C (2:58)

Yeah. Again, Cody, thanks for having me. So, Trey Lauderdale, CEO and founder of Atomic Canyon. We're a generative AI nuclear power company located in San Luis Obispo, California. In case you don't know, which I'm sure all of your listeners will, we are going through one of the most, what I'd say is exciting periods in modern history, if not the entire history of mankind, where we have artificial intelligence that's transforming multiple industries, transforming the way we live. That AI Renaissance resurgence, whatever you want to call it, requires a tremendous amount of power. So our view is nuclear power as a vertical is going to get absolutely transformed through the implementation of, of specific AI models that can help improve efficiency and workflow. So we're the leader in helping to blaze that trail forward and ideally by enabling nuclear power to be more cost effective and grow at scale. This AI super cycle can be powered by nuclear, which in my view is clean, reliable energy, baseload 24 by 7. So you will find a big advocate in nuclear power in me. So that's what we're doing. We're about two years old and we're having a blast. It's a lot of fun.

B (4:09)

Certainly I'm going to spend some time understanding what it means to be a AI company in the nuclear space because I think none of us probably want our neighborhood nuclear power plants run by AI just yet.

C (4:20)

Yeah, yeah.

B (4:20)

I'm really interested in hearing your journey. You and I first got to know each other and connected through the techstars network when you were doing healthcare stuff. And I reconnected with you maybe a year and a half ago as you were diving into this journey with Atomic Canyon. And it was fascinating to hear your personal story of how that came to be. So maybe start there and then we'll get into what the company actually does after that.

C (4:45)

You know, as an entrepreneur, I love telling my story so I can go on and on about it. While Atomic Canyon, we're just coming up on our two year anniversary, myself and the technology team we pulled together, we've spent probably about 16, 17 years before that deploying technology to the healthcare arena. It's now called digital health. Back in my day when I was starting companies in the space, it was healthcare IT. In 2008, I had the privilege of starting a company, Volt. It was the first company that brought iPhones into hospitals for clinical communication. And when I say that people, what does that even mean? People forget. In 2008, when the iPhone came around, I used to go and tell hospitals, hospital administrators of amazing facilities like Cedars Sinai in your backyard, Stanford, UCSF a little bit further north. I'd meet with their leadership and say, hey, your nurses and doctors need to transition from pagers. And these legacy voiceover IP phones. And you really need to be thinking about an iPhone in the hand of every nurse and doctor so they can view their electronic medical record, they can send HIPAA compliant text messages, they can receive alerts. This mobile technology is going to be transformative and time has changed so much. But back then I used to literally get laughed out of the hospital. The former CIO of Cedars is a good friend of mine. He was an investor and he told me, trey, I don't know if nurses and doctors would ever want to text. No one really text like remove that functionality. And lo and behold, texting became like one of the biggest features. And another CIO told me, trey, we'll never allow iPhones on our Wi FI network. Mark my words, BlackBerry is our device. You need to go build on BlackBerry. And I bring this up not to make fun of those former executives, but, but just to really emphasize when you're going through a technological transition, whether it's smartphones, whether it's AI, it's really hard for organizations to realize what is the world going to look like in 5, 10, 15 years. Went through that whole journey. We ended up being on the right side of technology. Eventually we scaled that company, deployed it to hundreds of thousands of nurses and doctors, and then we got acquired in 2019 by a hospital bed manufacturer, Big publicly traded company. I got to lead the digital business at that company. Hill Rom spent a couple of years in corporate America, which is great. Got to do mergers and acquisitions and get to see what it's like to transform. I'd say a sleepy medical device company, but hospital beds, not like the sexiest thing ever. Eventually we sold that company as well. After that, I did what every entrepreneur does. After a couple exits, I became pseudo retired. Was doing angel investing, vc, private equity work, and a lot of the themes I was saying was artificial intelligence entering healthcare. So some work with computer vision and radiology. There was a company that was doing automatic speech recognition, recording patient physician encounters, creating physician notes, storing that in the emr. And I started to witness AI has been around forever. But we came to this period where now suddenly AI was ready for mission critical applications. Also in that journey I witnessed, oh my goodness, even these little startups, they're using tons of GPU hours, like whether it's in Google or Microsoft, you name it, like the GPU usage just going up and up and up. Little companies. So I did a little bit of the back of the napkin math and realized if this AI thing takes off, we're going to need Gigawatts of power, unbelievable amounts of power separately. As luck would have it, I'm from Florida, my wife is from California. We have young kids. So once we had financial flexibility, after a couple exits, clearly moving wherever she wants and we moved closer to her family. San Luis Obispo, California, where I'm sitting right now. Little did I know when we purchased the house we're in, I'm 10 miles downwind of Diablo Canyon, which is California's last remaining nuclear power plant. So upon that discovery, my first response was actually fear is like, oh my goodness, what have I done? My knowledge of a nuclear power plant before this venture was the Simpsons. You know it's going to be green ooze and three eyed fish and pollution everywhere. So I did a bunch of research very quickly and realized nuclear power plants actually don't release pollution. They're actually incredibly efficient. They're safe. Everything I thought I knew about nuclear was totally wrong. And it was actually quite the opposite. A nuclear power plant is this huge economic generator of not just for power, but for jobs and tax revenue. It actually pays for a bunch of our schools and everything else. So I didn't think anything of it. But over time, as you live in the proximity of a nuclear power plant, they employ 1300 head of household jobs in a town the size of slow 40,000 people. That's a material amount. So I started meeting all these people in nuclear and you know, I'm a curious guy. So I'd ask, tell me about the plant and how does it work? And quickly realized besides nuclear being a great economic engine for our community, it's just a modern day miracle. The fact that we can split atoms, create heat, little San Luis Obispo, 2 gigawatts of power, 24, 7, 10% of the state of California's power right here. I started to ask, well, why isn't this the standard? Like why do we not have nuclear power plants across our entire world? And you quickly realize the first challenge of public perception. You know, people still need to become more educated on nuclear. That's changing quite dramatically, especially the last few years. So if you didn't get to, okay, what's the second problem? If we get through public perception, what's the next big challenge? Realized that nuclear just has an unbelievable amount of administrative burden. The regulatory work, the paperwork to operate a plant, to, to design a plant, to build a plant, to get one license, just unbelievable amounts of paperwork. And that's where the idea got accepted.

B (10:41)

Sounds like an industry you already had worked in, in healthcare, in Terms of heavy regulatory environment, yes, to some extent.

C (10:49)

But coming into nuclear, it's just exponentially more regulated.

B (10:54)

I mean, interesting.

C (10:55)

Trying to get something done in a nuclear power plant just takes huge amounts of paperwork. And the regulation is very, very intense. I'm not saying whether that's good. I don't think that's my role to state healthcare. I found it way less regulated. I mean, in a hospital, people die every day in hospitals. Like that's just the nature of the beast. Nuclear power plants, if someone gets injured, it's reported to the NRC publicly available for everyone, it's treated completely different. So that's how the company came together was I realized nuclear is going to have tremendous potential to be a clean energy source for this AI renaissance that we're saying. However, everyone's focused on what can nuclear power do for AI. I don't believe enough people are focused on what are the potential applications of this incredible technology. Artificial intelligence helping to improve nuclear power and to help make nuclear power more effective, less expensive, and to an extent, safer. So that's where I got the idea. And that was the start of the.

B (11:53)

Company you were hanging out in slo, meeting people, working in Diablo Canyon and kind of had this Aha, that, wow. It sounds like a lot of the internal process that they're having to navigate is maybe archaic. I don't know if that's too harsh, but behind the times from a technological perspective. And where did you then go from there?

C (12:14)

My first point of contact was with my wife and my family where I had to say, this is going to sound nuts, but I'm going to start a new company and I'm going to do it in nuclear power. To which my wife and my extended family, I think they all had to do wellness checks on me. Like, Trey, like, is everything okay? Is this a midlife crisis? Like, you're a healthcare guy, you've got a huge health care network, you know all the investors. Why would you pivot mid career into something you know nothing about? But there's something about entrepreneurship. And I tell aspiring entrepreneurs, you'll know when you fall in love. Like when you fall in love with this idea, it just, you become obsessed. You can't stop thinking about it. I found myself sitting in board meetings that I was at for healthcare companies and I'd be like, googling stuff about nuclear and learning more about nuclear. And in this day and age, information's at your fingertips. So I was able to become educated very quickly. And once I made that decision, it's like, okay, well how do you move forward? You just go like entrepreneurship. There is no perfect playbook how to build a company. You just have to start moving. So the first thing I did was I reached out to a the best AI engineers, people I worked with in healthcare and said, I'm launching this company. Luckily, through my exits, I have the financial means to back it out the gates. So I told them I'm doing it. I'd love them to come on. A bunch of engineers jumped in. I said, guys, what do we do? They said, well, the first thing we need is data. So I was like, okay, let's go find data. So we found the Nuclear Regulatory Commission has this data repository called the Adams database. It's a bunch of publicly available information about our fleet. Nuclear, very big into transparency. So There are about 53 million pages of documents. I reached out to the NRC and said, hey, just a heads up, we're going to be downloading all these documents. We're not a foreign actor or anything. We just could be building AI models. Surprisingly, the NRC responded to us and they were curious what we're doing. They're super supportive. We downloaded all 53 million pages of documents. And originally I thought this would be really easy. We'll be a wrapper on top of ChatGPT. We'll throw this into like an OpenAI Azure cluster. Bada boom, bada bing. We do generative AI. Let's go. Quickly realized, we started testing Microsoft, Copilot, Gemini, all the standard LLMs.

B (14:22)

Did you know what you were going to do with these documents when you downloaded them all? Like, what was the.

C (14:25)

Nope.

B (14:26)

So it was like, let's see what's in there and see what we got.

C (14:29)

The first step is how does the core AI technology even do a nuclear? Like, before we start problem solving, is the technology at a spot where we could introduce it to solve problems. That's kind of first step is like, you got to get through your technology risk. You got to de risk the technology first. We went in and we quickly realized that AI models off the shelf, they hallucinate, they make stuff up like crazy when you put nuclear documents into them. And the reasoning for that is they haven't seen enough examples. Nuclear is like its own language. The acronyms, the words, like sometimes acronyms mean different things based on the context of how they're stated. So it's a pretty tough and meaty kind of technical challenge to solve. So I asked my engineers, like, what do we do? And they said, we got to go build nuclear specific sentence embedding. Models. And I said, okay, I've never heard of that, but I trust you guys. What do you need? And they said, we need a lot of GPU. We need like a few thousand Nvidia H1 hundreds. I was like, okay, well that's pretty expensive. And from a preference perspective, I'm not a big fan of being an AI company raising 50 million pre seed from Nvidia. You give it right back to Nvidia. Like I don't think that's good hygiene for a company. So I was on a mission to find a supercomputer I could use for free. So just started networking through the Department of Energy, got connected with Oak Ridge National Laboratory, one of the premier laboratories that our DOE sponsors. They have a big group in nuclear and they also are home to Frontier, one of the world's fastest supercomputers.

B (16:00)

By the way, we just had Susan Hubbard on the show like just recently. So yeah. Anyone listening? Go listen to the archives, you know, just from a few weeks ago and listen to everything Oak Ridge is up to.

C (16:10)

Oh yeah, Oak Ridge and our national labs in general, they are just such incredible resources and assets for our country. So I presented to them like, hey, we've got this set of data. We want to build nuclear specific models. We would like to use your supercomputer. Ironically, their nuclear PhDs had seen the exact same problem and they said, can you solve it? I'm like, we think we can. So they gave us access to 20,000 GPU node hours for us to go build nuclear specific sentence embedding models. So team went at that, took about six months. It's really intense, like harnessing a supercomputer. Getting things to work in parallel is like, okay, I believe my team is going to de risk the technology now it's time to go find a problem to be solved. So I reached out to Diablo Canyon. Didn't have any connections there. I'd like to say it was just old school entrepreneur reach out to everyone. You got to be persistent. You're going to hear 100 no's. But here, before you hear a maybe.

B (17:07)

I literally am picturing you like knocking on a door and Smithers, like opening the door and saying hello.

C (17:13)

I wish it was that. Nuclear power plants have a very strict security zone. If I would have been allowed to go in and knock on the door, I would have. But I couldn't even get past the first security gate. And maybe over a beer I'll give you some stories there. But nuclear plants are locked down, so it was a lot of Virtual email asking people I know who they know. It's again, it's a small community. So kids on my soccer team, I talk to their parents like, oh, you know, somewhat Diablo and I have no shame. Like I would talk to everyone I could find. And eventually I got to meet with the leadership at Diablo and, and I came at them with a lot of humility and said, look, right now as a company we've done a lot of research, we're really good at AI but I don't know your problems. So I'm not going to come here and tell you this is why I can go solve. What I'd love to do is have the opportunity to come in, bring my product team and just observe and we will put in, the hours will come day shift, night shift. We just want to observe what your employees do, where the challenges are. And then after doing that a couple of months, we'll come forward and tell you where we believe AI is ready to solve problems and where it's not ready to solve solve problems. They agreed. We spent a whole bunch of time on site and we quickly realized there's a very baseline problem to be solved, which is any nuclear power plant, whether it's Diablo or any of the other great 54 plants we have across the US they have to store a tremendous amount of data. I mean, we're talking about billions and billions of pages of documents from the entire history of the plant on their on premise record management system. And when I say one system, there's usually many systems across multiple departments and their employees. Whenever they're getting ready to do any work product, whether that's in their regulatory team, their engineering team, their maintenance team, the first step before they do any work is they have to gather documents. They have to say, okay, well what is it that we did for maintenance on this generator the last couple years and what's the project we're going to do? And then we have to go and design how the change is going to occur and have to go implement that. Then we have to document what we did. There's a lot of documents flowing back and forth, but the first step is just finding the documents. And the search products that they had were unbelievably archaic. Like we are talking about Boolean S search. You have to know the name of the document, you can't see the contents of the document. And that's where we came and said that right there. Like to your earlier point, Cody, we do not say AI should run a nuclear power plant. We are not there yet. Living 10 miles from 1. AI is not ready for that. However, assisting employees, helping them find documents, which is super tedious, which they don't like to do. Nuclear engineers don't want to search for documents, they want to do nuclear engineering work, enabling a software technology, an agent if you will to help them find documents. Perfect place to start. So we came recommended that that be the solution we solved. They agreed to and we were off to the races. So from there the abridged version, a lot of challenges in nuclear because very security intense. So they didn't want anything in the cloud. So I know a bunch of Nvidia folks from my healthcare day. We designed and implemented the first Nvidia H1 hundreds, the GPUs. We put them on premise. At Diablo we architected a solution that uses a series of AI models, optical character recognition, computer vision to ingest and read all billions of the documents they have. We then use the AI we built at Oak Ridge. We put that on premise to understand the context and what's in all those documents. And then we built a really advanced search application. It makes it really easy to read and find and summarize what's in those documents. And there's all sorts of workflow, like when you find a document you want to know it might be a thousand page document, where inside that document is the relevant information I'm looking for. So we have all sorts of like really advanced workflow to make it easy to find documents. So that's the first product we built. We took Diablo live starting in May. It was a rolling go Live wrapped up this past August and it's been a grand slam home run. They won all sorts of awards and they're thrilled. We got hundreds of users on the platform daily. So yeah, a lot, a lot of success and we've been super thrilled with Diablo as a customer and the opportunity to serve them and others in the industry.

D (21:33)

Hey everyone, I'm Yin, a partner at mcj, here to take a quick minute to tell you about the MCJ Collective membership. Globally, startups are rewriting industries to be cleaner, more profitable and more secure. And at mcj we recognize that a rapidly changing business landscape requires a workforce that can adapt. MCJ Collective is a vetted member network for tech and industry leaders. Leaders who are building, working for or advising on solutions that can address the transition of energy and industry. MCJ Collective connects members with one another with MCJ's portfolio and our broader network. We do this through a powerful member hub, timely introductions, curated events and a unique talent matchmaking system. And opportunities to learn from peers and podcast guests. We started in 2019 and have grown thousands of members globally. If you want to learn more, head over to MCJ VC and click the membership tab at the top. Thanks and enjoy the rest of the show.

B (22:35)

From a security perspective, are you sending these queries like over the cloud to Oak Ridge? Are you building stuff on site there? Like how does the actual setup work for building software for an active nuclear power plant?

C (22:51)

Especially because this is AI. The security reviews, as they should, were unbelievably thorough. So to answer your question, this solution, per the request of Diablo and PG&E, is all on premise. So nothing leaves the firewall, it's all export control, highly sensitive data. We had to architect all sorts of security and permission. So when you're searching based on your role, there's only certain documents you could search or get access to. All of that had to be replicated in our AI what we built at Oak Ridge, the software that we built, it's our Fermi sentence embedding models are actually installed on premise on the hardware that's there and everything runs self contained. Nothing exits their firewall and that's how their solution was designed.

B (23:39)

So if I think about it, take the nuclear side out of it and just think about it from an AI architecture perspective. You used Oak Ridge basically as your giant training database to build all the training knowledge of these nuclear specific words. And then you took that and your inference search is actually living on prem at Diablo Canyon, which enables people to do these specific queries and get results back. Is that accurate?

C (24:04)

Cody, I am so impressed. You used inference search, you used. Yeah, we used Oak Ridge with the NRC data to train the specific models, then we implemented them and all the inferences done on premise with the H1 hundreds that are on site and then we design the solution. So there's a lot of headroom. So if we have to do any additional fine tuning on their specific data set that can ever leave and we can't use that for other plants. But we have all sorts of capability for like different sub departments and that kind of goes into roadmap of what we're building and kind of the next phase of the project. When I talk to people, everyone's like, well search that's kind of boring. To which I tell them, in nuclear we have a saying and I'm allowed to claim we because I've been in the industry two years now, so I'm officially part of the nuclear crew. Boring is good. We love boring and nuclear. So search is beautiful because it solves a real problem. And I can describe it to someone in simple terms. It adds value it and then it demonstrates that AI can actually work reliably. And as a result it gives us a ton of credibility and capability to then launch, you know, the next phases of the project and what we're doing at Diablo and other locations.

B (25:16)

And what is it about using AI and the type of particular technologies you're using is rag right? Retrieval, augmented generation, which is basically AI enabled search. What about it is an order of magnitude better for this use case than what search has had for the last 20 years. Why couldn't Google have come in and solved this problem with a vertically specific Google search engine 15 years ago specifically for nuclear.

C (25:43)

But we also saw this in healthcare. There's a whole bunch of problems to be solved before you get to the actual technology itself. For example, at Diablo Canyon they have a series of different locations of their data. It's not like there's like a master perfect record management system. There's like SAP, there's a documentum instance, it's like 15 years old, there's another onedrive, there's all these different data locations. And in many cases because they're on premise, the nuclear power plant is not forced to do upgrades every year. It's not like the cloud where you have no choice to upgrade so many of these instances. I won't say exactly, but we had one data set that was like a 14 year old data repository that had been upgraded. So a lot of the work we had to do at the beginning is building a lot of custom integrations and interfacing. So a lot of system integrations, number one, to be able to access the data. Second, when you look at the ingest process there's a lot of specificity we had to build around leveraging different optical character recognition models, computer vision models, handwriting. A lot of these are diagrams with actual handwriting. So figuring out what's the right tool, how does that to be fine tuned to be able to ingest specific components of the ingest process. Then a part of our technological secret sauce is the sentence embedding models. We have the latest benchmarks we have with Oak Ridge, they perform like 40% better than off the shelf search embedding models. So that then helps with more of the reliability and then finally the workflow and the end user training. So understanding a worker who's in a nuclear power plant, the way they're searching for documents and the way information should look when it's retrieved and what are the different filters that can be put in place. What I'd say is we've got unbelievable AI engineers who could do research and development separately. Our product team who actually designs the user experience is also phenomenally good. So I'd say what we bring to the table is you could think of us as combination of system integrator, getting the right data, understanding AI architecture, ability to do really advanced research and development to make the technology work, then separately a really great product team that actually can get the end user adoption. And as a result, when you bring all those ingredients together, you get AI deployments at scale that are successful. So clearly having at least hundreds of daily weekly active users on the product, it's the largest generative AI solution in nuclear. There's not that many of those that have been deployed.

B (28:10)

But don't throw shade at what you've accomplished there as a tiny little startup getting a software platform integrated into an active working nuclear power plant in the United States.

C (28:22)

It gets even better though, because that's been super successful. But that's what our team did just to get off the ground as we look at the future of where this goes, you brought up retrieval, augmented generation. That is equally challenging to get done at scale when you're looking at different workflows. So a way to think about this, Cody, is the first step is search. You gotta be able to find the documents, you can't find the documents. None of your AI is going to work. As we look at what comes next with Diablo, you now build more specific modules on top of the search platform to add even more value. And in that case you have, let's say your regulatory module, you have your 5059 module, you have an engineering change or design change module. There's all of these capabilities that are necessary for specific users and workflows that. And as a result we're going to need to have multiple implementations of different large language models to then go and facilitate the workflow in the user experience. Retrieval augmented generation will be incredibly important. But today there's no benchmarking of which LLMs do great against nuclear workflows. So we announced a project this summer that we've kicked off with Idaho National Labs. So Idaho also has phenomenal nuclear experts. So we reached out to them and said, look, there's no nuclear benchmark. So when Chad GPT releases ChatGPT 5 or ChatGPT 6, you see all these press releases, you see like, oh, ChatGPT can pass the medical boards, it can do better than 90% of AP physics students. Like there's all of these like benchmarks that are available nuclear, you have none of it. When Gemini releases their next model, we have no idea is it good at nuclear, not good at nuclear. So with our friends at Idaho, we're going to build what's called nuclear bench, which we're going to make publicly available. It's a series of tests and benchmarks to understand which LLMs out the gates do really, really well against nuclear terminology. And then that helps educate us as we're designing the next phases of our friends at Diablo Canyon and other nuclear power plants. Which models should we use for specific workflows? Then how do we build all the experience around that? So the first layer of the cake took a while to get in place. But it provides both tactical benefit of hey, there's value from AI at a nuclear power plant and a stated roi. Then strategically it lays the foundation and then build all the other generative workflows that people look at. So that's kind of roadmap of where we're going. And then I could share another big project that we have that we're super excited about.

B (30:57)

On the note of I and L, we've also had Dr. John Wagner who runs on the show as well. So anyone who wants to dive into the work they're doing in nuclear, which is really, really awesome, that's also a fun one to go dig up in our archives.

C (31:08)

I'm amongst DOE National Lab royalty with the people you get on this show. That is impressive.

B (31:14)

Again, as a tiny startup, you found your way into working with all these folks. So it's really incredibly impressive what you've been doing. Trey, which leads me to, I guess my next question, which is you've ultimately decided this is a venture type of business, right? Like, is this a company that we can grow and scale and build into a ultimately a venture returnable company? You've decided yes. You've now raised a couple rounds of venture capital in the U.S. there's what, 93 total reactors? How do you view the size and scale of this business over time? And how do you view your ability to grow it? One of the biggest nuclear builders in the world is China, but I have to imagine that's probably off limits for you as a US based startup, how are you thinking about the growth and scale of Atomic Canyon?

C (32:02)

That's a super exciting question and clearly one that as CEO I think about on a daily basis. So there's a couple different ways and directions to look about the growth opportunity. So first, as we were picking our second customer, I'VE realized as an entrepreneur, having gone through this a couple times, your first customer clearly incredibly important, but your second is also really important too, because then it becomes a question of are you a one off? Yeah, you pull it off once, can you do it again? So we want to be ultra selective on where we apply our precious resources for our second project. So there's a group within Nuclear inpo, the Institute of Nuclear Power Operators. Background there is after Three Mile island, all the plants got together and said, hey, we got to start sharing best practices. Like, we got to share more detail, more information and info was created. So every nuclear power plant puts in dollars, puts in resources towards info. So it's this great, incredible data repository sharing of best practices. So INPO got together with nei, which is the Nuclear Energy Institute, Big industry association is lobbying and other kind of best practice sharing. And then epri, which I'm sure everyone who listens to your pod knows about, all got together and said, hey, we got all these nuclear data sets. We really got to start sharing them better. We got to figure out how we can use AI to get this information and make it more publicly available. They put out an RFP to see who wants to take on that business. And when my product team and nuclear team found that, they said, trey, this is the second thing we got to do. This is right down our alley. So we applied and we ended up winning that business. So that project, Info announced it about a month and a half ago. We're really working with Info and these groups to use our AI platform to make their data more available. So we're super excited about that and that leads into the immediate growth story. So the first is, hey, there's 54 plants, 94 reactors in the United States. We believe all of them are going to embrace nuclear in the next three to four years. However, that's a relatively small town. You run through 54 plants pretty quickly. So as we look at the growth opportunity, the immediate one is international. I was just in Singapore for Singapore International Energy Week. Nuclear is a hot topic across the world and there's I'd say roughly another 200ish reactors that we can go sell our current software to. I think the next phase of growth and growth trajectory for the company comes in, I'd say, two general domains. So the first is we're building nuclear power plants finally. Thank God. We've been waiting for this forever.

B (34:31)

By we, you mean United States, not Atomic Canyon. Just to be clear. Not there yet, Cody.

C (34:35)

Not there yet, no. But we as the United States As a society, we are embracing nuclear. Like, every week there's a new announcement. There was the $80 billion to Cameco and Westinghouse. We start building AP1000s. And then you see other announcements that through the tariff, Japan's gonna support $200 billion, 100 billion to Westinghouse, 100 billion to GE, Hitachi, you got SMRs popping up everywhere. You got X Energy, you got Terra Power, you name it. So as society, we're embracing and building nuclear. So our view is two opportunities in that domain. Number one, I can't announce anything, but we're very active with a number of the SMR and the OEMs. And Oak Ridge, we signed another partnership with them. How do we use that supercomputer to help streamline the licensing? Because there's a whole bunch of these reactors AP1000s can get built, but we got to build a lot more of these reactors. So how can we use AI to streamline that reactor license approval process through the nrc. Secondarily, once you get them licensed and approved, there's a whole process to build reactors. Everything from the site selection, environmental reviews. Where can we build the reactors to the actual construction process. How do we use AI to optimize a lot of the work that we've done in construction? And then finally, when they go live, we're already building a lot of the operating software to help operate. We see those two big arenas. The licensing process, getting a reactor license is tens, if not hundreds of millions of dollars of paperwork. That's all going to get streamlined away with AI. Separately, when we're designing, building, site approval, et cetera, huge, huge potential for AI to be implemented and streamline those processes. So from our perspective, the current fleet is beautiful to start at because, number one, we get credibility. We get to demonstrate we can live up to whatever promises we make. We think the INPO projects can help accelerate us across the current fleet. We'll probably sign a couple more international reactors, but then the big growth opportunity comes from us supporting the deployment of artificial intelligence to help build reactors, license reactors. And in 2026, I think you'll see some pretty exciting announcements of who we're partnering with and what it looks like. But you know, Cody, our big challenge, just to be honest, there's too much demand right now. Everybody knows they need AI and what it comes down to. More companies fail from overextending themselves than from focusing and executing. So a lot of my time is spent saying, out of all the projects we can do, what are we going to be Focused on where are we going to achieve over the next few months. It's good problems to have right now. Really good problems to have.

B (37:14)

That's awesome. Well, congrats on really building this business in an interesting way. Starting with a single problem that you went to go solve growing up from there, but starting from a point of execution and deployment. That's how you learn. That's how you build a startup.

C (37:28)

Yeah, I couldn't agree more. You can kill your startup through analysis paralysis. There is no path that magically gets laid out in front of you. You need to go, you need to take action, you need to have traction. And the goal is just to continue moving the ball forward. And over time, assuming you execute and you have a pretty good mind around overall strategy, you could figure out where the business opportunities lay. And I do think there's advantage in coming in with a fresh perspective. I mean, I'm not biased by any nuclear background. So as I enter into the space, a lot of opportunity to decide where to implement, where to focus and what that looks like. So I think that's part of the advantage we bring is we're a fresh set of eyes with no biases as we approach this space. And we've hired multiple nuclear employees now, so it's not like we don't have nuclear experts on our team. That's something we augmented pretty quickly. But yeah, we're super thrilled. We think two years, it was a bit of a leap of faith to jump into nuclear now with all the activity and what's going on, I would not want to be anywhere else. Like this is the most exciting market in the world. Got to make the call that we thought it was going to be the right place and the call was right. So now we get to really reap the benefits of being ahead of the pack when it comes to building really advanced AI that's focused on this nuclear space.

B (38:46)

Trey, anywhere, you particularly are needing help right now. For anybody listening who's excited to dive in and figure out how they can.

C (38:52)

Support you, I believe it was Ballmer who said, developers, developers, developers, developers. Hiring AI, backend, full stack, front end. We're very, very selective on the engineers we hire. But without a doubt we, we are very, very actively hiring software developers like that. That is where our focus is. So our engineering team is going to be at least doubling over the next year. That's the big opportunity. Nuclear, there's a few constraints, like you've got to be located in the U.S. ideally a U.S. citizen, because you have to go into a lot of secure closed environments in the national labs. We are hiring. So if you are amazing 1 percenter software developer and you want to work for a company that has impact, we're helping drive clean energy in the United States. We're helping to make America the leader of nuclear and artificial intelligence. Come join us. We're a great place to work, family friendly.

B (39:47)

We're the best remote employee. Right. Like you're in San Luis Obispo, but I think your team is pretty spread out.

C (39:53)

Yeah, our roots have been set in San Luis Obispo, so we're not moving. Love San Luis Obispo. It is not the center of the world for AI engineering. We'd love to hire people here, but we've had to go remote, which, you know, we make it work.

A (40:06)

Share a little bit about the funding that you've raised.

C (40:09)

So our most recent round was led by Energy Impact Partners. In May, we closed a $7 million seed round of capital. Energy Impact Partners. They're the absolute best. So again, coming from the health care industry, I didn't know landscape of who's who, so really wanted to find a lead that had a lot of background and understanding of utilities. Energy Impact Partners is backed by a number of the largest utilities in the US So they are able to make all sorts of connections, tell me what events to be at, who to be speaking with. So I could not have hand selected better investors. Also involved was commonweal. They got a lot of connections into US government and building in these kind of tougher spaces. And then Tower Research Ventures participated. They're a hedge fund. As a venture firm, not a lot of people have heard of them. Why I brought them on if you want to. Who's the best at AI by far is hedge funds. Hedge funds have the best AI teams, so they actually bring us pretty good pipeline of developers. And then separately from that, Wischof Ventures, if you know Nicole, she's amazing. They participate as well. So yeah, it was a great group. We're super excited. Love my investors. They've been phenomenal to work with.

B (41:22)

Trey, really appreciate you making the time to join. Any final thoughts you want to share, what you're working on or advice for anybody interested in the space or whatever else comes to mind in closing remarks.

C (41:33)

Number one, Cody, thanks for having me on. Big fan of what you're doing. And when we look at nuclear is going to take an army of people to build this out. Like we're in a position where it's becoming pretty clear that nuclear is going to be one of the many energy solutions that are deployed as we're going through this period of rapid, rapid expansion of energy needs. However, nuclear's challenged because we don't have enough people. Like, if you look across the board, we just don't have enough nuclear engineers, like, to go build $80 billion in AP1000. Like, that's great. Who's going to design them? Who's going to engineer them? Who's going to build them? The only way we're going to succeed as an industry is through the augmentation of our work staff with artificial intelligence. So I can't think of a better opportunity, a better industry to be part of. So just thrilled to be here and we're having way more fun than I would have ever expected. But thanks for the time and we'll see where we're at in the next couple of years.

B (42:29)

Thanks, Trey. Appreciate you joining.

C (42:31)

All right, take care, Cody. See you later.

A (42:33)

Inevitable is an MCJ Podcast. At mcj, we back founders driving the transition of energy and industry and solving the inevitable impacts of of climate change. If you'd like to learn more about mcj, visit us at MCJ VC and subscribe to our weekly newsletter at Newsletter MCJ vc. Thanks and see you next episode.