In this Climate Positive episode (our 100th!), Chad and Guy talk with Matt Loszak, CEO and co-founder of Aalo Atomics, about their innovative approach to building factory-made advanced micro reactors to power the AI-driven energy demand surge. Matt shares his unconventional journey from nuclear engineering student to software entrepreneur to nuclear startup founder, explaining why he believes we're entering a "second atomic age" for clean energy. Matt discusses Aalo's strategy of vertical integration and mass manufacturing, inspired by SpaceX and Tesla, to deliver compact, liquid metal-cooled nuclear reactors that can be deployed rapidly for data centers and other applications. He explains how the regulatory environment has evolved with recent executive orders streamlining pathways to criticality, the company's ambitious timeline to achieve zero power criticality by July 2026, and their vision for 3-cent-per-kilowatt-hour nuclear energy at scale. Matt also addresses public safety perceptions, the advantages of particular reactor technologies, and how their 50-megawatt "Aalo Pod" architecture provides the redundancy and incremental buildout that hyperscalers need.
In this Climate Positive episode (our 100th!), Chad and Guy talk with Matt Loszak, CEO and co-founder of Aalo Atomics, about their innovative approach to building factory-made advanced micro reactors to power the AI-driven energy demand surge. Matt shares his unconventional journey from nuclear engineering student to software entrepreneur to nuclear startup founder, explaining why he believes we're entering a "second atomic age" for clean energy.
Matt discusses Aalo's strategy of vertical integration and mass manufacturing, inspired by SpaceX and Tesla, to deliver compact, liquid metal-cooled nuclear reactors that can be deployed rapidly for data centers and other applications. He explains how the regulatory environment has evolved with recent executive orders streamlining pathways to criticality, the company's ambitious timeline to achieve zero power criticality by July 2026, and their vision for 3-cent-per-kilowatt-hour nuclear energy at scale. Matt also addresses public safety perceptions, the advantages of particular reactor technologies, and how their 50-megawatt "Aalo Pod" architecture provides the redundancy and incremental buildout that hyperscalers need.
Links:
Chad: I am Chad Reed.
Hilary: I'm Hilary Langer.
Gil: I'm Gil Jenkins.
Guy: I'm Guy Van Syckle.
Chad: And this is Climate Positive.
Chad: In today's conversation, Guy and I are joined by Matt Loszak, CEO, and co-founder of Aalo Atomics, a company working to pair next generation nuclear power with the soaring energy demand from AI driven data centers. We talk about his journey from software founder to nuclear entrepreneur, why he thinks nuclear has been so misunderstood, and how smaller factory built reactors could help deliver reliable clean power quickly and cheaply.
This episode marks a couple big milestones for climate positive. It's our 100th episode. And my final time behind the mic. As co-host, this show has been one of the most rewarding parts of my work, giving me an excuse every few weeks to sit down with some of the smartest, most mission-driven folks in climate and energy.
So as we mark 100 episodes, I want to say a very personal thank you. Hosting Climate Positive has been a privilege and a joy. Your feedback, questions, and encouragement have shaped the conversations we have on the show and reminded me that there's a real community participating on the other end.
Although this is my last episode of co-host. Climate positive will absolutely continue. Our expanded team of Gil, Hillary, Guy, and Kenny have some terrific guest lined up, and I'm excited to join you as a listener for this next chapter. So with that, let's get into our conversation with Matt Loszak of Aalo Atomics.
Chad: Matt, thank you so much for joining us today.
Matt: Hey, uh, thanks for having me.
Chad: You have a, a very accomplished but unconventional background for the CEO of an advanced nuclear manufacturer startup. You were a software entrepreneur before jumping into nuclear. So, could you first tell us a little bit about your background and what inspired you to co-found Aalo Atomics?
Matt: First of all, it’s probably worth highlighting. I studied nuclear engineering and physics and university, so I was kind of more of a bandwagon software person getting back into nuclear after a bit of a hiatus, but kind of rewinding all the way. I first got interested in nuclear when I was a kid growing up in Ontario, because we actually used to have a lot of.
Coal power plants, and I lived through the coal to nuclear transition, so there was like 62 smog days per year. I had pretty bad asthma. I’d go to the hospital for it sometimes, but yeah, fortunately the government decided to turn off all those coal plants and we went all in on nuclear, and lo and behold, smog days went to zero and my asthma went away.
And I thought, my God, nuclear, uh, is incredible. So I studied nuclear engineering and physics and university and thought about starting a nuclear company when I graduated, but Fukushima had just happened. So I thought, ah, maybe not the best time to start a nuclear company. So that’s when I taught myself to code.
I still really wanted to do a startup. And so taught myself to code Googling, how do you build an app? How do you build a website? Some degree of my physics and engineering degree had helped with coding. We’d done kind of a lot of Python or MATLAB and things like that, but it was very different. Learning HTML, CSS, JavaScript.
I did PHP, which was kind of a little bit of a interesting choice. Yeah, and so did two software companies. The first one got millions of users but didn’t have the best business model. So the next one I thought, okay, it has to actually make money. So started something kind of like Rippling or Gusto for Canada.
It’s called Humi. It’s essentially HR payroll benefits software for Canadian businesses. And we processed over $10 billion of payroll. Uh, we sold the company for kind of low nine figures, but around 2021 or 2022, I was starting to think maybe now is the right time to start the nuclear company. It still wasn’t obvious.
I mean, we were still shutting down gigawatt scale nuclear plants globally, and I was thinking to myself, you know, what are the odds that a startup can have success if we can’t even keep multi-billion dollar paid off assets alive in the nuclear space? But it felt too meaningful not to do. I think nuclear is the most misunderstood technology probably of all time, and so I wanted to dive in and try to help push it forward and reunite with it after the short hiatus in software.
So yeah, and the rest has been history.
Chad: And how’d you come up with a name? Is it Aalo or Aalo? I’ve listened to a few of your interviews and I don’t remember, but how’d you go with the name? How do you say it?
Matt: It’s Aalo. It means the light in a few different languages. So kind of Persian or Bengali. And I thought it would be a nice nod also ’cause my now co-founder and CTO Yasser is Bengali and the kind of initial technology selection was a reactor that he was developing at Idaho National Lab.
And we’ve really evolved the design a lot since then. That is kind of 300 times the power output and twice the physical size. And there’s a lot of changes to make it a good fit for powering data centers. And I’m sure we’ll get into that. But the seed of the idea was there and then, but the reason I, you know, I wanted to call it Aalo the light, is because we really try to hire optimists and we’re pushing for this bright future.
And I actually think it kind of becomes a self-fulfilling prophecy. More than half the people at your company are pessimists and don’t think things will work out. It’s not gonna work out. And if people believe it’s possible, then there’s a chance. And said differently, there’s nothing more dangerous than a well-spoken pessimist, is what I’ve found.
So yeah, really wanna push for a bright future. I mean, it’s interesting how like in the 1950s the future was this bright, amazing thing. Everyone is like, the future will be awesome. There’s diners that look like spaceships and cars that look like spaceships, and there’s spaceships and we’re going to space and the future is bright and, and these days.
There’s less optimism about the future. Like, AI will be great, but maybe it’ll be bad. And like there’s lots of concern around that. So I think long story short is we really want to push for a bright future and make the future positive.
Chad: That’s great. You have described Aalo as a part of this second atomic age for clean energy. So as you look back over the last few years as you were starting this company, what changed in the world that convinced you that this was the right time to build factory-made advanced micro reactors?
Matt: Well, I actually wasn’t convinced it was the right time, but I thought it was too meaningful not to do.
And we really timed it nicely because when we sat down and we were looking at the potential markets to go after, initially data centers were on the list, but they weren’t necessarily at the top. And there was GPT-2 and then GPT-3 around when we were planning this out. And it was like, first it was a joke.
It had typos and it wrote about unicorns or whatever, and then GPT-3, you’re like, wait, this is getting better pretty fast. And I remember one moment that really stuck out to me, which was when there began to become an emergent property of reasoning coming from the models. So for example, GPT-2, you would give it this thing of saying, here’s the case study.
So you say, I’m on my chair in my living room, I have a cup on the armrest, and I have my wedding ring in the cup. I take the chair and put it upstairs in my bedroom and I take the cup and put it upside down on my bed and then put it on the chest of drawers. Where is my wedding ring? And GPT-2 would be like, it’s in the cup.
Or something like that. But GPT-3 suddenly said, well, it probably fell out onto your bed when you turned it upside down. And that is no longer just next probabilistic token prediction. It’s got an emergent property of reasoning. And I think at that point, a lot of people were like, holy crap, we need to scale this even more.
And so what you have now is like this complete race amongst all the hyperscalers to achieve the highest scale possible. And this requires a lot of energy. So the scale of demand is 40 gigawatts in the US alone in the next five years, and that’s like creating 40 new cities worth of power in a five-year period.
And the last time that happened was never. So we timed things nicely because nuclear is a really beautiful solution for their needs there. I mean, renewables are great, but they use 100 times more land. Our solution is very compact, which makes siting easier and ironically helps fight Nimbyism. We don’t need water, so less water constraint.
I know it’s a bit of a constraint for data centers. It’s not a huge deal — as much as it’s overblown perhaps. And then we don’t need a magma pocket like geothermal or fast moving water like hydro. And then obviously it’s clean, unlike oil and gas. And right now, let’s be honest. The data center’s being powered by gas turbines, but they’re having a lot of pushback from the local communities and there’s low hanging fruit for a few gigawatts of that stuff, but certainly not 40 gigawatts.
To achieve 40 gigawatts, you’ve gotta do new fracking, new pipeline. Long story short, we think that our solution is really ideal, clean base load, small, deployable quickly, and it’s just a question of executing and getting this proven, uh, as fast as possible to meet the demand of the hyperscalers.
Guy: And Matt, the, uh, regulatory environment for nuclear energy seems to, you know, have improved under the administration’s May 2025 executive order, mandating the 18-month NRC review timelines, designating data centers as critical defense facilities, and then, you know, setting this broader 400 gigawatt nuclear target by 2050. How has that type of policy and regulatory prioritization impacted your team?
Matt: Yeah, it’s a great question. So the NRC has really evolved over the past few decades. So initially there was no such thing as the NRC. There was the Atomic Energy Commission and they could both regulate and promote nuclear. And under their watchful eye, all the first 52 test reactors were built at Idaho National Lab.
They tested all sorts of different types of coolants and fuels and things like that, and then for certain reasons, water-based reactors became the initial technology, you know, rabbit hole that things went further down into. And the NRC was created in around 1975 to clamp down on that type of technology.
But the regulator being designed around that tech made it hard for innovation and hard for other technologies he had explored. There was kind of this catch-22 in the 2000s and 2010s where they would want you to have nuclear test data from your innovative reactor before they gave you a license.
But how could you have that test data without having built it and gotten a license to begin with? So the DOE was also formed around the same time, 1977, and actually came from the Atomic Energy Commission, and the DOE is kind of the closest thing to the AEC today ’cause they can regulate or promote nuclear.
So our team came from the Marvel program, some of our first five hires, and that was the first program that ever got the DOE’s approval for construction for a new reactor design. That happened in 2023. We brought those folks on and we started saying, hey, for our first reactor, we should also do DOE authorization, and maybe there’ll be a way to do it somewhat with a commercial lens, but obviously still just an experimental reactor.
And that became very prescient because — now I’m getting to your question — this past year, maybe six months ago, the Trump administration put out these four executive orders and they were very logical and they kind of did what you would hope or expect, but it basically advanced some of these more advanced designs.
And so the first one was around pushing a bunch of companies to achieve criticality, to turn the reactor on by July of 2026, which is like six months away. And they kind of streamlined a lot of pathways to have that happen. So for example, you could have an OTA, another transactional agreement with the DOE directly instead of working through a national lab, which helps to remove a bit of red tape, directly lease the land, things like that. And so we’ve actually now started construction on our first experimental power plant, the LOX, and this is the first new reactor building being built at this national lab since the first atomic age, 50 years ago.
Guy: And it’s wild. Grateful for your efforts there.
Matt: Yeah, super exciting. So that was one of the executive orders — on the criticality. Another one is saying, well, maybe there’s a way to do gigawatt scale commercial reactors under DOE authorization, if it counts as defense critical electric infrastructure, meaning in a wartime, if it were to break out, you could use some compute for military efforts.
And then suddenly the infrastructure is not just normal civilian infrastructure, it’s also kind of got a DOD angle to it, and that could also enable DOE authorization. But you know, obviously our whole model is to really focus on the factory and mass production. So of course, when we deploy thousands of these around the country, we’ll need also NRC approval. We need both.
The last executive order that’s relevant to us in this sense is the idea of harmonization between DOE and NRC. So they’re both gold standard regulators. If you have an operating reactor that the DOE deems safe, then the NRC should probably agree. So it’s meant to increase the overlap in the Venn diagram of DOE and NRC requirements such that you can more easily parlay the DOE approval into an NRC approval once you’ve got this thing built, both at experimental and a hundred megawatt commercial scale. So it’s a very exciting time for nuclear ’cause all that is happening, and I think we can’t ignore the fact that a lot of this is driven by demand for AI and the country acknowledging the importance of that. So exciting times.
Chad: Yeah. And we’ll go into your product and business model in a minute, but you mentioned you have the demand from AI and for low carbon, no carbon energy broadly given climate. You’ve got the regulatory environment favorable, at least in the US, and public policy working at this point to your benefit. There is still a public perception of nuclear that’s really shaped by legacy fears, safety fears in particular. So could you talk us through how you are addressing those with the design of your product?
Matt: First of all, try to cut through the noise and identify what are real safety concerns and what are fake ones or propaganda. The worst thing would be designing tech to address propaganda. I think that would be silly, and surprisingly that has been done. For example, burying nuclear waste miles beneath a mountain is engineering around propaganda, which I think is ridiculous. However, if you look at the past three well-known nuclear incidents, while it is true that nobody died in the past two — Three Mile Island, Fukushima — which surprises a lot of people, there were relocations, and really that’s what we’re trying to engineer: prevent relocations.
So it’s kind of cutting through the propaganda — actually wait, there were no deaths in those last two. But then, okay, how do we avoid relocations? That’s our kind of engineering objective. And by the way, Chernobyl is kind of a strange case because it’s kind of like they didn’t have a containment dome, which is kind of like having a car with no seat belts. So you can very easily just make that change and say, okay, this can be made much safer that way.
On the topic of safety, nuclear today is as safe as solar and wind on a deaths per kilowatt hour basis, which is a morbid statistic — like how many people die for every unit of energy that is produced. Nuclear is actually on par already with solar and wind. So you can kind of ask, well, okay, how much safer do you really wanna make it, kind of thing. But you do want to avoid relocations, and so part of the way that we’re doing that is with liquid metal coolant. There’s not been a radioactive release with this coolant before.
There have been incidents with smoke or a fire, which is natural if you deploy a lot of something, but it’s not a radioactive release. It’s more of an industrial hazard, just a pain for the utility. But there’s lots of advantages to sodium as well, so it makes it more mass manufacturable. The reactors can be more compact and at atmospheric pressure, which helps to avoid a major excursion, so it’s not a pressurized system like traditional reactors.
Additionally, sodium is very thermally conductive. It’s 100 times more thermally conductive than water, molten salt, or gas. So it helps you avoid a Fukushima situation where if you lose backup power, our reactor can actually keep itself cool. It doesn’t need backup power to keep circulating the coolant. So there’s things like that that help optimize the design.
But just coming back to the topic of Chernobyl, for example, right? So there were 45 deaths immediately, and then there were a couple hundred thousand people who may have died in their seventies and eighties, decades after the incident. And that sounds like a lot, right? But at the same time, every year 2 million people die early from oil and gas PM 2.5 particles in the air.
So again, I’m not saying 200,000 deaths from Chernobyl is okay, but I’m saying we have to put this into perspective and optimize for the safest thing for humanity. And this is from me where I had asthma and it went away when we went to nuclear. Like people don’t often realize nuclear cleans the air and actually it also produces medical isotopes that help fight cancer. So it’s the only energy source that has a net negative death count, if you consider that. But I think we do have to engineer and make sure that there’s no relocations and we engineer around radiation releases.
Guy: You touched on this briefly previously, but the idea that the real innovation here is around the manufacturing process and the plant, that manufacturing plant really being the product. Can you talk us through kind of some of the design considerations that went into the system to make it mass manufacturable and then really drive down that end cost per kilowatt hour?
Matt: Yeah, for sure. Just one quick note to make before we dive into that. On the nuclear waste side, I think this one is basically pure propaganda. Like nuclear waste is very easy to contain. There’s a tiny amount of it. If you took the entire waste from the entire nuclear production in the past 70 years in the US, which is powering 20% on average of our country, 300 million people, right? 20% of that for 70 years, all that waste fits on a football field. The real waste is six inches high on a football field.
Guy: Wow.
Matt: So it’s a tiny amount of waste, easy to contain, never harmed anyone. I’d much prefer that than oil and gas, which puts the waste into the air we breathe. Or even renewables after end of life have to be recycled or buried underground or something. So I’m very confident in the way nuclear handles its waste, and I think that is basically pure propaganda.
Okay, so what design decisions did we make to make this mass manufacturable? Sodium as a coolant, like I was saying, allows you to make the reactor more compact and allows you to have a much better supply chain. So we try to make sure that we never choose unobtainium — things that are hard to get in the supply chain. And part of that, for example, atmospheric pressure means our vessels can be less than an inch thick, like 5/8 inch, 3/16 inch grade stainless steel. Traditional PWRs, pressurized water reactors, need six-inch thick vessels that are very hard to make, with like a six to 12 month lead time. On the contrary, we can plate roll our own vessels in our factory in hours, not six to 12 months.
Secondly, a water-based reactor of the same power output would be twice as long. And you have to think not only about mass manufacturability, but also transportability. And some of our competitors, I won’t name names, but finished their reactor design and then realized it was too long to ship on normal roads without taking down power lines. So ours would be like half the length with the same power output.
And then, again, this is kind of tied to your question, is supply chain. We chose off the shelf uranium dioxide fuel, not TRISO, not HALEU. These are like higher enrichment or special chemistries, which have certain advantages, but the supply chain is really poor for those, and we think that’ll be a mistake in terms of scalability to stand a chance at servicing even a fraction of the demand of these hyperscalers, at least for the next five to seven years.
So those are a few examples. And what I wanna highlight there is like one of our competitors is publicly traded, and they achieved a 24 billion market cap. But what’s surprising is they are a fully remote company. They don’t have a factory. And so to us this seems like the wrong strategy. We’ve learned from SpaceX and Tesla that vertical integration, taking control over how things get made, is the solution to better economics, a bigger moat, and proving technology and really achieving certainty on schedules and timelines.
If you look at Vogel, the last nuclear plant built in the US, it was 10 years over schedule and $15 billion over budget because it was kind of a similar model where there was a designer who then sent the design to dozens of suppliers, each with their own factory, and they made all these modules, which then came to site and didn’t fit together properly. And lo and behold, there were reworks and slowdowns.
So we tried to bake in DFMA — Designed for Manufacturability and Assembly — into the reactor design and the plant design. The whole balance of plant as well. We make it in the factory to the highest extent possible, and that way we take raw materials into our factory and we output modules, like 40 or so per pod, and we know those modules will fit together ’cause we’ve done that work in the factory.
It’s a bit more ambitious. It takes more capital. We spend half our capital in the factory and half on the product. But we think it’s the right approach because the alternative is you finish your first plant and then you say, okay, now we’ll mass manufacture it, but then that involves design changes, baking in DFMA, and then suddenly maybe it has to get relicensed on the regulatory side. And we think it’s just way better to have these things develop hand in hand and have these two products from the beginning, the factory and the power plant, which we call the Aalo pod.
Chad: You noted that the challenges with traditional larger scale nuclear are often costs and timeline overruns. And so you’ve set pretty aggressive goals — 3 cents a kilowatt hour energy, as well as a timeline for groundbreaking to electron production of about 60 days. How’d you come up with these goals specifically, and how achievable are they, and by when?
Matt: Yeah. The company mission is 3 cents per kilowatt hour, so it’s not our first-of-a-kind target or anything like that. I think we’ll be like around 15 cents initially per kilowatt hour and come down to maybe seven at half of a kind with our first product. In the long term, we want to find the right balance of economy of numbers and economy of scale, so we can make a reactor actually probably five or 10 times more powerful while only being slightly larger. So you can imagine what the economics will look like there if it’s the same amount of stainless steel and hardware producing 10 times the revenue. That’s kind of our thinking on the LCOE target of 3 cents per kilowatt hour, which we think is doable at scale with the right product.
And again, this has not been possible in the past ’cause there wasn’t enough demand from a single type of customer, from a consistent type of product that would allow you to come down a cost curve in mass manufacturing. Every other reactor built in the US in the past 70 years is bespoke. They’re all kind of one-off. So how could that ever come down in cost? That’s our thesis. We think there’s something special here in terms of the environment being different, and from that, something new can emerge.
And then in terms of July next year, we’re not gonna have an entire power plant done by then. What we’re gonna achieve is zero power criticality. So what that means is we’re going to manufacture the vessel and the fuel. It’ll be full scale, there’ll be a building, so it’ll be, you know, reactor operators, everything. We’ve done it ourselves at full scale. If other competitors claim they’re achieving criticality without having done it all themselves, I don’t really count that personally. And there’s some interesting marketing happening right now in the space around that.
But yeah, we will have built it all ourselves. It’ll be full scale. And what zero power criticality means is it’s kind of like getting in your car and turning the ignition, but not yet hitting the accelerator. So the reactor will be started up, so to speak, but it’s not gonna be producing very much electricity and not very much radiation. So that’s what’s happening there.
But yeah, one more thing worth mentioning is we’re pretty unique in the sense that we’re doing DOE authorization for this experimental plant, which will not only use heat, but it’ll also actually have a turbine, produce electricity, and power a data center. And we think this might be the world’s first home-built nuclear-powered data center.
Guy: Love it. And can you remind us when — what date in the future — you expect to be fully energized next to a commercial data center?
Matt: It could happen end of next year, early year after. So this will be 30 megawatt thermal, 10 megawatts electric, which is one reactor and one turbine. Our commercial product is five reactors, one turbine, 50 megawatts electric, with 150 megawatts of thermal coming from the five reactors going to a single 50 megawatt electric turbine. That’s the Aalo pod. And we chose that architecture because data centers tend to get built out in these 100 or 200 megawatt shells, right? So we wanted redundancy at the pod level for redundancy at the reactor and turbine level. That way we can refuel one reactor at a time and not short the whole power plant when you’re refueling a single reactor, as you would be with a gigawatt scale nuclear plant.
And secondly, you can build it incrementally much faster, ’cause the data center shells take like six to nine months to build. So we’d much rather build smaller nuclear plants incrementally in parallel versus wait 10 years for a gigawatt scale plant to get built before your first electron. So those are the major value props. Our power’s a bit more expensive than an on-time, on-budget gigawatt scale nuclear plant, but we think we have the better product in terms of product market fit for these customers.
Guy: Got it. And hopefully that declining cost curve with time as y’all really ramp the manufacturing and the deployment. So certainly exciting times ahead. So we’ve talked a bit on the data center front, and it’s interesting to hear more about the thermal output from the system. What other industrial applications, and more broadly, do you see the pod kind of best suiting going forward?
Matt: We kind of think of ourselves as — one day we want to be like Amazon, you know, the everything store. So this could really be like the everything energy product. But they didn’t start with everything, right? They started with books because no bookstore could compete with an online situation where you can have like a trillion SKUs that can never be done in a physical bookstore, and then they expanded from there.
So we think of data centers in the same way. It’s an amazing introductory wedge market. They have a high willingness to pay in return for speed. Clean is nice, and then making siting easy is a good benefit. But we’ll come down the cost curve with that customer and then we’ll go after all these other markets and it really opens up the world. So as we do the larger product and come down in LCOE, we can do gigawatt scale utility power with our initial product. We could even do smaller municipal power because if you’re a small municipal utility, you don’t want a gigawatt scale plant, but you could use a 10, 50, a hundred megawatt. That was never economical before. You’d say, hey, we’ll be 15 cents initially. You come down to 7 cents a kilowatt hour and they say, okay, we’ll come back when you’re at seven.
That’ll open up markets — desalination, industrial process heat. We also produce high temperature steam, which starts off at like 500°C and then can either be converted to electricity or can go straight to certain chemical processing plants or what have you. I think at 500°C we can be relevant for roughly half of all industrial process heat applications.
Guy: Yeah. Huge for industrial decarbonization, that low to medium temperature heat.
Matt: And then for the rest we could maybe use electricity to get to higher temps. A bit more expensive there. And then desalination — it’s interesting ’cause in Texas where we’re based, they use a lot of water for fracking and things like that. So we can help to do manufactured water as well and clean up some of the oil and gas mining processes and so on. So it’s really exciting.
But one more thing I wanna say. It’s wild to say this on a podcast, but I think what AI is going to do for us is going to be insane. And this comes back to the light and optimism. I know a lot of people are like, it’s gonna be terrible, but I actually think there’s gonna be so many cool things that happen in the next 10 or 20 years with AI, especially as we hopefully power a lot of it. So for example, people might not need to work anymore. Work might be optional. AI might solve all disease. Maybe our loved ones will be kept alive for longer. You guys saw how it went from text generation to image generation to video. There’s some talk of video games that you describe a world and you can just go walk around in that world in this like real time photorealistic video game, I mean.
Chad: In Star Trek they called that the holodeck, I think.
Matt: Yeah. Right. And then speaking of Star Trek, once it starts to discover new physics, imagine the technology that could be available to us. I mean, maybe there’ll be anti-gravity boots. You can go walk on water or fly around the sky or something like Iron Man.
So I think the future is going to be very good. And it’s just crazy how society right now is built primarily upon dead plants and dinosaurs, of which there’s only one or 200 years remaining of supply if we were to use all of it. And with nuclear, it can last us 4 billion years. So I think when we use a lot more energy, a lot of good things will happen. That’s, I think, one of the most exciting things about this.
Chad: Well, sticking to this optimistic future where the Terminator doesn’t kill us all, you’ve raised now over a hundred million dollars and obviously have pretty aggressive ambitions. So let’s say we’re in 2035. AI is good for humanity, you’re growing. What does your company look like then? How many pods have you deployed? How many gigawatts online? What is your role in global energy supply?
Matt: Our goal is to have these gigawatt factories deployed around the world. Initially we’re trying to get to a few gigawatts per year, which is ambitious, but also kind of not — in the sense that the biggest solar factories produce 150 gigawatts per year of capacity. Now that is not as high a capacity factor and there’s lots of stipulations, obviously. But I think if this demand growth continues, then we should be positioning ourselves to potentially have factories that can even do 10 or a hundred gigawatts of nuclear per year, maybe per factory, maybe in different locations around the world as well.
I think it’s gonna be really interesting to see what happens with data centers in space, because when you put solar panels in space, they achieve like four to 10 times the output without the atmosphere and nighttime clouds, and Starlink can use normal solar panels because the economics of a Starlink satellite is such that it can burn up in the atmosphere every three or four years, and they just put new ones up there.
Same with GPUs. GPUs get replaced on that cadence. They’ll be addressing questions around the rest of the infrastructure for cooling and things like that. I think a lot of people think that inference will largely happen in space while training will happen on Earth, ’cause for training, you gotta send the whole internet’s updated data all the time, and that would clog up the entire Starlink infrastructure if that was done in space.
And then also probably in one or 200 years we’ll be using so much energy on Earth that the waste heat alone and the heat from consumption of energy will cause too much warming on the planet. So we’ll have to do energy generation and utilization in space probably within 200 or 300 years. Otherwise we just literally overheat the planet, forgetting CO2. So I think there’s things like that that’ll happen on that timescale.
I know you asked more about 10 years. And I do have theories as to how AI will remain positive for humanity and not do the Terminator scenario, but that’s probably outta the scope of this conversation, maybe over a beer.
Chad: We’ll have you back on to discuss that. Well, Matt, thank you so much. We’re almost done, but first we have the hot seat, so we’re gonna ask for your immediate reactions to the following statements or questions. First one is: something I thought was true that I no longer believe.
Matt: Well, I have to go with the topical one, which is when I was studying nuclear at university, I even believed some of the propaganda that had been told to me before, around the stigma on nuclear waste, or is radiation uniquely dangerous? But the more you look at it, like I kind of say there’s two types of people in the world: people who love nuclear energy, and those who haven’t looked closely at it yet.
So even when I studied nuclear, the propaganda can still remain. And if you’re handling something radioactive, you can be like, is this like uniquely dangerous? But the reality is it’s not. And it’s like, if you were to eat a chunk of nuclear fuel, that would be bad. But if you drank gasoline, that would also be bad. So it’s like there’s danger all over the world and I think the potential for good that nuclear has to do for humanity is immense. And I think the more we shift out of the falsehoods and the propaganda, the more we unshackle ourselves to enjoy a really incredible future.
Guy: Appreciate your shifting in perspectives there. And then, uh, one more. What is something you are particularly proud of, Matt?
Matt: I’m definitely proud of Aalo. I think it’s been really great so far. I mean, two years ago we were two people. Today we’re a hundred people and I think we’re really well positioned to have a big impact in the space. Yeah, I could probably go on there on different topics, but Aalo is definitely one of the top ones for me.
Chad: Next one. To recharge, I?
Matt: Well, I, uh, try to stay pretty active and healthy and eat well. So I do like a 10 minute run every day around the parking lot. Like, you know, some kind of maniac just running around the parking lot of our factory, 10 minutes every day.
Chad: Do you sleep in the factory too, like Elon did early on in his days?
Matt: We’ve had some pretty late nights, but I, I haven’t set up a sofa here yet to sleep on, but I probably will at some point. I do, you know, a little bit of 10 minute yoga stretch every day. Try to do weights every day. I find that really helpful for, uh, keeping fresh and clearing the mind after a long day.
Chad: Do you wanna do the next one?
Guy: Sure. What is a book or article or teacher that influenced you most over the years?
Matt: I had a, uh, professor at university who taught a pretty badass sounding course called Lasers, and he was incredibly gifted at teaching and also kind of helped encourage entrepreneurial behavior. And so I had a bunch of conversations with him outside class about how to combine physics and entrepreneurship and startups and invention, and having someone believe in you and kind of say, hey, I think you should pursue some of these things, was a really nice, supportive force.
I think we tend to underestimate the importance of even a single voice early in our journeys through life in terms of encouraging you to nudge in a certain direction. So when I was considering quitting my first job outside of university, some people were saying, ah, maybe you shouldn’t leave your day job. But I’m super glad I did. I think it’s people like him who helped give the confidence to take the leap and try and figure things out.
Chad: Awesome. So last one, I promise. Thank you for doing these on the fly, by the way — we usually provide them in advance. So, to me, climate positive means…
Matt: Hmm, climate positive. Well, probably just means not killing ourselves. I mean, the climate is our environment and we wanna make sure that we build lots of cool stuff that makes humans happy and helps us to discover new science and explore and live better lives. But if that comes at the cost of ruining the planet, then that’s obviously a problem. So to me, that’s probably what climate positive would imply.
Chad: So avoiding the Terminator scenario in all circumstances.
Matt: Yeah. Avoiding death by climate. Yeah.
Chad: Awesome. Well, thank you so much, Matt. This has been really great. Appreciate you taking the time and wishing you the best of luck.
Matt: Awesome. Thanks, Chad.
Chad: If you enjoyed this week's podcast, please leave us a rating and review on Apple and Spotify. It really helps us reach more listeners. You can also email us at climatepositive@hasi.com to let us know what you thought. I'm Chad Reed and this is Climate Positive.