Valar Atomics创始人 Isaiah Taylor：核能将如何开启能源丰裕时代

Welcome to War 250. First advanced reactor to ever make power by a startup.

We did something pretty awesome today. The first ever AI chip powered by a nuclear reactor.

First trico reactor to turn on in over 50 years in the United States.

How much of the premise of Valor is possible now comes from increasing demands for power in the United States, largely driven by AI compute. Energy being a commodity, the demand is set by the price. If you can figure out how to make energy cheaper, you will have demand. When SpaceX started going, people were like, there's no way they're going to hit the numbers that they're hitting today. Most of the nuclear industry is a modeling and simulation industry. They're not focused on hardware iteration, hardware execution, building the simplest and safest reactor that allows them to scale. And Valor just demonstrated that we are.

Hi listeners, welcome back to No Priors. Today, I'm here with Isaiah Taylor, the founder and CEO of Valar Atomics. We're here in their Utah nuclear facility, where we're going to do a tour and talk about why speed and scale are the only path to making nuclear reactors in the United States again, why energy is the only input to our hyper-techno-industrial future, and why there's an open regulatory environment for progress in energy R&D for the first time in years.

Welcome, Isaiah. Thanks so much for doing this.

Of course. Welcome to Valar Atomics at San Rafael. Thanks for coming.

So, what are you doing here at Valar?

We are making nuclear reactors built for planetary scale. Nuclear fission is kind of an old technology. It's around the 1940s that we really started working on it, and we've built a lot of nuclear reactors, but we've never built them on a mass scale.

Nuclear has never had its Ford moment or its Tesla moment, if you want to put it that way. That's what Valar Atomics is working on doing. We have nuclear reactors that are more manufactured than constructed, and they're extremely safe. We think the combination of these 2 things is going to allow us to scale nuclear fission and make energy 10 times cheaper for humanity. So that's what we're working on.

We're going to dig into why you think this is possible and why you're making it happen after decades of no real activity in this space. But first, for you personally, you don't have the credentialed decades of nuclear research background that is often required to work in the space. Why you, and what triggered it?

I have been reluctant to start this company for a long time and eventually did out of sheer frustration that nobody else was working on it with the pace required and understanding the scale that would be required. I think any problem that you want to work on that's as hard as this, you have to really, really believe in it. After a certain amount of time, I just realized that I was never going to stop thinking about this, and so I might as well build it.

It started for me with my great-grandfather, who was a nuclear physicist on the Manhattan Project. My great-grandfather and great-grandmother were both in the secret city during World War II. I just grew up thinking about nuclear as this incredible technology.

I've always wanted to build thousands of machines. I didn't know which ones I was going to build. It was just very clear to me since I was about 6 that I was supposed to build thousands of machines.

At some point, I went from thinking about nuclear as a totally solved problem, which was my first outside view of the industry. It was like, okay, people know how to build reactors. They're doing fine at it. Then I suddenly realized that this was an entirely unsolved problem that we sort of thought we solved in the '60s. There was a good run into the '70s, and then we suddenly stopped.

I came to that realization, and it was very disorienting for me. I actually remember the afternoon. I was like, wait a minute, did we actually stop building reactors in the '70s? Why did nobody tell me that?

Since that moment, I had been obsessively trying to understand why. I came to some conclusions about why when I was in high school, and I still hold those conclusions today.

Then it was a 10-year journey of waiting for someone to solve the problem, watching every nuclear startup that popped up, and trying to figure out if they were going to solve the problem. Eventually, I realized that no one was going to solve it at the pace and scale that was necessary, and so I started Valar.

For people who come from outside the nuclear industry, why did we stop building reactors in this country?

We stopped for a pretty simple reason, which is Three Mile Island. We were building a lot of reactors. We were doing really well. We had a nuclear incident at the Three Mile Island reactor, and in that incident, we essentially lost the ability to cool the reactor.

In traditional light-water reactors, you have to always be able to cool them. If you shut a reactor down in traditional nuclear, you have to keep cooling it even after shutdown. In this case, the cooling system failed, and so you had core meltdown.

Nobody died. In fact, not only did nobody die, nobody was injured, and there was no radiation dose to the public. But it was sort of an optics and PR situation that was mismanaged in many different ways, and the public became fearful and lost interest in nuclear.

I think that actually turned around 15 or 20 years ago, when suddenly people looked back on that and said, okay, actually, we do know how to build reactors really safely. We've taken the learnings from that and incorporated them into new designs.

But then there was a second problem, which was that once you stop an industry, it's very hard to start it again. In the intervening time between Three Mile Island and when people gained interest in nuclear again, we also really changed how we build things in the United States.

We went from a country that was very good at large-scale civil infrastructure—bridges, roads, huge power plants, dams, highway systems. We were really good at large-scale civil infrastructure, and we kind of atrophied at that skill set. We got much better at advanced manufacturing.

One of the first realizations I had about this was that if you're going to reboot nuclear and you actually believe in the atomic energy century, which I absolutely do, it's probably not going to look like what it did in the '60s. It's not going to look like civil infrastructure because we're not that good at it anymore.

What we're way better at is producing more complicated and more functional pieces of manufactured equipment. So you're going to shift from the civil-infrastructure way of building to a manufacturing way of building.

I think some people did realize that, and this is generally what you'd call SMRs, or small modular reactors. What was missing is that people did not realize you're going to have to pursue that model through hardware iteration.

That's where Valar is unique. Yes, we believe in SMRs. We believe in reactors that are manufactured. But we also know that in order to get there, you have to do it through hardware iteration. You have to turn some plants on. You have to turn on a 100-kilowatt reactor like the one behind us. You have to go cold critical, like we did in Project Nova, and then you have to take the next steps after that.

It's going to be driven by hardware iteration, and I think that's what nobody else realized before us.

That's a whole company philosophy for you, but it has to interact with a regulatory environment where I think the assumed status quo is that you can't iterate because you can't turn anything on. It's going to take you 15 years to go through a permitting process and several billion dollars to get a no, or to try again. Why isn't that true?

There's this chicken-and-egg problem that existed for nuclear in the last 20 years. You need data in order to go to the regulator, but in order to go to the regulator with data, you have to have run a plant. So there's this question of how you get the empirical data.

The way that people tried to solve that was with modeling and simulation. If you look at the nuclear industry outside of Valar, it's mostly a modeling-and-simulation industry. When you think of a nuclear company, there are nuclear companies out there that everyone knows the name of. When you look under the covers, it's actually a modeling-and-simulation company.

They produce very, very precise what we call paper reactors, which have really good predictions of how a theoretical thing might behave. That was sort of the hack to try to get around this regulatory chicken-and-egg problem.

But the alternative is that you actually just do what Congress originally intended. If you look back at the early days of nuclear legislation, there were actually 2 pathways for nuclear in the United States: the testing pathway and the commercial pathway.

The commercial pathway everyone knows is called the NRC, the Nuclear Regulatory Commission. The NRC is solving for large-scale commercial deployment of mature systems. So if you try to do the iteration under the regulator that's concerned with large-scale commercial deployment of mature systems, that's going to be tough.

You're going to be asked commercial-deployment questions when you're just trying to figure out how to bend some steel and get some atoms splitting. But there's another pathway, and that pathway is the Department of Energy.

Most people don't know this, but the Department of Energy was originally called ERDA, the Energy Research and Development Administration. People don't know that because it was only called ERDA for about 2 years. There was a little 2-year period where it was called ERDA, and it was renamed shortly after, but its origin is actually a spinout of the Atomic Energy Commission.

Where the DOE comes from originally is Congress taking the Atomic Energy Commission, which was 1 agency responsible for both commercial deployment and testing, and saying, “Actually, let's split these out. Let's make the NRC in charge of commercial deployment. Let's make ERDA in charge of testing, and we'll have these as 2 separate agencies.” Then, 2 years later, Congress said, “Actually, we're just going to call ERDA the Department of Energy, right? The DOE.”

But the origin of this agency is testing nuclear reactors. That's what it's for. Everyone forgot this. It's just been sitting there in legislation for about 40 years, and no one's really taken advantage of it, except for 1 government program here or there where they want to do a NASA space reactor or something like that.

But from a legal perspective, it has been waiting for an administration to say, “Hey, this might be a way to accelerate nuclear R&D.” That's exactly what the Trump administration has done. The reactor that's behind us here was turned on under an executive order. It was EO 14301, which called for 3 advanced reactors to go critical on American soil by July 4. We built it under Department of Energy authority under that executive order.

As we're sitting here, we're making 100 kW, splitting something like 10^17 atoms per second in this reactor behind us. So, this is the control room. There are 2 sections. This is the control half. This door's closed. We have to have the senior reactor operator's permission to enter here, which we just got. Thank you, Brent. Out here is sort of our observation room.

It looks like the control room in Hawthorne.

It is the exact control room in Hawthorne. We literally picked this up, put it on an airplane, and flew it here. The entirety of the plant, including the control room, flew with us on the C-17s. It was the first time that's ever been done, which is pretty cool.

How many people operate it?

We've got 1 operator at the controls, who's the reactor operator controlling all this. Then you've got a senior reactor operator who provides oversight and ensures that we're staying within our safety limits and everything. But we always have to have 1 person responsible for controlling and maintaining the reactor safely.

So, in a couple of days here, we're going to do a scram. We're going to press that big red button right there. The big red button scrams the reactor. It basically drops the rods, right? Dropping the control rods will introduce a lot of boron carbide into the core. When we scram it, the rods will drop, and boron carbide will enter the core. Boron is a strong neutron absorber, so there won't be enough neutrons to maintain criticality because the neutron population gets absorbed into the rods. That'll cause the reaction to stop.

Now, in a nuclear reactor, this gets to the core of what Valar is doing differently. Why do we really believe that the technology we have here is going to scale? Traditionally, when you scram the reactor in a light-water reactor, your work is nowhere near done yet. The next day of operations is incredibly important because when you turn the reactor off in a traditional nuclear reactor, you still have a lot of heat being produced. You have a lot of heat being produced because in a nuclear core, you still have about 5% heat production from recently split atoms.

When a reactor is running, you have a bunch of atoms being split. They're producing a ton of energy, and even after the reaction shuts down, you have recently split atoms that are continuing down their decay chains. This is called decay heat. Decay heat is about 5% to 6% of what your continuous power was before you shut down.

In a traditional reactor, the reactor would melt down if you didn't continue to run the cooling pumps after turning the reactor off. There's no more fission occurring. You don't have a chain reaction anymore, but you do have this decay heat. If you were just to let it sit there, that heat would build up and you would have a meltdown. That's exactly what happened in Fukushima and Three Mile Island.

The way they deal with this in traditional plants is they have very carefully built cooling systems that they continue to run for about 24 hours after shutdown. They try to build in all these redundancies and make sure that, in any circumstance, the cooling systems never fail. That works pretty well, and nuclear is the safest form of energy ever made.

Even with that caveat, Three Mile Island melted down and nobody died, right? Fukushima, arguably, maybe 1 person died. It's sort of a debate among scientists whether that person died from radiation or not. So, even though traditional nuclear is very safe, we want to do even better.

We want to do better because we are interested in scale. We want to make thousands of reactors, tens of thousands, hundreds of thousands. When you're doing that, you really just want it to never melt down for any reason, right? The best way to do that is to make active cooling systems unnecessary altogether.

That's what we're going to demonstrate on Friday. We're going to scram this reactor. Immediately after we scram, we're going to turn off the entire electrical supply to the plant. We're going to turn off the circulator. We're going to turn off the RCCS pump. Every safety system in this plant, we're going to shut off, and we're going to watch what happens.

We have a cheat code. We know what's going to happen because we did this in Hawthorne before we had neutrons in it. We ran the plant at full power with electrical simulators. We basically have these electrical resistors in the plant in Hawthorne. We put about 15 city blocks' worth of LA power into this and got it up to full nuclear temperatures and pressures.

Right. I remember seeing the sim. Yeah.

We held it there, and then we did exactly what we're about to do, which is turn off all safety systems at once and watch the plant for 2 days. Exactly what we had hoped happened: The RCCS panels, which are these water jackets around the core, went into passive circulation mode. That means the water boils, the steam leaves, it condenses, that removes heat, and you have natural circulation with no moving parts.

So, there is zero input from the operators, no moving parts, and no electrical control. It's just the geometry of that system that slowly removes the decay heat over a period of 2 days. We're about to demonstrate that right here, starting on Friday, after 72 hours at full power.

That'll be a really exciting demonstration, and it's really the basis of scale, right? If we're going to do this many times, we want to make sure that, just from the basic physics, this reactor doesn't melt down. Not because of operator control, not because of really good engineering, but because the physics of the plant make it safe from meltdown.

One of the huge reasons for the slowdown in investment in real R&D in this space has been concerns about the safety profile. You and your team are very respectful of that, but not particularly concerned that this is an insurmountable problem. What do people outside nuclear misunderstand about the safety of these SMRs and your design in particular?

There are 2 big misunderstandings. One is that people generally perceive nuclear as stable, cheap baseload, but kind of dangerous with downsides. I think that's just a complete misperception.

Nuclear energy is the safest form of energy empirically. If you look at the amount of power generated versus the number of deaths, it's even safer than solar, which is very strange because how does someone die from solar? Most solar is installed on roofs, and people sometimes die falling off a roof while installing solar. If you add those deaths up, it is actually a higher death per unit of energy than the totality of nuclear energy in the world. That's kind of crazy. Nuclear, as it is today, is already the safest form of energy.

The only problem with existing nuclear is that I think the existing philosophy—again, it's the safest in the world, but we can do better—says that if you look at risk, it's actually composed of 2 different things. It's composed of the odds of the thing happening and the consequence if it does happen.

Traditional nuclear has focused on risk reduction by reducing the odds. They say, “Okay, a meltdown could have some bad consequence, so let's make sure that it never, ever happens.” All of the effort into risk reduction goes into making the odds of anything ever happening very low.

The alternative way that you can reduce risk is by reducing the consequence, and we would argue that that's a much better way to reduce risk. Odds are stochastic, right? Even if you do an enormous amount of risk reduction on the odds and all this engineering, something's still going to happen. Something's going to come out of left field, and you can't predict everything.

And so advanced reactors tend to focus on consequence more than odds. Now, we still try to protect against things like—we have site security here, for example, so someone can't come and compromise the control room. But from an engineering perspective, we also say, well, what if someone did compromise the control room? How do we still make sure that it's perfectly safe?

And so advanced reactors really focus on that second thing. Our safety basis when we go to the regulator is that everything in the plant has failed. Absolutely everything, right. Everything.

We know what happens.

Yeah. In fact, it is safe, right? So we have regulatory limits where we say, okay, in an accident scenario, you can't dose the public with radiation, right? That's really ultimately what a nuclear regulator is trying to prevent. You can't dose the public and workers with radiation.

And so our safety basis says, "All right, let's look at the reactor design and pretend that everything in it has failed. Are we dosing the workers and the public with radiation?" And the answer is no. That's super important because even though we do still try to reduce the odds—we have site security, we make sure that our operators are well trained—we start with the worst-case possible scenario, where everything has failed, and we say, "How do we make that fail-safe?"

And that's really a physics problem. It's really about the geometry of the core and the materials that you use. TRISO is a huge part of that, and that's how we pursue safety. So I think that people miss that, A, nuclear is already very safe, but B, the new breed of nuclear is much more focused on just making sure that these plants are intrinsically and inherently safe, regardless of the odds, regardless of what crazy thing comes out of left field. And that's a much better way to scale.

What are the major causes of reliability issues for a reactor?

Of this design? Let's see. The helium circulator is one area that people have had trouble with in the past. The heat exchanger between the main helium loop and the secondary power-conversion loop is something that Fort St. Vrain has had issues with. I don't know, Jess, what do you think?

Jess

Moisture as well. Fort St. Vrain had problems with moisture. Graphite is hydrophilic, meaning that it loves to take moisture. Even in hot, dry desert air, it still sucks up moisture.

So we have mitigated that risk by having the helium purification system. Part of that is taking the moisture and trying to condense that moisture out of the system and then put it into some supply tanks—or, excuse me, some storage tanks—outside of that. We know it's a known fact that for several hundred hours of operation, we'll be slowly leaching that moisture out of the graphite blocks as we heat them to maintain the temperature.

So, as you know, it's graphite ceramic on the block on the way in. It's kind of difficult to migrate that moisture from the center of the block all the way out. That's why it takes so long: just keeping it up to temperature, and it slowly migrates its way out of the blocks. But if you don't capture that moisture and remove it from the system, then you can imagine any carbon steel that is in the system will rust. Then the rust particles will migrate and either cause a short in the motor or a blockage in the system somewhere.

So anyway, that's one of the lessons learned from Fort St. Vrain: you need this system. I will say, though, helium-graphite systems have a lot of natural advantages. There's basically no chemistry going on other than the bit of moisture that we need to pull out of the graphite blocks.

Helium is inert, right? And unlike a water-based reactor, where you're very carefully trying to make sure you have dry steam—if it's a dry-steam plant—and wet steam is going to start just destroying all of your thermal machinery and everything. Helium is a very inert, simple chemical to work as a working fluid.

Now, there are downsides to it. It's not very dense. You have to have more pumping power, these sorts of things. But the fundamentals are in our favor in terms of building a very simple plant.

When we think about an abundance of nuclear power, I think a lot of interested listeners will have heard of a number of SMR startups and people building larger-scale reactors. How do you break out the different use cases and the different designs here? Are people aiming at different power levels and efficiencies? Is it just a completely different philosophy of company? How do you look at that landscape?

Yeah, I would make one big division in nuclear companies, and it is people who believe that nuclear is essentially a hardware-execution problem and people who believe that it's a design problem. Valar is solidly in the camp that we believe this is a hardware and execution problem. This is really an issue of: can you build reactors, can you turn them on, can you get them running, can you operate them well, can you produce them at scale? Can you go from making 1 to 10 to 100 to 1,000? That is the problem of nuclear.

The second class of nuclear companies, I think, are mistaken to think that this is a problem of making the most beautiful design, making the most sophisticated design. The design will have the most perfect efficiency. You look inside some of these design documents, and there's so much complexity. The materials that go into it are super rare, and you have to set up entire supply chains that don't exist just to get the thing running.

And I look at that and I say, okay, maybe over time the reactor grows in complexity, and you start to use some more rare materials to try to get higher and higher performance. But the problem of nuclear today is the Toyota Camry problem, right? We don't want to make Lamborghinis. We want to make a very simple, very cheap, very safe reactor that we can make literally tens of thousands of. And actually, that is going to make the cheapest energy in the world, right?

You maybe will get a little bit better performance out of a more complicated reactor, but I'm going to beat you on cost because I'm making 1,000 of them, right? And that's really what we care about. Valar is in the business of making energy 10 times cheaper for humanity, and we want to continue doing that forever. There will always be ways to make energy cheaper and cheaper. That's very much our goal.

And I think the big difference in philosophy is: do you tackle that through building reactors and through iteration, or do you tackle that through design? We believe it's entirely a hardware problem.

So the company is less than 3 years old, just about to turn 3. Where in the timeline of hundreds of reactors are we?

Yeah, we just turned on number 1. So we went critical back in November. That was our first criticality. That was what you'd call a cold criticality, or critical pile. And just a few days ago, we made power for the first time in our first reactor. So that's the W-250 reactor. Obviously, the next goal is to go turn on another one.

But there's a sort of success metric here that we talk about in the company, and we call it tick rate. Tick rate is how we judge ourselves and how we judge other people in the nuclear field. Tick rate comes from video games, right? And it's sort of the time in between changes in the game state, right? For us, what this means is how long from the founding of our company to the first time we split an atom, to the second time, to the third time, to the fourth time. The goal of the company is to get that tick rate as low as possible, right?

Our first-ever mark, from filing in Delaware to first atom split, was 2 years and 4 months. From Project Nova's first atom split to the second time here in this reactor was about 7 months, and the goal is to get that window as short as possible. We will get it down to minutes. This company will get to the point where we have a new reactor turning on every few minutes.

And when we do that, if you look at the economics of nuclear, it really is driven by how quickly and cheaply you can produce plants, right? The uranium is very cheap, right? The fuel is very cheap. The question is, can you produce plants quickly and cheaply? And that is going to be driven by iteration and by scale, which ultimately is driven by tick rate.

So I'm very proud of the team. We've done some enormous, incredible things that no other nuclear startup has been able to do yet, but we are nowhere near done. We're very hungry, and we're going to make that tick rate 6 months, then 4, then 1, and then into the minutes.

Welcome to W-250. Where you are standing right now is the first advanced reactor to ever make power by a startup. The first advanced reactor ever built outside of the national laboratory. This is the 5th new nuclear device to make power in the United States since the year 2000. Valar is the only private company founded since the discovery of nuclear fission that has ever made nuclear power.

This is really the thing that is unique about what we're doing: we're a new team, we're a new company, and we're attacking nuclear fission from first principles. All the other companies in the space who've actually made power before are 100- or 150-year-old utilities, defense contractors, and large engineering corporations, and they haven't really done fundamental R&D. What they've done is they've picked up government work, commercialized it a little bit, and done some supply-chain integration.

But this is like a startup mindset applied to nuclear fission. There are a lot of things about this plant that are different that you'll be able to see. The first one is what we're looking at right here. This big, awesome structure is called the Modular Citadel.

I'm going to talk to you for a few minutes about concrete. It's going to sound like I'm a big concrete nerd, and I guess that's okay. But this is actually some of the more innovative stuff that we've done in this reactor. The purpose of this concrete is a bioshield. The reason we're able to stand here and not be receiving unhealthy doses of radiation right now is because there are 78 inches of concrete in front of this reactor.

If this concrete were not here, we would not be able to stand here. But it's actually very special concrete. The Modular Citadel is a series of precast blocks. Each of these blocks was actually created in a factory. It's our factory.

We have what we call the Citadel factory in Salt Lake City. We have a production line where we could create 3,000 blocks a year that are identical to these, and they arrive here on a truck and then get placed by this crane up here. That's why we have this crane. It can place these blocks.

What's really awesome about this is people have thought about doing this before in nuclear. They thought about precast blocks, but the problem is, if you're going to put 2 concrete blocks together, if you were to zoom in on a microscopic level, those blocks are actually not touching, right? Nothing's actually touching on a microscopic level. You have a little gap between them.

Gamma rays can slip between the gaps, and neutrons can slip between the gaps. What you can't see here—and I'll show you some photos afterward—is that if you were to look with an X-ray in that seam right there, you would actually see that the concrete itself follows what we call a tortuous path. So it's a sine wave.

The concrete on every seam follows a sine wave curve, and on the vertical edges as well. If you were looking at that vertical seam, you would also see a sine wave. If you were to look in the corners, you would see a sine wave. There is no straight path from the inside of the reactor to the outside through the Modular Citadel for a ray to follow straight through the concrete.

What's even more interesting is that traditionally, when people have thought of this, it's really hard to get that casting shape right and to get it to actually be solid enough to bear all of this weight. The way that they get around this is they would grout it. They would grout it with more concrete. You'd have sort of a wet concrete mix that you use to seal them together.

This has no grout. These blocks are literally just stacked. In fact, there's no mechanical lock between them either. There are no bolts. There are no screws.

We do have this frame around it. This is called a seismic frame. We did this—we actually don't need it according to our engineering calculations. The Utah State Building Code had a disagreement about whether or not we needed a seismic frame, and you're like, “Oh, let's just put it on just to be safe.”

But the entire Citadel has no fastening of any kind in the blocks. What that means is that we went from 3 months—which is what it would normally take to build a bioshield like this—to stacking this in about 42 hours.

How do you explain the pace of what you're capable of doing here versus other people that are trying and have been trying?

Yeah, the pace is extraordinary. Of course, I'm proud of my team. Of course, I think Valar is the best nuclear company out there, but I would argue empirically that we have the highest pace of any nuclear company in the world, and it's not close.

The reason for that is 2 things. One is that we are obsessed with simplicity. We want our designs to be as simple as possible. We would rather have a simpler machine that's easier to build, that we can build at higher scale, than a machine that's more efficient, that's more performant.

We have ruthlessly deleted complexity, deleted parts, and deleted systems where they're not necessary. The other thing is that we picked a super-safe architecture, and this is, I think, a point that's hard to tell from the outside because every reactor looks the same from the outside and every CEO says that the reactor is safe.

To some extent, they all are, right? Because they're highly regulated and we make sure that they're safe. But there are levels to this, right? It goes back to the previous conversation we had about odds versus consequence.

A TRISO-fueled, graphite-moderated, helium-cooled reactor of the geometry that we've built here is just about as safe as humanity knows how to build a nuclear reactor. That actually allows us to move faster, right? That's going to allow us to scale as well.

The lower that we can keep that consequence, the greater we're going to be able to scale and the faster we're going to be able to move. The last thing is just the way that we have shaped this team. I have built this team very carefully.

I'm deeply involved in every hiring decision, and we have filtered nuclear as an industry for people with extreme agency. People with extreme bias to action. We have a lot of people from outside nuclear, right?

A lot of the people here now have more nuclear experience than any other startup because they've actually turned a plant on, but when they walked in, they had never seen a nuclear reactor before. What they had done is built hard things in the real world, right? So we try to find those people who built hard things.

Within nuclear, I particularly love to find people who started in nuclear because they loved it and realized it was really slow, and then they left and were like, “Screw this. I'm going to go do something interesting.” Then they built a ton of hardware somewhere else, and I convinced them to come back because I know that they love it.

I'm like, “If you come back here, you're going to be able to actually build and actually execute, but in something that I know you love.” Also, there are lots of people from the traditional nuclear world, but those people we filter very hard for.

Do you want to do science? Do you want to write papers, or do you want to build reactors and turn them on? The entire Valar team is laser-focused on that.

How much of the premise of Valar being possible now comes from the fact that there are increasing demands for power in the United States, largely driven by AI compute?

Yeah. I'm super grateful for all our partners in the AI space. NVIDIA, of course—we've got a big event that we're talking about today. We'll talk about that more in a minute, what we did here with NVIDIA.

But I would like to point out that, with energy being a commodity, the demand is set by the price, right? If you can reduce the price, you increase the demand. That will be true about energy forever, right?

From now until the end of time, we know that energy is the only scarce thing in the universe, right? It's the only irreversible thing in the universe in terms of a resource. If you can figure out how to make energy cheaper, you will have demand.

I view Valar as having a fundamentally infinite market because if we can figure out how to make energy at 1 cent, you induce your own demand, right? We begin to imagine ways to do things better with cheaper energy. If you can make it a tenth of a cent, you've induced even more.

Yes, it's awesome that we have these tailwinds of customers who are really hungry for power. I think it's also great from a public-awareness perspective, where suddenly everyone remembers that power is super important, right?

But I also know that fundamentally making energy cheaper is what induces demand for more energy, and that's really going to be our strategy for the next century.

Can you talk about what you are doing with NVIDIA?

Yes. We did something pretty awesome today. We powered the first-ever AI chip powered by a nuclear reactor. We took an NVIDIA Blackwell. I'm super grateful to them for providing the system to us, and we connected it directly to our nuclear reactor.

We actually hosted the nuclear website. Depending on when this podcast drops, you'll be able to see it. If you go to nuclearbite.com, that website is directly hosted on the nuclear reactor. It is entirely coming from that chip provided by NVIDIA. We've always been asked if we're going to sell merch, and I've always said that we're not going to sell any merch until we've actually split an atom. So actually, the nuclear website is the only place you can buy Valar merch. You actually have to buy it hosted from the reactor. So we take billing in the physical world very seriously.

You consume some nuclear power. Nice.

Actually, on the website, if you go on it right now, we tell you exactly how many atoms of uranium were split in order to deliver that web page to you.

The other thing is, if you're watching this podcast too late, you're not going to be able to see it anymore because once the reactor turns off, the website is directly hosted from there. It won't be available anymore. So there's some exclusivity to it as well.

So, you've got to tune in while it's live.

That's right. You have to tune in while the reactor is running. We'll keep it running for a few days as we continue our final testing, and we'll go through various tests over the next few months.

When you go talk to the big buyers of compute and thus power today, they will say nobody can deliver at scale until around 2031 or 2032. Until then, it's going to be diesel gensets and maybe solar and batteries, and we've got to figure out how to make it until that period of time.

What do you believe that they don't believe about the next 5 years?

Yeah, I think they've looked at all of these nuclear companies of the past, and again, most of the nuclear industry is a modeling and simulation industry. That's really what it is. If you think about it, companies are what they do, right? I actually did not allow Valar Atomics to call ourselves a nuclear startup until we had split an atom, right? We are legitimately a nuclear company because we split atoms.

We didn't allow ourselves to call ourselves a nuclear startup until we'd split the first atom. Before that, we were a company that was planning to split an atom. Now we're a company that has done it—about 10²⁰ atoms so far. That number will keep going up. But I think there's just a big gap in mindset: Are you serious about scale?

I would also point out that exponential curves are very counterintuitive, right? If you look at the predictions of how many satellites would be in orbit when SpaceX started going, people were like, "There's no way they're going to hit the numbers that they're hitting today." Exponentials are just really, really hard for humans to understand.

So I would say that if you're looking at the nuclear space today, you don't expect people to have this problem solved in 2031. In fact, I would say even in 2035, a lot of these companies are still not going to get there because they have the wrong mindset. They're not focused on hardware iteration, hardware execution, and building the simplest and safest reactor that allows them to scale.

Valar just demonstrated that we are. We actually had to invent our own grade of concrete in order to do this. The reason is that we need a high enough density to block gamma rays. The concrete also needs to be very strong because these are self-stacking blocks, and we can't use rebar because if we use rebar, it becomes activated by neutrons. You'll actually be producing nuclear waste.

We took 2 engineers on our team—1 mechanical engineer and 1 nuclear physicist—and had them tag-team this. We said, "Go invent a grade of concrete that has the density to block gamma rays, the strength to actually be a concrete block that's self-cast, no rebar, and the right atomic makeup so that it will not become nuclear waste when it's irradiated."

They literally flew all around the country, collecting rock samples from all around the United States. We called it the rock hunt. It was about a 3-week period of flying from place to place, literally collecting bags of rocks. They would come back on a Friday night, set up the conference room with a bunch of buckets of acid, dissolve the rocks, and then take them through spectroscopy analysis to figure out the exact atomic makeup. Then they'd go back out the next week and collect the next sample.

By the way, the 2 team members are 23 years old and 21 years old, respectively, and they've done things that the nuclear industry has been dreaming of for literally 30 years. When we unveiled the system for the first time, we got DMs from people who had been in nuclear literally their entire careers saying, "I've been writing about how someone should do something like this since the 1980s, and I've never seen anyone be able to actually pull it off."

That's a really awesome thing about working at Valar. That's also why we get the best talent in the world: If you are a really, really talented person who cares about nuclear, you want to see things actually get built, and you need to go where you can actually build things.

You're not a paper-and-simulation company; you're a nuclear company. But you also seem to be a road-building company, a building-building company, a concrete-pouring company, and a rock-finding company. How do you think about having those capabilities, and why do you make that decision? You're very opinionated that if we must, for speed, we will verticalize.

Yeah, this is the secret weapon of Valar, and I'm going to give you the secret weapon because I fully believe that no one will be able to copy us in this. We are willing to verticalize anything that is necessary on our path to scale. We've proven that a couple of times in this reactor so far, in areas where we were told repeatedly that it was utterly impossible.

The fact is, when a problem is important enough, you can go solve it and you can verticalize it. When a problem is as important as making energy cheap, you can get the most talented people in the world to come and help you solve that problem, right? You can assemble the brightest minds to tackle it. You can get the best capital formation to help you acquire the resources necessary. It's really a mindset thing.

We've talked to and know many of the most talented nuclear engineers in the world. What we keep finding over and over is that, yes, there are amazing designs, but not much has actually been physically built. If you have the muscle as a company to build things in the physical world—if you know how to do fundamental clean-sheet design all the way through to prototyping, testing, and manufacturing—you have an edge that nobody else has in nuclear today.

We look at really anything in the stack. It's not that we want to verticalize everything. We'd prefer that most of the plants commoditize. In theory, it would be great if we could just buy a kit, right? If anyone is selling a nuclear reactor kit, I'm a buyer. I'm going to staple it together, make energy, and that's great.

You're just going to make lots of plants.

Exactly. But that's not how it works. In actuality, there are a lot of these components that are wildly overpriced and have super-constrained supplies. A fundamental edge that Valar has is that we know we are competent to verticalize any individual piece that prevents scale. Once we identify that, we attack it ruthlessly.

And you don't think any piece of it is intersecting with more regulatory pain or too complicated for you to figure out? When I think about your fuel supply chain—

It's definitely intersecting with regulatory pain, and we run at that pain. We just run toward gunfire on the most complicated things every time. We're 2½ years, almost 3 years, into this company, and we've done something that no other nuclear startup has ever done, which is make nuclear power.

It comes from this fundamental mindset that the team has: Yes, there are complexities and regulation. Yes, this is hard equipment to manufacture. Yes, it's difficult to do both of those at the same time, where you have complex equipment, regulation, construction, rock hunting, pouring concrete, and building buildings. But the company that figures out how to do all of that complexity at scale and at pace is going to win.

We're sitting here right now less than 3 years from the founding of the company. If you look back at this in 10 years from now and Valar's worth $500 billion, something like that, you will be able to see the throughline: We took the hardest parts of nuclear, which are the site, the regulatory process, the fuel supply chain, the instrumentation, the shielding, and all these things that everyone tries to outsource, and we said, "Actually, we're going to become masters of that." We're going to hire the best people in the world to tackle it, and that becomes our edge.

And in the actual manufacturing, this is another really good example of what makes Valar unique. This is our control skid. These 3 boxes right there are basically analog-to-digital electronics boxes. If you know anything about electronic circuitry, it's pretty simple stuff. You've got some signal amplifiers and some basic rectifiers in there, and basically we take some analog signals from our detectors and send them to the control room.

Each one of those boxes costs $450,000. This is when anyone talks to you about cost in nuclear and says, "Oh, nuclear is expensive." They're talking about ridiculous things like this, right? This is really where Valar's going to win. We accepted the $450,000 cost on this one because we're moving really fast, and spending $1.5 million for these 3 boxes was more worthwhile to us than taking a couple of months of delay.

But this box over here is a totally different story. This one was bid out at $5 million. They told us it was going to cost $5 million to buy that box right there. That is what's called our reactor protection system, or RPS. This is basically the brain of the reactor.

This system right here is super unique. It has no human input. It's basically a self-deterministic machine that decides whether or not the plant is in a safe state. It has 3 different sections that vote against each other to determine whether the plant is in a safe state, and if 2 of them vote that the plant is not in a safe state, it shuts down.

We were quoted about $5 million, and we were told it would take about 2½ years to build. We said, "Well, all right. Maybe I'd be willing to pay $5 million, but I'm certainly not going to wait 2½ years." After haggling with this vendor for about 2 months and trying to convince him to go faster, we eventually got the team together and said, "Guys, we're going to have to build our own RPS."

We sat down with Joe, who runs instrumentation and control.

Awesome guy, Brown dropout. Really good with electronics, he brought together 5 team members, and we locked ourselves in the conference room. 6 weeks later, we had a working RPS, and we spent about $400,000 on it.

Everywhere in nuclear is like this. There are totally fake costs from an industry that is totally anemic and doesn't know how to build anything anymore.

By the way, the conclusion to that story is kind of unfortunate. When the vendor realized that we were going to do it ourselves because we just could not get our heads around a 2-year timeline, they freaked out and started telling everybody in the industry that we were an unsafe company that was going to kill people. We would dare to do something ourselves that was their special sauce, and they were charging the market $5 million for it.

I don't know if you guys use any of these investor search platforms where you can look up a company and people have done expert calls. They contacted all of those expert-call companies and volunteered to do an expert call on Valar, where their CTO

Shitty.

Yeah. Their CTO basically went and ranted about how we're the worst company on earth because this is a massive threat to them: the fact that their $5 million system we can just make for a couple hundred grand with a team of 5 in 6 weeks.

The more that you look at nuclear, the more it's like this everywhere you look. It's a fake industry that has not been building anything for 40 years, and the little bit that they have been building, they're charging a 100× margin on. But if you just take it from first principles, have smart people in the room, and are willing to build it yourself, you end up with this, and the plant is running. There is something very inspiring about the idea of, like, given the physics work. Yeah.

It should be possible, and if it is possible to do it faster here, the best people will come. Yes. And the capital will come.

Exactly.

This is an expensive project, as far as I can tell. How do you think about the fact that you're an equity-financed, venture-backed company today? How do you think about the fact that debt vehicles, project financing, and operational proof of this don't really exist in the ecosystem to perhaps support as much scale as you'd like, as quickly as you'd like?

Yeah. So this is actually one of the things that I noticed years before starting the company. The traditional plan for a nuclear startup is you try to do some engineering and some design and some customer LOIs, and you try to assemble this package. It's essentially paper, but it's a package that's attractive enough to get somebody to come and fund the project. That's the essential plan of all nuclear startups: they want to convince a debt financier or project financier to fund their nice paper package of designs and partnerships.

I went into this company believing that, and by the way, I believed that was possible 10 years ago. After watching this fail over and over and over in other startups, I came into this company knowing we're not going to do that.

One of the unique advantages that the United States has is a risk-on equity capital environment. The type of risk that we are asking investors to take is: we know the physics works, we know that there's infinite demand, and how we go from here to there is technology execution. Venture capital in the United States is the best at underwriting tech risk of anywhere in the world. That is how Silicon Valley has come to be what it is. That is how venture capital has come to eat as much of finance as it has: venture capitalists know how to underwrite technical risk.

It's execution. When we say tech, we're not talking about physics. We're not talking about science. We're talking about mechanical engineering, thermal hydraulics, instrumentation, and manufacturing methods. These are complicated. It's nuclear, but it's significantly less complicated than a rocket engine, for example.

I think this is one of our fundamental edges and advantages: we're risk-on about this, and we have a cap table that's risk-on about this. We are willing to go build reactors on our equity balance sheet because that's going to allow us to prove it years ahead of anybody else. While our competitors are still trying to convince fundamentally risk-averse financiers to finance a project, we will be on our 5th reactor.

After 5 reactors, project finance is an option. Debt is an option, but it's only an option once those early backers have put the capital in the ground and demonstrated it. That becomes an enormous moat that's frankly going to be impossible to beat.

One of the more different things that you've decided to do here is try to build your own plant independently and bring customers to it at as much scale as possible in a gigasite. Part of the belief there is, well, if I can provide power at scale, and especially cheap power, the load will come to me for many things. Can you describe the rationale behind that strategy?

Yeah. So it comes from a very similar argument that I was making before, where I'd watched these startups try to get that deal right for a very long time. Frankly, there's just too many parties. There are too many people and too many parties.

That slows your speed. It's like, well, let's negotiate about this site and this permit.

Exactly. In theory, could we go to a hyperscaler who wants a gigawatt of power in a place? Yes, and certainly we will at some point. But we could also just put a gigawatt on the ground, and we can do that on our own timing, at our own pace. That pace is going to be faster than anyone else in the world.

Then the question flips around: if I have a gigawatt of power with land and fiber, is someone going to put a data center there? Yes, absolutely.

In the long term, we'll maybe start to do some of these deals where a dedicated customer wants their own gigasite for their hypercluster or something like that. But it'll be driven by speed even in that case. We'll go to the customers who say, “We want it delivered in a year,” not the customers who say, “We're thinking about this in 3 to 4 years.”

So really, speed and scale. If you honestly ask me any question and you keep digging back, you will find speed and scale. That really is what nuclear is missing. It's missing its Ford moment, its Toyota moment, its SpaceX moment, where you have a company that realizes that the actual thing nuclear is missing is scale deployment. That's really what we care about.

With the gigasite, we've got a site here. We're going to build more reactors here. We have a planned gigasite nearby. We get to go with conviction and build power, and I'm absolutely confident that load will follow.

How do you, if you just think of your primary objectives as speed and scale—

Yes.

Given safety, how do you drive that personally as CEO? What do you wake up every day and think about in terms of problems to solve or what to focus on?

Yeah, this is one of those things that I think is probably the most irreplaceable attribute of a company: its internal pace. We call it tick rate.

It's a metric that we've designed around going critical in new systems, but you could probably translate that to any industry. In software, it's probably something like new continuous-integration builds per hour.

I think it's the hardest thing to replicate. You have to start it from the first day of the company. It has to come from the CEO, and the first 5 team members have to have it deep in their bones. The first 20 team members have to have it deep in their bones. They have to be trained so that when they hire people, they're looking for that in those people's bones.

That's also not enough. Once you're in that setup, I have to constantly—and this is really what I do—go around the organization and speed things up.

My chief of staff is laughing over here because she's seen this happen many times. I go war mode on things. I look around and say, “Okay, that's going slow.” We'll set up a physical war room and say, “How do we turn a 6-month timeline into a 4-week timeline?” Maybe we'll hit 8 weeks, but it's not 6 months.

I've realized there is no autopilot for that. There's no automatic mode that makes a company faster inherently. There are lots of things you can do to set it up for success, but—

But you have to push.

I will always have to push, and I will always push. It's because I know that the physics are here. I know that energy will be 1,000 times cheaper than it is today. I know that nuclear fission will make energy 1,000 times cheaper than it is today. Somebody will do that, and it's really up to us as a team whether or not that's going to be in the future decades or the future centuries.

That time is up to us. It's up to us how quickly we go down that path, how quickly we make energy more abundant, and that feeds into everything else. It's highly motivating for us, and we are tackling it as fast as we can.

For people in their daily lives, what happens for them if this works and you get very cheap, abundant nuclear energy? Imagine that future for us.

Energy is the fundamental input to the quality of human life. If you look backward into time and ask how humanity went from one standard of living to the next, it's always unlocked by cheaper energy, right? So when we went from essentially a solar-powered world, which is what we were in the agrarian age, literally human muscles were solar-powered because you were either eating plants or eating animals, and both of those things were essentially chemical energy built up by the energy of the sun through photosynthesis. Life was hard, right? Life was really, really hard.

When we got into underground hydrocarbons, you had this massive explosion of quality of life, right? We have heat in the winter and AC in the summer. We're able to drive cars to get to the hospital. Even really simple things—we're in an aluminum-strut building right now. Every one of these struts that you can see behind us is made of aluminum.

There was a time when aluminum was considered a precious metal, when kings would literally make jewelry out of aluminum because it was that rare. The thing that changed it to a structural metal that we make buildings out of is energy becoming cheaper, right? Charles Martin Hall here in the United States figured out aluminum electrolysis, and cheap enough energy means that you can actually just separate that oxide from the aluminum and have pure aluminum metal. That is entirely downstream of energy.

I would say that the standard of living we enjoy today is entirely downstream of the fact that we figured out how to make pretty cheap energy with hydrocarbons. Now, our goal is not pretty cheap energy. It's insanely cheap energy, and that's going to be unlocked by nuclear.

What it will look like, I think, is fundamentally unpredictable, but I have a few ideas. First of all, I'm very, very excited about transportation. If you think about the fundamental limits of why I can't go see Grandma every week on a plane—why is it just so hard to do? Why is it so expensive? Why is it so complicated? It really just comes down to the fact that we've kind of squeezed as much juice out of jet fuel as we can, right? It's just hard to imagine jet fuel getting 10 times cheaper than it is today, and it's polluting the environment and all of these different things.

If you could make energy 10 times cheaper, then you probably could go visit Grandma every other day. Just on a Saturday, it takes 30 minutes, and she lives across the country, but you go and visit her on the weekend.

I'm very excited about what I would call hyper-techno-industrialism. This is a term I invented for basically everything becoming free.

Okay, I haven't heard it.

For basically everything becoming free. This sounds like I'm talking science fiction here, but an approximation of this will happen. Let me put out a thought experiment for you.

Let's take a physical good that's manufactured today, right? Let's take this microphone. This microphone came from a factory, and the factory had three inputs, right? It had people in the factory who were working on it, assembling things. It had input materials, and I would include in those input materials the plastic on here and this foam thing. It also had the machines that maybe injection-molded that plastic. The last thing it consumed was energy, right? Energy is one of the 3 things that made this.

What's interesting is that with the introduction of AI, we're actually converting the human input element to energy, right? Instead of a person having to pick something up and put it onto the next tray, you're going to have a robot pick something up, put it onto the next tray, and consume energy. That human will be able to coordinate hundreds of robots, right? Their personal output will go from picking up an object and putting it onto a tray to essentially picking up 100 objects and putting them onto 100 trays. The trade-off we're making there is that we're increasing the consumption of energy.

You could still say, "Okay, but you are consuming physical inputs." What's interesting about that, though, is let's dig into that. There were some physical inputs and some machines that went into that factory. Well, where did those come from? Those don't grow on trees. They also came from factories. You can ask the same question: What went into that factory? It was energy, it was machines, and it was people, right?

Apply the same rubric, and you realize, okay, actually, it's really just machines and energy. Well, where did those machines come from? If you keep asking this question, what you realize is that energy is the fundamental input.

When we figure out AI and robotics that allow us to do semiautonomous manufacturing, energy will become the cost of all things, right? The cost of buying a thing will become the cost of the energy used to make it. To the extent that we can make energy 10 times cheaper and then 10 times cheaper, that is the extent to which we will be able to make basically everything free.

That's very exciting to me, because if we're going to go explore the stars and set up a place for us to live on Mars and explore the universe, we're going to need a lot of stuff, and it's going to need to be a lot cheaper than it is right now. So that's very exciting to me.

Exciting to me, too. That's a great place to end. Thanks, Isaiah.

Thanks so much for having me.