Michael Crabb [00:00:59] Hi, this is Michael Crabb and I'm joined today by John Chaplin. As I think our listeners know, usually Titans of Nuclear is a feature of folks who are in the nuclear industry; their backgrounds, their experiences, what they do. Today, we're starting something a little bit different: our first Nuclear Nexus series. And this one will be focused on a really timely subject: data centers. I'm kind of excited to dig in.

John Chaplin [00:01:25] Yeah, it's kind of hard to pull ourselves out of the nuclear world and realize we're adding solutions to the marketplace. What are the best ways that that fits in? I think that's a good subject matter to dive down in on. Data centers are a perfect match for nuclear, and yeah, looking forward to the conversation.

Michael Crabb [00:01:42] Yeah, and it's funny... I'm sure your parents are the same, right? Putting something on the cloud. As a new parent, I think I've got 4,000 pictures over the year that are sitting on the cloud. But the cloud's not actually in the cloud; it's infrastructure that's built in the real world and has real requirements and real implication on the environment and the community that we're in. It's the reason we're able to communicate to our audience right now. And so, I think the intersection of how nuclear power can impact the cloud and the data centers that power that and enable it is really quite interesting. We'll see.

John Chaplin [00:02:21] Yeah. And I'm looking forward to, throughout this series, diving down into experts, learning more about the applications where nuclear fits into this data center space, and how it can be leveraged to really expand the infrastructure profile, and also, facilitate the new AI development that's really driving the growth.

Michael Crabb [00:02:39] Yeah, that's a great transition to the scale of the conversation, right? So, we'll have a host of different folks who are from the nuclear space, from the data center space, the development space, perhaps even the political and regulatory space talking about how these two industries can come together. And it's the scale of both of these things that is so tremendous, right? I think data center power, globally, would be a top 20 energy consumer in the world, from a country perspective.

John Chaplin [00:03:05] And that's today, compared to where it's going to go with AI having a 10 to 20 fold increase in that power consumption to be able to run all this computing for these training of these models.

Michael Crabb [00:03:16] Yeah. No, it's really crazy. I mean, I think some of the stats that our team pulled together are worth emphasizing. In 2022, data centers used 300 terawatt hours. So, about 1% of total global energy consumption. In the UK alone, they're 3% of the UK energy use. And as you said, that's just the tip of the iceberg, right? We're going to use 10x more moving forward.

Michael Crabb [00:03:43] Okay, so what are the some of the things that we need to focus on, I guess, as we talk to these experts? What can folks expect?

John Chaplin [00:03:50] I think really tackling and understanding how nuclear can really come in and solve those issues. The immediate issues, but also the long-term issues as it scales out. Near term, there are a lot of issues around power sustainability, being able to get that carbon-free power, but also just grid constraints and being able to meet the needs of this growing demand that's coming from this industry.

Michael Crabb [00:04:14] Yeah, I think the grid constraint one is one that isn't fully understood in the nuclear space, right? It's not really a problem that we have or have dealt with and mass, but more and more I'm hearing that's really driving data center location and data center developments. And there's been a few recent announcements with that intersection, right? The Amazon acquisition of the Talen data center, outside of their nuclear facility. And that's just an example of, "Okay, well, there's an existing power generation unit. How can we locate near them in order to benefit from the power?" You know, Microsoft, Google, many of these folks have similar campaigns. Tell us more. What do you mean when we say "grid constraints?"

John Chaplin [00:04:57] I think legacy nuclear that's large build has always been long-term planning and ability to basically connect to the grid their full load; they're able to plan that. As we move to more distributed generation with small modular reactors or micro modular reactors, the ability to be able to connect to the grid becomes a larger issue that can't be planned. And you want to be able to get to those grid connections quickly. And with renewables coming in, you have tons of different energy generation sources trying to connect to the grid all at the same time, and that causes a long backlog.

Michael Crabb [00:05:28] Yeah, and it's really both sides of the equation, right? Us on the supply side, and in the nuclear space, you hear about the supply issues, right? Generators trying to connect, but load has to connect as well in this sort of complex transmission system. And both parties are frustrated, because I think the bottom line is there just isn't enough capacity, right? There isn't enough space in the straw, so to speak, to have as many electrons as are needed.

John Chaplin [00:05:54] Right, and I think that's where nuclear can come in, because it can do behind-the-meter solutions, connect directly to that power consumer, especially those data centers, and be able to facilitate their growth while they're still waiting for that grid connection.

Michael Crabb [00:06:07] Yeah, totally. And it's cool that we're seeing that actually in practice, right? Like literally, there are running data centers that use nuclear power. And so, excited to scale that together.

John Chaplin [00:06:18] Both seeing it in practice and also seeing... A lot of the hyperscalers realize that this is a significant strategy that they need to deploy and are building their own internal teams to evaluate and prosecute that strategy.

Michael Crabb [00:06:28] Yeah, absolutely. Okay, well, that's grid connectivity. We'll have real experts talking about that over the coming weeks. There are some other really key value propositions, I think, at the intersection of both nuclear and data centers. I think a key one is price volatility, right? Or, price reliability in this case.

John Chaplin [00:06:44] Right, exactly. When you're thinking about data center development and the economics of their development, power consumption is a large cog to what they're selling. So they need to have that, one, reliability of what their cost is going to be. So, under long-term power purchase agreements, you might be able to get that from solar or wind, but since they are intermittent resources, you still have to understand how to balance that and manage that price risk for when those technologies aren't producing. But with nuclear being that physical baseload power source, you're able to really get that consistent profile.

Michael Crabb [00:07:17] Yeah, fixed cost is really important. And I've sort of observed that there's almost wrong-way risk when you've contracted a low price for renewables. Or, when renewables are on, they tend to all be on and market price is really low. And then when they're off, your PPA does nothing for you, but energy prices go through the roof. And so, you're sort of paying too much in both circumstances, and nuclear can really create a fixed profile for consumers who, as you said, a large part of their customer wallet is energy.

John Chaplin [00:07:47] Yeah. And with the changing landscape in generation sources, connection backlogs, understanding how that's going to affect the market is nearly impossible. So, the volatility in the marketplace is only going to become more extreme as we move forward.

Michael Crabb [00:08:01] Absolutely. And then, we can layer on top the desire to be sustainable on top of all of this, right? There's really a global recognition of nuclear's role to play in a sustainable future. And the way I see it, there really is no other option.

John Chaplin [00:08:21] No, I mean, we've touched on some of the reliability and security, but also being able to have that consistent power load. When you're running these models, some of them can take weeks or months to train. If you lose power and you stop running that model, that's loss of time, that's loss of revenue. Those are costs out the door you're not going to get back.

Michael Crabb [00:08:41] Yeah. Well, and the reliability concern is obviously a key focus for data centers even outside of the AI stuff. Like, Instagram went down for two hours and everyone freaked out, right?

John Chaplin [00:08:55] Right. Yeah, with data centers, even for just cloud services, I think I've heard some of these hyperscalers say that reliability is their main concern. If you go to Google and the search bar isn't there, that is an absolute critical importance to Google. It's always got to be up and it always has to be available. So, having these data centers rely on intermittent resources that might not be available when they need them is a critical risk that they have to manage.

Michael Crabb [00:09:21] Yeah, and I'm kind of excited to dig in with some of these folks in the field doing this work. Reliability levels are like... Being reliable means different things to different people, right? And we're in the nuclear industry, very proud of 94%, 95%, 96% uptime and reliability. But that's not 99.999%.

John Chaplin [00:09:41] Right. And there are always strategies on the back end. You hear about data centers adding n+1, n+2 redundancy and having different strategies for how they manage that. And I think it's always going to be a mix of different energy sources and different solutions, but nuclear stands to really benefit be a large leg of that solution.

Michael Crabb [00:09:59] Yeah, really the anchor, right? I mean, you can get 95% from one solution and then you can fill in the gap. And you really were touching on size as well. That's maybe the final aspect in both the data center market and the nuclear space. And they're going in opposite directions, but I think they're meeting in the middle.

John Chaplin [00:10:17] Yeah, today some cloud services have been 10 to 60, maybe even 100 megawatt sized for power consumption to be able to run those. But with the advent of AI, those are increasing by multiples. Now we're seeing 400, 600, 800 megawatt data centers being developed. That is a large power drain on the grid and requires a lot of planning. So I think there are definitely power sources like SMRs that can come in and meet that demand.

Michael Crabb [00:10:46] Yeah. And I just think there's something poetic about historically focusing on 1,200 megawatt nuclear power plants and 10 megawatt data centers. And yeah, we'll meet somewhere in the middle at 500 with and n+5 reliability as our fleet of nuclear power plants. Yeah, it's really exciting. And I think nuclear's ability to really be site agnostic is really important as well. Because data centers have requirements around latency, around proximity to networks, to network gateways, so it'll be really interesting to see how the different business models evolve.

John Chaplin [00:11:24] Yeah, absolutely. Especially that's true for cloud. With AI, they have the ability to be a little bit more flexible because these are really spent training and doing large compute without the need to connect to the network. But as we see more of that build out, we'll see these large, distributed generation sources, whether that's AI or continuing to build the cloud infrastructure in large demand areas like major cities.

Michael Crabb [00:11:50] Yeah. It's going to be really interesting to hear how these concepts are being implemented in real life. Yeah, we look forward to sharing some of those stories. I can't wait to hear from the experts.

John Chaplin [00:12:03] Yeah, me neither. I think the other thing that we didn't really touch on is the de-carb goals. We hear a lot from data centers, and I think the tech sector has been a leader in pioneering solar, pioneering wind. And I think they're going to be a leader in pioneering nuclear. They're really pushing for the market demand, that 24/7 time match. To date, we've been trading credits, and solar and wind has been able to trade some of those renewable energy credits, during when they're all producing, say, solar in the middle of the day, to make people have a clean profile throughout the day. But if you really are going to push for a requirement for 24 hour, nuclear is really the one that stands to benefit the most.

Michael Crabb [00:12:44] Yeah, at the time, everyone knew RECs was sort of an interim goal, right? And then, we got really excited about it and we forgot it was an interim goal. And then, we realized we really weren't doing what we said we set out to do. And so, yeah, it's going to be exciting to see how that pushes the data centers and others towards nuclear.

John Chaplin [00:13:04] Absolutely.

Michael Crabb [00:13:04] That wraps up our first episode of The Nuclear Nexus for data centers. A really relevant topic; excited to hear from the experts. And we'll have some other conversations in similar veins at a couple of key conferences coming up. At Data Center World and Datacloud ESG Summit. I'm sure there'll be a few more as this starts to gain steam. Looking forward to sharing those topics and more with the audience in the future.

‍