generally, across all of that, the sensors, the sensing stack has gotten a lot cheaper and more performant over the last 10 to 15 years across all sensor types. That's driven by better batteries. That's driven by better memory, right? So, you can increase the sampling rate that's driven by better cameras.

The sensing stack itself has gotten much cheaper and much more lightweight so that you can stick them on drones. Right. You can stick them on these other sorts of carriers. And so, we're able to collect a lot more high-fidelity data over larger areas, at a much lower cost today than we could even like 15, 20 years ago.

John Mulliken:

You mentioned earlier, alpha fold, right, which is all about protein folding. And it sounds to me like what you're doing is you're looking for features in the ground that might be similar to that. They solved this problem that was previously thought to be unsolvable with this new type of model. How does that affect your work?

Joel Edwards:

Right. AlphaFold is a really good analogy because AlphaFold incorporated physics into the learning. Whereas a large language model, like a GPT model, has no physics. Right. So, when we build, when we try to build predictions, we put physics into the sandbox. We bound it by physics. And then we also bound it by the data that we have on-site at that location.

And so, between those two domains, we make predictions about what the reservoir might look like underground. So, we're iterating over that. Even with the data you have, and even with physics, there's still a lot of possibilities for what that reservoir can look like. And that's the, that's the magic of applying machine learning tools in this space, as a human, when you look at physics, and we look at the data on hand.

It's a ton of work to build a single model that can fit that data and be bounded by physics. takes a lot of effort. Machine learning tools can basically do that at a much faster pace, but you're having humans in the room making sure that the sandbox is correct. That it's not hallucinating, that it's not cheating. All these models are very, very good at cheating. They're very good at learning things that you don't want them to learn. Right. So yeah, I think, I think alpha fold in the geologic context, like these are different applications of machine learning in the physical world.

Mike Toffel:

And so, what of all these is proprietary to Zanskar versus public? The datasets, some of them sound public or at least licensable by competitors as well as you. Do you have uniquely proprietary data, or is the proprietary aspect of your efforts really in the algorithm or elsewhere?

Joel Edwards:

Yeah, the data in the data science context, the independent variables, right? Those are generally data coming from publicly funded data collection efforts by the USGS and others. We're encouraging the DOE and others to go out there and collect more of these independent variables that can benefit everyone. The Zanskar piece is that we go out and we collect that dependent variable, meaning when we make a prediction, we go out and we test that prediction. And that prediction for us is Heathrow.

We're trying to predict locations on the Earth where there's more heat flow, more heat coming out of the ground, based on all these independent variables, the properties of the rocks at that location.

John Mulliken:

So, you have two different kinds of models, right? My understanding is that you're looking at the earth in two very different scales. Just give us a sense for what those different types of models are.

Joel Edwards:

Yeah. We've split it into two model-building phases, and maybe other people are trying different sorts of splits here, but we have that regional model-building phase where we're trying to just discover locations that have excess e-flow. That's the whole problem space. So that means you're looking at the scale of like basins, or in other words, I guess basins is more of a westward term because we have the basin in range, but think on the scale of like counties.

And we're trying to predict in a county, a location in that county where there's excess heat flow at the scale of a city in a county. Maybe that's the right way to put it. That's the first prediction problem. And then the second prediction problem is when you find that location that has excess heat flow, you have found a geothermal system. And now you need to model what that geothermal system looks like underground so that you can put a well into it to extract heat out of that system over long timeframes.

Mike Toffel:

So, we've been talking about this thing in concept, but I wonder if we can walk through an example. So, for example, you have this New Mexico project called Lightning Dock Geothermal Area. Can you tell us a little bit about the origin story there and how you found that as a promising location, and then what happened after that?

Joel Edwards:

Yeah, Lightning Dock is one of those accidental discoveries. It was discovered by homesteaders in the 1940s. They were drilling water wells, and they accidentally hit steam. Um, and the first publication of this was in 1954 or 56, there was a Nature paper that talked about these homesteaders hit steam in this location in New Mexico. Isn't that weird? Wow. That's really odd. Nobody would have expected that. So, credit to the homesteaders for finding this.

And then a few decades later, these home centers built greenhouses at the location, and they started using the heat, and they grew roses. And it actually became the largest rose-growing operation in the U.S. It's not around today. I don't know the details of why it didn't survive. My understanding is that the 1990s were not great for rose-growing operations in the U S, something to do with NAFTA. I don't know the details. This is what the families, you know, children say. But anyway, so they used the heat for roses. And then in the early 2000s, a geothermal company started to look at the site to make power, and a power plant was built and constructed in 2013, and then expanded in 2019. And unfortunately, the power plant had reservoir issues a few years later, and it started to have anyways, production issues. They weren't generating as much power as they thought they would. And so, the power plant came up for sale in 2024, and Zanskar bought it. And we were excited about the system because when we put the system into our models, we thought that there was a lot more upside to the system than the current well-filled was getting out of it. So, we bought the field in 2024. So, we didn't find it. The homesteaders did. But we were excited about its potential. End of 2024, early 2025, we sputtered a well, and we drilled a new production well at the site, much deeper into that same system. And we hit bonanza, whatever the word is, we hit a big resource, and we plugged that well into the power plant, and we completely repowered the entire power plant with that one well. And that well today is the most productive pumped geothermal well in the country. So, there are over 200 pumped geothermal wells throughout the United States, and it's the most productive, meaning it's pulling the most thermal energy out of a system of any single well in the United States.

John Mulliken:

So, it went from really being what you call a stranded asset, meaning it wasn't worth much to anyone. And it sounds like you actually were able to pick it up fairly inexpensively into something that was highly productive.

Joel Edwards:

Yeah, that's right.

Mike Toffel:

So let me just take a detour here for a question about game theory and contracting. So, you bought it from a prior company that had been trying to do what you were doing, which is to take the heat, generate power. They run out of steam in a cajoling way, in a casual way.

And then you said, actually, we would like to buy this for much more than you think it's worth. Like, wouldn't that cause that company to say, hang on a minute, Zanskar thinks there's a lot of value here. Maybe we should go and do our own due diligence to see why they think that, and actually do a little more test driving. And then they would be enjoying the riches that you're enjoying now. Why did that not happen?

Joel Edwards:

Well, Zanskar didn't really have credibility at the time. Zanskar is a startup, and we were entirely focused on finding new systems at the time and how to model those systems. And we didn't have an operating power plant at the time. We had no credibility that we knew what we were doing. So, I'm sure they looked at us and thought, good luck.

Mike Toffel:

So, they basically thought you were probably overpaying, and they were delighted to walk away from what they thought was a stranded asset.

So, you'd mentioned blind before, like a blind well or a blind site. Can you just unpack what that means?

Joel Edwards:

Yeah, just simply, there's nothing; there's no natural feature at the surface to tell you that there's something hot under the ground. No hot springs, no geysers, nothing. That's the lightning dock system was nothing, well, the farmers, there was nothing to tell them that there was a system there. And this is, this is when you look at just resources generally, oil and gas, minerals, water, geothermal.

The majority of systems do not make it to the surface of the earth and express themselves. The majority of resources are hidden underground. That's what makes them particularly hard and challenging to find. And oil and gas, if you look at the oil and gas history and minerals, most of these always start with the natural surface feature expressions in commercial applications. So, oil and gas production started in Pennsylvania, not because Pennsylvania has the biggest resources, but because there were literal oil seeps at the end. And then it moved to South Texas, the Spindle top discovery. There were gas seeps at the surface. Right. And then, in the 1920s and 1930s, oil and gas started to make blind or hidden discoveries like West Texas, right? The Santa Rosa discovery well was in 1923, 1925, something like that. And that, that discovered the Permian basin.

But there was nothing at the surface to tell you the Permian Basin was there.

Mike Toffel:

Got it. So, talk us through a blind discovery because the Lightning Dock is quite different. That's where there was already knowledge of a waning well that you sort of boosted with your technology. Talk us through an example that sounds like it's more your core business, which is finding these blind opportunities.

Joel Edwards:

Yeah, that's Sanskar bread and butter. This is where we started our first venture capital dollars went into this discovery problem, probably particularly because we felt like we could do it, at lower costs than if we started with, let's buy a power plant and drill a deep, expensive well. But the discovery starts from scratch; it really starts with you building regional models, and you make regional predictions.

And then you go out boots on the ground, and you test those predictions. So, you need to go to those locations, and you need to measure heat flow. And this is the thing we've been doing for five years. And the majority will say, we have missed way more than we found. Like what's that Michael Jordan quote? You miss the shots. don't shoot or whatever. Like Michael Jordan missed more game winners than he's made or whatever. Like that's definitely true for Zanskar. We have missed way, way more than we've hit,

We basically built a workflow that is much lower cost for measuring heat flow at locations than our predecessors. So, like our misses are much less painful because they're much cheaper. But the discovery problem is essentially making regional predictions, driving out to that basin in wherever, Utah, Idaho, New Mexico, Arizona, out West somewhere. And you have field crews that stay in that area for a few weeks to a few months, and they're making all these measurements across the area, looking for this excess heat flow prediction, testing whether it's true or not.

John Mulliken:

So, it sounds like there hadn't been a blind discovery for 30 years or something like that. Tell us where you made a blind discovery?

Joel Edwards:

Yeah. So, we've been quite a few at this point, and we have announced publicly one in Western Nevada. And the reason we've announced it is because we've done enough drilling to confirm that yes, this site can generate utility-scale power. So, we're comfortable announcing it, and a site we've named Big Blind. It's in Western Nevada. And it's a site where if you or I went out to the site and we spent some time there, you would not know that there was a system underground.

It's a site that we originally discovered back in 2023, actually. And then over the course of the last couple of years, we collected data, and then we eventually drilled confirmation wells into the resource to confirm that it was power quality. And it's a really, uh, it's a very big system. It covers many square kilometers. And this system has similar reservoir attributes to other large systems in Nevada that generate over 100 megawatts. So, we're in the early days of the system. We're going to be following up and drilling more wells into the reservoir and so forth. But it's a very large hidden system in western Nevada.

John Mulliken:

And is it deep? One of the things we talked about is that drilling is expensive. The further you go, the more it costs. It sounds like a lot of previous-generation wells hadn't been that deep. I mean, what's it like to go drill for a blind system?

Joel Edwards:

Our confirmation wells went down about 3000 feet and hit reservoir or power quality temperatures. And so, by that standard, it's a pretty shallowly expressed system. But what we're excited about with geothermal systems is that these systems do go deeply into the crust, and the deeper they go, the hotter they get. And we think the bigger they get. So, this again is a similar story to oil and gas, where the original Permian Basin Discovery well was a very shallow well. And when they first looked at that resource, they thought it was a shallow resource, probably small. A hundred years later, you're still getting oil and gas out of that basin because you've moved deeper and deeper and deeper into that basin. And so now you have production wells that are below 10, 15,000 feet in that basin. And now they're, so forth and so forth. That story is going to play out in geothermal, and it just has not yet.

So, if you look at the geothermal well fields that are in operation today across most systems in the Western US, they're very shallow wells. They're three- to 5,000-foot-deep wells into shallowly expressed systems. But we know those systems go very, very deep below 15, 20, 25,000 feet. And so, as these power prices stay high, there are all economic reasons to go deeper into the system and make them bigger. And so, I think that same story that played out in oil and gas, where you start shallow, and you move deeper, I think that's going to play out in geothermal.

Joel Edwards:

And these systems that are, maybe making 30, 40, 50 megawatts today, I think in the future, they're going to make a lot more power because you're going to move deeper into them.

John Mulliken:

Zoom out for us for a second. Mean, you've mentioned that there might be a terawatt of power out there in these geothermal systems we just haven't found yet. I mean, what gives you the confidence that that might be true? Just mean, what is a terawatt? How much is that?

Joel Edwards:

That's a lot. Terawatt is a thousand gigawatts and a gigawatt is a thousand megawatts. So, it's a million megawatts. And your average geothermal well is, five to 10 megawatts. So, a million divided by 10 is a hundred thousand geothermal wells, which would get you to a terawatt. And right now, in the U.S., we have, you know, roughly, you know, we have somewhere around 500 operating geothermal production wells. So, we are what's a hundred thousand divided by 500.

2000 times, 200 times smaller, we need to grow a lot. Oil and gas have that order of magnitude of wells. They have hundreds of thousands of wells. So, the analogy is that we would scale to oil and gas size in terms of the number of wells, and we would get a lot more power, a lot more heat out of the earth. Yeah, go ahead.

Mike Toffel:

 So let me ask a follow-up question on growth and the sort of business model. So, as I'm hearing this story, there are several stages to actually get to operating well. There's the technology stack that you mentioned, which uses mostly publicly available or subsidized data applied to your algorithm to say, here are the most promising areas. Then there are the test wells to see whether those predictions are accurate or how accurate they are.

Then there's the construction, the design and construction of the plant, and the operating plant. And then there's the operation of that plant. So, in this typology that I just made up, there are four steps, perhaps. And one could imagine Zanskar specializing in just stage one, with the AI machine learning piece, or vertically integrating into all of them, or choosing a subset. What are you doing, and how did you make that decision?

Joel Edwards:

Right now, we're doing all four steps in the process. We operate a power plant, we discover new sites, and now we're developing those discoveries into power plants. So, we started at the beginning and at the end, now we're doing the middle as well. If geothermal were a bigger industry, it wouldn't be so niche. I think there would be a lot more opportunity to discover and then bring in partners that do the middle part, and then bring in other folks that do the end part. Depth does not exist in geothermal. Think if you fast forward and the industry grows, that will exist in geothermal. It does exist in oil and gas. It's super common. It's super common in the mining industry. We have junior mining companies, which are small outfits. They raise a couple of million dollars in the public markets. They have a team of 10 to 20 people who go out. They prospect, they stake claims, and then they sell ownership rights to larger companies that come in and do more and more, and so forth until it's an operating mind with a different operator. Right now, Zanskar is doing it all because we have to do it all.

John Mulliken:

You've mentioned that there are real benefits to doing some of the operations yourself. Tell us a little bit. I mean, does running Wells make you better? Does it help you with the other parts of your business?

Joel Edwards:

Yeah. I probably should have mentioned that. Thanks, John. You're trying to understand a reservoir at the end of the day. And when you discover a reservoir, you know certain attributes about it, but there's a lot you don't know about the reservoir. And as you can imagine, if you operate a reservoir, if you have a well that's pushing and pulling mass out of the reservoir, you're going to learn a lot about that reservoir. So, absolutely. If you can do it all, then your intellectual property you're going to understand these systems much better than people who don't do it all. So that's the other benefit of trying to do it all is we're still in the early days. We feel like there's a lot about geothermal systems that we don't know yet as an industry. And so, we want to be on that sort of frontier discovery piece. Because there are probably ways that we could discover them that we don't quite see yet, because we don't quite understand fully how the systems form and live and die.

Mike Toffel:

So, what I'm hearing is that, in a sense, there are spillover benefits by operating across all of these domains that help your prediction algorithm; actually, you have a deeper understanding of the outcome.

So, we've been talking in the US context, and in the West mostly. Is that a matter of convenience? Because you're based in Salt Lake City, and perhaps your partners are, and say, let's look around. There's a lot of opportunity here. Or is this the best opportunity in the world? How do we think about the US versus other markets?

Joel Edwards:

Yeah, what's that famous quote from like 150 years ago, like go westward, young man, something like that. Anyways, that's definitely true for geothermal. The rocks are hotter out West for geologic reasons. Internationally, there's quite a bit actually. Think there's a message out there that all the best locations have been tapped. Geothermal doesn't have much more. We know where everything's at. And the reality is, if you look at a map of the globe, Mercator projection, whatever projection you want.

There are actually quite a bit of geothermal known areas with hot rocks would estimate that probably a fourth to a third of the continental crust has really good conditions for geothermal systems. And we've developed a very small fraction of that. So, it's, it's a very large, in your mind, the Western half of central America, the Western half of the Western third of South America, like that Pacific ring of fire, think East Africa, think Turkey. There are a lot of areas where you have these ingredients for like hot rocks.

John Mulliken:

And would your model translate? Mean, could you imagine operating out there, or could you imagine a different offering that might be better for those locations?

Joel Edwards:

Yeah, that's the goal. Generalizability is the North Star when we build these models. We're starting with the Western United States because the Western United States is actually a really big area, and there are multiple different geologic provinces in the Western United States that host geothermal systems. So, we're capturing many different conditions for geothermal systems. And we are trying to build a generalizable model, a model that can be copy pasted to these other areas. We haven't gone out and tested these models in those areas yet. So, I can't say that we're generalizable yet, but certainly that's the goal. Like we want to help. I don't know if we'll ever be out there drilling those wells in those areas, but we would certainly love to help other folks find these systems and increase the installation capacity in these other areas.

John Mulliken:

What is the team that does all this work? Mean, who is it who's building Zanskar into this operating model across all these different aspects of geothermal?

Joel Edwards:

Yeah, the team is today roughly 90 to a hundred folks, and about a third of the team is in the field. So these are folks on drilling rigs. So, drilling site supervisors, drilling engineers, these are folks out collecting the field data. So, you know, geologic texts and so forth. That's a third of the company. The next third of the company is R and D.

These are the folks building the models. So, these are PhDs in geoscience and data science and so forth. And then the next company is the development folks. Sitting down at kitchen tables, meeting with landowners. These are folks visiting the BLM and doing all the permitting. These are folks dealing with transmission and dealing with power offtake and so forth. So, about a third.

John Mulliken:

Okay, then one more question, then I'll turn it over to Mike. Mean, one of the really fascinating things about this is you're using data science, machine learning, and what we broadly speak about as AI. And AI is driving a huge amount of power demand, which is really challenging for the grid in a number of ways, but you're also bringing a large amount of capacity online. Mean, how does that tension work in our goals with respect to the climate? And are there ways in which you're able to particularly work with some of these data center customers?

Joel Edwards:

Yeah, geothermal has a lot of the attributes that fit really well with data centers. You know, the data center folks, they don't want to increase emissions, right? They want something firm and predictable. They want something that's fairly power dense, that's not going to take up a bunch of land, right? They want something that's like domestic supply chains, domestic labor force, and domestic resources. So geothermal has that full suite of attributes.

And I mean, if you look at some of our peers, and I won't name names, but they're contracted out to a data center, you know, folks. So, I think they are going to be a big buyer of geothermal power, and the nice thing. I think there are some positive attributes here. You know, data center folks typically have a higher price point. So higher price points mean you can drill more wells. You can take more risks, right? You can go for more power, so they can pay for more geothermal power. And then you get the knock-on effects of, you're starting to scale geothermal, and then you're getting economy of scale benefits. You’re better at predicting where these sites are. You're getting more productive well fields. Your power plant you have this issue where you have only a very small niche supplier of geothermal power plants because the industry is so small. Well, you start scaling geothermal, all of a sudden, the bigger power plant providers are going to be really interested in having a geothermal product. So, you're going to start getting cheaper and more performant power plants. Like if the data center piece can be that catalyst and pay for that, like the initial growth phase of geothermal, then you get economies of scale, and then you get lower costs, and then geothermal can start to service a lot more, you know, lower cost buyers, and so forth.

Mike Toffel:

So, I have two follow-up questions on this. So, one is about relative cost to other carbon-free or near carbon-free power sources, which people think of as wind, solar, and maybe nuclear as well. How does geothermal power compare in a total cost of ownership or levelized cost, whatever metric you use?

Joel Edwards:

Yeah, we are not as cheap as solar and wind. And we are not as expensive as nuclear. We are more expensive than natural gas was five years ago, and we're maybe slightly more expensive than natural gas today. So natural gas today has doubled relative to five years ago, right? The installed costs.

So, we're in that we're in that 50 to 100 dollar per megawatt hour range, Solar wind or below that? Nuclear is above that natural gas is in that range right now I guess I guess so and when today are probably maybe on the lower end of that range because they've come up quite a bit in last few years in terms of cost And so yeah, if you look at the cost of a geothermal facility, it's roughly 30 to 50 percent well field Constructing the geothermal well field that gets you the heat so you're paying for your fuel upfront. That well-filled is your fuel for the next three to four decades. All upfront, and then the power plant is the other 50 to 70 % of the costs. And power plant, I'm lumping all that into the substation, the interconnect, the transmission, all that.

Mike Toffel:

And do you expect, you mentioned the economies of scale and so on, do you forecast this cost declining in the coming years?

Joel Edwards:

I don't forecast it going. How many examples are there out there where an industry grows, and it gets more expensive? The inverse of economies of scale. I'm curious whether there are examples where it got bigger, but it also got more expensive.

Mike Toffel:

Well, it could get bigger and not get much cheaper.

Joel Edwards:

That's true. So yeah, what's that slope going to be? That's, that's a big question. I think, in the last few years, basically what happened in last few years is that geothermal finally stopped drilling wells, the way geothermal has drilled wells in the past, which is using rotary cone and other things. And they started using oil and gas tools on geothermal rigs or geothermal well fields. And we saw a 30 to 50 % cost decline in at least the drilling phase of those wells. So, we are drilling wells much faster than we used to. The counter to that is that the steel and the cement are higher today than it was five years ago.

So maybe on the net, the well fields are a little bit cheaper than they were. The power plants are on the order of $3,000 a kilowatt. So, 3 million a megawatt, which is pretty dang expensive for a power plant. It's hard to see it. Like you scale geothermal power plants, and they get more expensive. I think there's some, there's natural gas power plants with the big expensive turbines that you have to forge an issue. You know, you don't have enough forges that can make them.

They're like somewhere I've seen numbers in the order of like two to two and a half million a megawatt today. Five years ago, they were like one to one and a half million a megawatt. Uh, so somewhere, you know, geothermal. I think it can come down. It's at three today, three million per megawatt today. It can come down maybe to two and a half to two. So, it's somewhere in there. I don't know where it's going to land. You're right. I don't know if it gets down to a million a megawatt or if it, if its asymptotes out at two or two and a half.

Mike Toffel:

So, one last question on the data centers is regarding siting. So, we talked earlier about the attributes of where to site a data center. Sorry, where to site a geo, where you guys look to site to tap into geothermal. Data centers, some of them have some flexibility as to where they are situated. So, is it more advantageous as a system for them to site adjacent to you and avoid the line loss between you and the data center, but then you incur

Some cost to link the data center back to the grid, and where is it going? Or does it just not make sense, and let data centers have their own priorities, and then you just hook up to them through long lines?

Joel Edwards:

Right. I think what you're describing is what data centers are doing in 2026. I'm from the Midwest and from Iowa. And there are several examples where I grew up where they have plotted the data center down right next to an old coal power plant or whatever. And you're like interesting. So, I would imagine that that's going to play out in geothermal, too. But I'm not making a statement that that is anyway.

But I think if you look largely, historically, at resources, geologic processes control where resources are. And that is detached from where humans choose to live, where the load is. So, in oil and gas land, we discovered huge oil and gas resources in West Texas, the Permian Basin, or in the Bakken, or whatever, in these remote areas that we then had to build a ton of infrastructure to transport those resources to the load where people live.

And it's the same thing in mining. So, in oil and gas, you build pipelines. In mining, you build railroads, right? The resource, the mine, is somewhere you build a railroad to connect it to consumption. In geothermal, we'll build transmission lines. We build a power plant where the resource is, then we transmit that resource to the load with transmission.

Mike Toffel:

Except if people want to use the heat element as opposed to the power.

Joel Edwards:

Yeah, you're not moving that very far. No, totally.

Mike Toffel:

aquacultureWhether it be agriculture or aquaculture, they need to be close to you.

Joel Edwards:

Or recreation, have you ever hopped on a plane and flown out to Iceland and then taken a dip in the Blue Lagoon, right? Like people do that. So, then they're flying to the resource.

Mike Toffel:

All right. That's true. What's next? What should we look forward to in the coming months and years ahead for Zanskar?

Joel Edwards:

Yeah, I mean, we're packing it up. We're done. We're good. We've done our job. And I'm just kidding. No, the exciting part is really, it's starting to get exciting because you've done all this work for five years. You found all these sites, and now you get to do the fun part. If you get to drill them and develop them and turn them into power, that's what's coming. So, we've only announced a few sites publicly. There are a lot of sites behind those that we found that are going to be announced over the next couple of years.

We're really excited to show the public these, get the public excited about these discoveries, and get these things in motion. One of the things we want to do is that we have the geothermal industry in the U S has not done a good job in, and speaking of Iceland, in contrast to Iceland and Iceland, when you go to a geothermal power facility, they have the blue lagoon next to it. Right. It's like, hang out at the power plant. Right. It's nice. It's fun. Or they have greenhouses or whatever.

In the U S, our geothermal power plants are like industrial sites. They're like fenced off, like don't enter. It's like not very inviting. I think we should change that, think we should make our geothermal power plants like a welcoming, like a public-facing thing. And you can do that with a hot spring. You know, you can physically, excess heat, a lightning doc could physically build a hot springs pool, right? We can build greenhouses. Like, there are things that we can do to make it much, much more public-facing. And I think.

If you look at, I don't know, some of the grassroots stuff around data center opposition today, I think it will be important for us to make these facilities welcoming to the public to get that sort of social license buy-in. And I think we as an industry need to do that here in the U.S.

Mike Toffel:

So, you're still a startup. You've gone through a couple of rounds of funding. Tell us about the funding journey.

Joel Edwards:

Funding journey. So, we're a startup venture-backed, which means you raise money in tranches, and the tranches get bigger and bigger, and the tranches are unlocked by hitting milestones that you commit to when you raise the money. So, as you hit milestones, you raise a bigger tranche of money. And so, we raised over a hundred million dollars a few months ago. And then, along with that, that was equity. So that came into the top co. Along with that, we raised a credit facility, which means that the money is not tied to the top coat; the money is tied to projects. And that's really exciting for Geothermal because typically when you're building these projects, you don't get to access credit until the well field is constructed. And remember, the well field is like up to 50 % of the total project costs. So that's a lot of money.

So, if you can pull that forward earlier into the project and pull credit earlier into the project, then you lower the total cost of capital for that project. So that's what we're really excited about this $40 million credit facility: it's pulling it forward, it's pulling it earlier into the project cycle than it was historically.

Mike Toffel:

And this credit's being extended by banks, or by whom?

Joel Edwards:

These credits being extended by sort of private equity type groups.

Mike Toffel:

So, it reflects their greater confidence in your ability to deploy assets to actually hit successes.

Joel Edwards:

Exactly. Yeah, the models are a little more auditable, right? You can look under the hood, understand what's going on. The sites look good. They have all the attributes for development that they need and so forth. So yeah, you're sort of D. This is the original goal of Zanskar, is like, can we de-risk that like upfront stage that like sucks for everyone? Like, let's make it more palatable for the capital markets.

Mike Toffel:

Great. For those who've been listening and think geothermal is the future, I hadn't ever thought about that. And wow, there are some new AI models that are helping to reduce the cost of development. Where should they go? How would they learn more to explore whether a career in this space is for them?

Joel Edwards:

Yeah, that's a good question. And my initial response is don't explore too long, just get in and do something, and see if you like it. And I don't know if that's the right answer or not. We have had folks show up to Zanskar that are like way overqualified, and they were just willing to like get in the door and just to see what it looked like under the hood. And they joined, and some of those folks have like advanced engineering degrees, and they're out in the field collecting data, right?

Because they just want to know what this thing looks like. That would be my recommendation: explore it a little bit, talk to folks on the inside, and then just get in the door and see what it's like under the hood, see if it's something that when you look under the hood, you're like, I believe this thing is going to scale. This is going to work. I'm ready to commit some time in my life to helping this thing scale.

Mike Toffel:

Great. Well, Joel, thank you so much for joining us. John, thank you so much for introducing us to Joel. It's been a really interesting conversation.

John Mulliken:

Thanks, Joe.

Joel Edwards:

Yeah, thank you, guys. It was super fun. Appreciate the time.