Artificial Intelligence and Nuclear Energy

00:00:03:20 - 00:00:28:04

Chris Keefer

Welcome back to Decouple for a very special I was going to say live, but definitely not live but in person in-person interview featuring my good friend the JP 1000 and the can do kid. Joking aside, James Krellenstein, and myself. James, welcome to the show. Thank you. Thank you for having me at your house. And and the show is going to be a banger.

00:00:28:06 - 00:00:53:00

Chris Keefer

We are going to be tackling the issue of AI and its implications on energy, of course, focused on nuclear. This is a emerging area and I'm really excited. I feel a bit behind the eight ball getting to it and covering it. But who better to explore it with than my friend James here? I've got a lot of thoughts and ideas down and I'm going to attempt to do a little bit of an introduction to sort of ground where we're going with this.

00:00:53:00 - 00:01:19:10

Chris Keefer

But as things classically go in decouple, we're going to go off on some tangents in the weeds, but basically I view this situation. I mean, it seems to have come on us as it is a bit of a surprise. And even the developer, Sam Altman, recently saying that nothing short of an energy revolution would be required in order to power this this new phenomenon, the emergence of a baseload demand, something that needs to be power.

00:01:19:10 - 00:01:39:16

Chris Keefer

24 seven in a set of societies in the West where we have bends or fragile lives in the grid and emphasizing baseload or even saying it's a myth how this impacts on climate action. We have a recent report out from Jp morgan called Electric Vision, which talks about, you know, we've added a bunch of capacity in terms of clean energy.

00:01:39:17 - 00:02:00:13

Chris Keefer

Fossil fuels continue to grow, but we have James might dispute this slightly, but we've failed to really electrify certainly at the scale required for these ambitious or frankly other people might say unrealistic net zero goals. So we're going to sort of tackle that. I think we're going to try and get to what are some of the limits to growth for AI.

00:02:00:15 - 00:02:29:19

Chris Keefer

You know, it's fascinating to me that demand in the West certainly has been pretty flat for quite a while. You know why that is and whether demand with AI is just going to continue linearly. I think that's a common human bias when thinking of the future that trends will continue linearly. And that was certainly the mistaken thought of utilities in the seventies and eighties, I guess up into the seventies when we were adding 4 to 5% per year of of growth to the grid.

00:02:29:21 - 00:02:52:04

Chris Keefer

All right. I mean, I'd originally conceived of this as a sort of a the risk of kind of energy apartheid. I think we're going to go a little bit into what this, you know, emerging massive addition of demand is going to mean for everyday people and human needs. And talk about some of the positives and potential negatives of AI.

00:02:52:04 - 00:03:10:21

Chris Keefer

That's that's kind of way too big to go into scope. But we'll try and stay focused. So James, a long intro there, but I just needed to gather my thoughts because I haven't been doing enough thinking about this. So again, warm welcome back.

00:03:10:23 - 00:03:54:00

James Krellenstein

Thanks. Thanks so much for having me. And for having me in your home. And for me just came back from Bruce Nuclear Station two days ago. I saw a can do in person a couple of days before and so, you know, I think when we're thinking about computation and computing, we don't necessarily think about computing energetically class. We think of a lot of other things. But, you know, computing itself just thermodynamically does require energy. And there's something called Landauer Principle or Landauer limit, which basically says that for a minimum, for any amount of computation, the second law of thermodynamics says a minimum amount of energy needs to actually be put in to actually do that computation and that that you just can't get around that first principle, second law of thermodynamics.

00:03:54:00 - 00:04:25:13

James Krellenstein

Now, we should be clear that we're about a billion fold away in terms of what that theoretical limit is and versus how much actual energy are now spending to do a bit a unit of compute. But what has been happening over the last decade or so is we have been seeing an enormous growth in not just computation in power that's in our devices that computers or phones or embedded devices that surround us, but also a huge amount that has been migrating to the, quote unquote, cloud right on right.

00:04:25:14 - 00:04:47:06

James Krellenstein

Than having the computing done locally, having actually the majority of the storage out. It's some big data center and then transported to you through the Internet or through fiber optics and, you know, whatever the last mile is copper, but it's all being stored at a big data center. And we have seen compute use go up a lot over the past decade.

00:04:47:08 - 00:05:26:21

James Krellenstein

But we've had also the amount of energy per unit compute decrease, and this is called Comey's law right after the Stanford professor, Professor Comey, who noted in 2010 or 2011 that it seems like about every 16 to 18 months the amount of power that it takes per unit of compute. Harvey's right. So what has been going on simultaneously over the last decade Now, to me, in 24 months, maybe as Moore's Law starts to fail or we've started to see is is that yeah, compute load has been historically increasing the amount of compute per capita I don't want to talk about it has been increasing, but the actual amount of energy per unit compute has been

00:05:26:21 - 00:05:54:10

James Krellenstein

decreasing at roughly the same rate. So we've seen sort of running in place, right? We've been doing more compute, but being more efficient. And then so therefore the actual amount of energy used for datacenters has been relatively stable. Now we should be clear that from a baseline the amount of energy use in terms of global warming impact, let's say, you know, it's been estimated that a minimum of 3.5% of humanity's GHG or global greenhouse gas emissions are from datacentre based operations.

00:05:54:11 - 00:06:20:19

James Krellenstein

So that, according to these estimates, is more than shipping or more than aviation not combined single handedly. That's just how much power is being utilized for for compute just in the regular status quo. And, you know, Amazon reports itself has to be one of the world's largest consumers of electric power purchases of electric power. Just because Amazon, you know, with eight of us is running such large data center operations.

00:06:20:21 - 00:06:52:13

James Krellenstein

Now, what has happened over the last 24 months, maybe a little bit. Sure. Is that we start to see the air boom occur. And in particular in these large language models, more more, more broadly, neural network based computing, rather than that which is requiring a totally different architecture of compute. We've really gone from the CPU base scale scalar computing to really having a lot of these neural networks running on these large like graphical processing units.

00:06:52:13 - 00:07:19:16

James Krellenstein

They were originally historically GPUs like by NVIDIA or AMD, but mainly NVIDIA. And that has there's two things that are going on. One is that we've lost a little bit, although we're probably better, we're better now on compute efficiency, but more importantly and more broadly, the amount of computation that we need to be doing to do an AI based task versus a classic non eyeball based task is just so much dramatically more.

00:07:19:18 - 00:07:50:06

James Krellenstein

It's estimated, for example, that a Google search or a good classical Google search, right, takes about 20 times less power than an AI enabled Google search. Right. And these are external estimates, peer reviewed. That is a massive step change in the amount of energy that we're going to need to do for the same task. And that's partially driven by I mean, you can pause there and talk about what is driving that.

00:07:50:06 - 00:08:12:03

James Krellenstein

But what we're seeing is, is just AI is accelerating the amount of power that we need, right. To do these classical tasks. So, I mean, you're talking about A.I. requiring 20 times the power for for Google search. But once again, you know, the Google search might be much more useful and much, much more utility. I mean, a Google search.

00:08:12:03 - 00:08:36:03

James Krellenstein

The weird thing to compare it to your chatbot query that you just normally do is taking about ten times the amount of power that a Google searches. And that's just giving you the example of the step change that we're having in terms of compute use that's translating to a massive increase in power use. And I mean, the cynic in me, I was just looking up some stats of 20% of all Internet searches are for porn.

00:08:36:05 - 00:08:59:06

Chris Keefer

I'm not sure what percent is consumed by TV cats. And obviously I'm being a cynic here. There's a lot of highly useful application trends and I think people are talking about the potential for, again, a real productivity step change with the applications of AI. And I think we'll probably get into kind of what those are later. But yeah, with this emerging increase in baseload demand, let's talk a little bit about that.

00:08:59:06 - 00:09:22:12

Chris Keefer

You've talked a bit about the scale, but how is it being met? I mean, we heard about this Susquehanna deal with Amazon. The tech companies had been, I think, a little reserved about using nuclear energy up until maybe they're being forced to now because of the nature of their demand and the scale of their growth. But, you know, Apple has claimed to be 100% powered by renewable energy in the past.

00:09:22:17 - 00:09:43:06

Chris Keefer

A bit of a disingenuous claim in my humble opinion. But but where are we going with this? Because, again, I mean, this podcast focuses probably a little too much on nuclear. We're going to get there. How are things currently really being powered and what sort of conversations, well, are emerging in terms of, you know, and how panicked are they in terms of meeting this burden?

00:09:43:06 - 00:10:21:09

James Krellenstein

So I think there's a couple of things going on. I mean, we really have been seeing large, large growth in electric power used for data centers in, for example, the state of Virginia, which in the United States has the most data centers in for right now at least. This is really, you know, the air revolution. You know, while the chips are made in Taiwan or at TSMC or maybe a little bit at Samsung, but almost all at TSMC, you know, the the air revolution right now is being run by us companies primarily, and it is using US technology designed by NVIDIA primarily and at US data centers in Virginia has historically due to the co-location

00:10:21:09 - 00:10:46:19

James Krellenstein

of where sort of the ARPANET, which was the predecessor of the Internet, was the Department of Defense sort of program, we still see actually a huge domination based on sort of interconnected connections of data centers in Virginia. And, you know, just over from 2020 to 2022. Right. We saw power use jump from around 116 terawatt hours in Virginia to about 133 terawatt hours.

00:10:46:21 - 00:11:17:06

James Krellenstein

So that's like more than 15 terawatt hours. And growth just in that one region alone. Almost all of that is dominated by the increase in data centers. So this load growth that we're starting to see being driven by air, it's not a hypothetical thing that we're like looking to the for it is happening now. And I think this is this is this is sort of coalescing with two other sort of macro forces that are going on, which is that the huge amounts of policy in the US, we're not electrifying.

00:11:17:06 - 00:11:40:22

James Krellenstein

Maybe as fast as you thought. We are. We definitely are seeing increased electric vehicle power demand, power charging demand, which I think is going to continue to grow, maybe not as fast as some people initially thought, but it definitely is happening. I think one of the things that we have to be really interested that what could be a game changer is the proliferation of cheap Chinese sort of electric vehicles that are being mass produced at much lower prices than the American others.

00:11:41:00 - 00:12:02:02

James Krellenstein

They start really having a significant market share here in the United States. We could start seeing a real take off in electric vehicles. And we are starting to see also heat pumps really become more and more of a method of heating in the United States, particularly in the northeastern United States, like cities like New York City have fossil fuel bands.

00:12:02:02 - 00:12:28:22

James Krellenstein

Right. You can't build new housing with non-electric heat with non heat pump based heat. So we're starting to see a combination of these these triple electrification. And this is really changing the world now because for, as you were saying, for really decades in some ways we've been watching power efficiency, right? Really cheap load growth stable even as actual economic activity dramatically increases.

00:12:28:22 - 00:12:49:15

James Krellenstein

And we don't have to look, we just look around this room to see that that changed. If you look at what a television one flies with a cathode ray tube based TV, we replace that with LCD and flat screens. We really reduce the power use of that dramatically. Lighting has gone through a revolution with the proliferation of light emitting diode based LED based lighting.

00:12:49:17 - 00:13:19:15

James Krellenstein

Now we sort of had all those efficiency gains sort of now baked in, and now we're seeing on the electric power side just multiple drivers of electric power use that in some cases are enabled by large amounts of government policy. In the case of straight commercial demand. And that is causing for the first time, I think, in a really long time utilities are trying to figure out and maybe in some cases struggling how they're going to meet load in a lot of regions.

00:13:19:17 - 00:13:52:11

James Krellenstein

And that's a very big difference than where we were 30 years ago or 20 years ago, or even in the first nuclear renaissance where load still was relatively flat in terms of loan growth. Right? I mean, one could say that's certainly not an absolute decoupling, but there's been a partial decoupling of economic growth from electricity use. You know, there's been a certain amount of Jevons Paradox there, and maybe that's covered over in terms of the flat demand by the offshoring of a lot of heavy industry, which I understand used to be a pretty big source of low growth, certainly globally.

00:13:52:11 - 00:14:12:05

Chris Keefer

You know, power use is increasing. I think obviously because of the economic development moving at a breakneck speed in China and India, for example. So, you know, power demand globally has been going up. But you're right. I mean, this is this has been this flattening. I'm thinking a lot of we should be clear, like it's a flip flattening of maybe electric power.

00:14:12:07 - 00:14:38:03

James Krellenstein

And even even if we talk about total energy, most energy still is not consumed as in the form. And its end use is electricity. Right. What's going on in the electrification sort of revolution that may or may not be happening? Yeah, but but it definitely is there. Definitely. We are seeing more and more people want space heating and water heating, which is a major a large portion of the residential energy use historically has not been in the United States.

00:14:38:03 - 00:15:03:12

James Krellenstein

Really electrical, I think like for space heating, it's like more than 90% of space heating has historically been non electrical in nature, just like to use. And now that's transitioning more and more as as heat pumps start dropping in price. And as they're encouraged, we're electrifying that. So even though a total energy demand might be going down or be stable, the electricity share of that pie is growing and growing.

00:15:03:18 - 00:15:25:08

Chris Keefer

Right. Ideally, I'm thinking a little bit about the soft path from Amory Lovins, the idea of megawatts and really some of his ideas went against the concept of electrification because that's inefficient, certainly in terms of in terms of thermal power generation. You know, from pure thermodynamics perspective, you're losing 60, sometimes even 70% of the energy that's put in.

00:15:25:10 - 00:15:52:08

Chris Keefer

And I'm thinking about places which have been conducive to electrification so far, and those are places I think there really followed a hard path. And I think about Quebec with its massive investments in its hydroelectric infrastructure, its ability to export. I mean, up until recently with the droughts that have really impacted the basins for these hydroelectric facilities, a surplus of power, and certainly the the Northeast states in the US really starting to depend more and more on that for clean energy.

00:15:52:10 - 00:16:14:15

Chris Keefer

But are we going to be experiencing a return to more of the kind of hard power path or off that hard path in terms of not only being difficult but needing to have hard power? Yeah, so it should be clear a couple of things, right? One, you know, the thermodynamic thing that you just talked about, you know, that's of concurrency to work or heat to get rid of or just most heat can't be convert.

00:16:14:15 - 00:16:41:00

James Krellenstein

I mean, unless you're at a really high country heat source can't be converted to useful work. So like one of the things that's cool about electricity, like you you it mechanical work, like you lose at a nuclear power plant, you're losing, you know, 66% or slightly more of the heat is being wasted, quote unquote. But you can't really actually ever convert that heat into useful work just because the second law and just general practical efficiency, you know, internal combustion engines, 20 to 30%.

00:16:41:06 - 00:17:13:16

James Krellenstein

But what is cool is, is that the you know, the turbine is making that into mechanical work and you lose only a very little amount in the actual generator side of things. Right. The actual conversion, mechanical work, it's electrical power and then back is pretty efficient. So I think electrification does make a lot of sense for thermodynamics. The edge cases on heating, we're talking about work or just somewhat thermal stuff, but that's what the heat pump is enabling of to pump heat against the thermal gradient and therefore you get a coefficient of performance depending on your two of three or four, meaning that for every one watt of electricity you're putting in, you're getting four

00:17:13:16 - 00:17:40:06

James Krellenstein

watts of three watts of heat back. Right. Which is great. So I, I, I'm not sure itself I don't think there's actually I don't think there's ever a soft path that would have ever happened. I don't think that people were really ever willing to I have never seen people ever really willing to change their lives in a meaningful way to reduce car like to reduce carbon emissions.

00:17:40:08 - 00:18:02:18

James Krellenstein

And like my example of this, always with 2020, right. We saw during the COVID lockdowns. Right. Which were devastating and had huge amounts of social implication or miserable. Thomas Anyone who abided by them and total global carbon emissions dropped by less than 5%. Right? Right. And that was like way beyond what we'll ever have social license to do.

00:18:02:19 - 00:18:23:00

Chris Keefer

We're not going to be able to do a soft path to decarbonization. What is referring to in the soft path is some of Emery's ideas, again, about because of that, I don't really know Emery's stuff. Yeah, that thermal efficiency was like he did was like, you know, we should do more sort of small, as beautiful, localized, and yeah, maybe we should just have coal burners in apartment buildings to provide seats that's more efficient.

00:18:23:00 - 00:18:44:09

Chris Keefer

Anyway, I don't want to get to say, Yeah, sorry, we got sidetracked off the air. Yeah. What I'm going to say is I think this is short. TLDR is this is a very good secular trend for nuclear power and maybe even for the general for a clean energy revolution. I would say that we need that debate about right.

00:18:44:09 - 00:19:24:10

James Krellenstein

But this is driving actual growth. That growth can hypothetically be used to power further decarbonization, if we want to call it, or it could not be. That's the real question that we have to ask here is, is that what's going on? I don't think there's any world in which we want the power to go out. And so the real question is going to be if the air revolution, this electrification revolution that we're seeing happening, is it going to increase carbon emissions dramatically or are we going to be able to your policy and through actual the economy, actually drive this to real sort of decarbonization and further out clean energy infrastructure?

00:19:24:10 - 00:19:50:15

Chris Keefer

Best question. Yeah. And again, I think I'm a bit of a cynic on the side of I don't really see climate leading to I mean, as leading to a lot of investment in renewable energy. We've added capacity and we have brought emissions down, I think modestly in a number of areas through fuel sparing mostly, but in terms of certainly as a motivator to build nuclear, I find it insufficient because of just how hard nuclear is.

00:19:50:17 - 00:20:12:18

Chris Keefer

People build nuclear for really pragmatic reasons, primarily, I think over its history for energy security reasons. But I do find that the simply the dollars and cents of it, these big tech companies are filthy rich. The kind of valuations of tech talk, for instance, I think it was like nine or 10 billion and all LNG exports from the US were like 23 billion.

00:20:12:18 - 00:20:42:04

Chris Keefer

Those is the recent static came across and it's just the energetics you know, like energy is required to power servers. That makes it possible. We've got some weird valuations but all that to say the tech companies have shit tons of money and if we're competing for creating power for particularly hard to decarbonize sectors versus building good data farms which you know, powerful people in this country and our societies want, I see one trumping the other in terms of the implications, but I'm willing to be proven.

00:20:42:05 - 00:21:11:05

James Krellenstein

So I'm not sure. We don't know yet. I mean, it's just happening. I think there is the companies, at least hypothetically, have our clean energy targets. And, you know, historically, these companies have done renewable energy credits, right. But we are starting to see them matching for 24 seven clean power. Right. And the question and we are seeing this in the Susquehanna deal, and I bet you met and Google and Microsoft not far behind the harm buying power from nuclear power plants.

00:21:11:06 - 00:21:52:11

James Krellenstein

Right. Because the nuclear power plant, of course, as I think this audience would know, is a really good source of 24 seven highly reliable baseload generation that could power something like a data center. And going back to the I sort of think one of the more interesting things that we've been learning about my company, about what's going on in the air revolution is that we generally think of the power load, the highest compute load is going to be what we call training the data, the large language models that actually make up GPT four or are or you know, the other models or a sports bar of what's Google's model from its name when it's getting

00:21:52:12 - 00:22:22:05

James Krellenstein

all the trouble but the that training takes a huge amount of energy to do and that training can really be what we're doing in training. Just so that people know is that we're basically making the model weights that are necessary to actually build these large, large neural networks of power, these large language models. The the and that's historically been a driver of much of the power use that we're seeing out of these AI, out of the amount of energy that excuse me that AI is taking.

00:22:22:06 - 00:22:50:04

James Krellenstein

What's happening though, now is this if you think about how you do an AI model, it's like, now you're going to generally have inference, which is what you're querying. GPT four right? That's like, you know, like write me a song about, you know, nuclear power and it makes you some corny song about, about nuclear power. I mean, but then the actual training is the actual building of that, the ingesting of that massive corpus of data to make that, to teach that how to do that.

00:22:50:06 - 00:23:13:01

James Krellenstein

What's happening is as we're starting to see the dramatic increase in the utilization of of large language models and other A.I. systems, we're starting to see that even though the amount per query is much less for inference based tasks, does, if so many queries that are coming up that we're trying to equalize the amount of energy used from training versus inference.

00:23:13:03 - 00:23:37:18

James Krellenstein

And the thing about inference is if you talk to a large data center APC, what they'll tell you is that they need to have these companies are targeting a couple of milliseconds for the quality of service for the latency between when a user queries a AI based data center and when it responds, and that limits where those data centers can be locate interesting by the speed of light.

00:23:37:20 - 00:23:56:00

James Krellenstein

So if you think about a five millisecond or two millisecond US target, which is one of the more ambitious excuse me, upper targets of what we're seeing, that means it has to be in a couple hundred kilometers from where the user is, just because you can't actually communicate faster than the speed of light. Now, what are these? We don't really exactly understand.

00:23:56:00 - 00:24:20:14

James Krellenstein

It's a couple of milliseconds is pretty imperceptible. We're not exactly sure what the future applications of AI that people are envisioning that are causing these requirements, but this is causing for a pretty big training. You could do. You could train halfway across the world. Right. And, you know, and transport the model waits over a couple of hours and speed up on your data center and speed up a new node that's running this model.

00:24:20:16 - 00:24:44:21

James Krellenstein

But in in terms of actually doing inference, we are starting to see the need for it to be co-locate co localize or closely localized to where the actual user is. And this is also going to be happening as we start seeing more multimodal AI models that are not just generating text but sound and video. It's just like a like a sedan, but we don't have an edge.

00:24:44:21 - 00:25:06:00

James Krellenstein

Servers close to where the kind of distribution network, you know, to where the user base is, which means that we're not we're having a double challenge now is a huge amount of energy consumption, but we have to localize it more than we would normally think we would have to. And this is totally fascinating and new to me. I did not understand this localization requirement.

00:25:06:00 - 00:25:24:21

Chris Keefer

It seems ridiculous for 4 milliseconds. I think it has be. I that's what they're targeting and I think it has to do with what they're imagining AI is going to be doing couple of years. Sure, sure, sure. Because you think about energy intensive industries like aluminum smelters, right? They co-locate next to big hydro dams, I mean, or big coal plants sometimes.

00:25:24:23 - 00:25:44:00

Chris Keefer

Iceland, for instance, I've heard, had a lot of server farm growth because of its abundance baseload generation due to geothermal, but mostly mostly hydro. And I think I remember you actually saying this one of the big rationales for nuclear in the early days was you can put it anywhere, right? I mean, if you have a water source, let me say that.

00:25:44:00 - 00:26:06:13

Chris Keefer

I mean, there's air cooling and things like that, but it's much more flexible and you don't need a rail line shipping in massive amounts of coal, for instance, or you don't require the massive hydroelectric resources. Exactly. So, again, this is another reason why nuclear seems to be an interesting match. And I mean, again, from a ethical, moral perspective, I am really concerned we're going to get to this, I guess, from a values perspective.

00:26:06:19 - 00:26:27:20

Chris Keefer

You know, if we're feeling a squeeze on demand growth and on power and that leads to higher prices, who loses out and who does the power go to the tech billionaires and their interests or bond poor? But that's actually a sidetrack. I don't want to get there. Let's talk a little bit about this Susquehanna deal. This is the co-location of a server farm.

00:26:27:20 - 00:26:54:20

Chris Keefer

I think presently 160 megawatts planning to go up to 960. The a lot of criticism because I think Jesse Jenkins, for instance, and other commentators are saying this actually isn't doing anything for climate change, is just adding a baseload demand. And a lot of people are saying also applying, I think additionality, there is no additionality existing. So yeah, at this rate, I mean, like I mean, just laying out like that plant was running runs at 96% capacity factor on average.

00:26:54:20 - 00:27:21:02

James Krellenstein

Right. It's a really great performer. I think you look at a review a year ago, something like 88% capacity factor because you're too big, too big boiling water reactors in Luzerne County, Pennsylvania, on the Susquehanna River, of course. And the idea is, is that that plant has been operating for decades and it's serving load right now. And if you just add a new load without any new capacity, that's not really doing anything from a climate change.

00:27:21:02 - 00:27:40:20

James Krellenstein

I think you heard some nuclear advocates say, well, this will make sure that the plant keeps online. Now, on the other hand, we had the Inflation Reduction Act, which are already giving a production tax credit to nuclear generating stations. I don't know. It's 2032. I don't think Susquehanna was ever going to go out of business. And we already are hearing talks about Three Mile Island unit number one being reactivated.

00:27:40:20 - 00:28:04:15

James Krellenstein

We're obviously seeing Palisades already being in the process of being reactivated. I think the Biden administration to their credit, is really trying as hard as possible to get the plants that we have bought back, raise them from the dead. But going back, it is absolutely true that from a climate change perspective, this is not on net probably it's just it's just a new load on the grid and other people are going to be using dirty electricity and no one else.

00:28:04:17 - 00:28:33:18

James Krellenstein

They're not like turning off the lights somewhere else to compensate for it. So it doesn't it's not doing anything for climate change except increasing electricity use. And that use that subsequent is not serving will be served by other presumably fossil or hopefully renewable assets or something, but it's not going to do anything. Now I have the Bulls case for nuclear though, is this only 90 gigawatts or 93, 94 gigawatts, depending on how much we're kind of Vogel in this right of nuclear capacity in the United States.

00:28:33:20 - 00:28:56:03

James Krellenstein

And that means that eventually the tech billionaires are going to run out of data of nuclear power plants to co-locate their data centers up. And that's a really good thing for nuclear power if you're in the nuclear industry, right? This is what we've been looking for in some ways, because we have what we have is demand and we have direct off takers who have very deep pockets.

00:28:56:03 - 00:29:37:13

James Krellenstein

Right. And who have been interested in nuclear. So I view this in the long term as a very good secular trend for nuclear. That may actually be the real nuclear revolution that we've been waiting for. And if we could couple this with the need to possibly make these data centers closer to the population bases, there's a lot of places in the Northeast, United States, for example, that where where hopefully we're not going to be able to, you know, just ship gas up to it because there's a huge prohibition in many places for building new gas, like gas pipeline capacity that necessarily, you can argue with the merits of that might not be a good thing.

00:29:37:14 - 00:29:56:00

James Krellenstein

It might be a good thing. I probably it's more mixed than I think a lot of people think. But it's certainly going to be a how are we going to make that electric power? Well, a lot of it right now could just be, you know, using fracked gas and just blowing up a combined cycle gas turbines and, you know, with gas so cheap, that is what the sort of flip side of this.

00:29:56:00 - 00:30:25:18

James Krellenstein

And you see a lot of the decarbonization advocates really getting concerned because what's clear on so many of these ISOs is and Archos, whether it's New York, New England, PJM, they don't know, period, end of story, how they're going to meet meet load growth. And there's gigawatts and gigawatts and gigawatts of deficit right now on on these grids based on even before the API boom that this is only exacerbating this.

00:30:25:18 - 00:30:47:14

James Krellenstein

And so secularly you know for society it's a great thing but if we're looking for an actual need to have nuclear, this is the best environment that we've been in in a couple of decades. I hear that a lot. That's again, because of this increasing demand like nothing is better for nuclear than demand projections that are going up and up and up.

00:30:47:16 - 00:31:07:16

Chris Keefer

You need customers for kilowatt hours out of highly capital intensive projects. If you're going to put down 10 billion for a gigawatt, you want to be damn sure there's going to be customers at the back end of that. And now it looks like that's demand. Demand is emerging and PPAs, I think Microsoft announced a PPA of $115 per megawatt hour.

00:31:07:18 - 00:31:42:10

Chris Keefer

That's going to incentivize a lot of building of nuclear. And I guess this gives maybe that you're going to see a chance to dust itself off and maybe follow some of the recommendations that the couple has been making. But it's interesting, as you were saying, the demand is coming so fast, gas can be built far quicker. Do you think nuclear can catch up or are we just going to have air cannibalizing existing nuclear power and gas and then gas filling in the gap, or is gas going to be a transition fuel to nuclear, getting its act together and being able to build up for that demand, you know, in the 2030 and 2040?

00:31:42:11 - 00:32:02:13

James Krellenstein

Well, so here's here's I think the challenge is that nuclear projects I mean, it's always the same challenge. It's it's can we deploy nuclear fast enough that that data center is being built? And these days, you know, we're building gigawatt scale data centers. And the other flip thing is that dozens of states at this point have banned new data centers.

00:32:02:15 - 00:32:22:08

James Krellenstein

They don't feel like they have the power or capacity to actually meet and they don't bring any jobs. I mean, they they've been few jobs, right? Yeah. Like how many? Very, very tiny, like dozens and dozens, Right. I mean, security. And then you have, you know, nerds come in occasionally to fix things and at the same time, essentially jobless, essentially jobless.

00:32:22:08 - 00:32:46:12

James Krellenstein

And they just suck huge amounts of power. Right. So and they just don't have the power. So so the question that the nuclear if the nuclear industry wants to reach this moment, it needs to be able to prove that it can actually meet that demand in a reasonable enough time scale. Otherwise gas plus maybe renewables. But in a lot of these areas, you're not going to even be able to deploy renewables for the siting issues.

00:32:46:14 - 00:33:13:11

James Krellenstein

Right. So the question that we are facing is can the industry get a sensible plan together to be able to deliver fast enough to meet this demand growth? And this brings us to I think, what the big sort of debate that's going on is that a lot of places like Microsoft in particular are thinking that micro reactors are going to be the pathway that they're going to take to basically power their new data centers.

00:33:13:13 - 00:33:37:21

Chris Keefer

You know, this this has been a interesting trend over the years. I've tended to lean on the side of large reactors basically because I've sort of repeated this ad nauseam. But I'm a defensive pessimist. I think everything is going to go wrong. It's a present. It does go right, but it makes me very conservative. It makes me try and investigate what are the unique circumstances under which nuclear has really deployed itself quickly and economically and know.

00:33:38:00 - 00:33:56:12

Chris Keefer

I tend to be brought back to that bias. I'm certainly someone who holds strong, strong opinions, loosely held. But it has been interesting because I've had people push back and say, listen, like the United States and this was as recently as maybe three years ago, two years ago, utilities in the United States are never going to deploy, never going to deploy gigawatt scale nuclear ever again.

00:33:56:12 - 00:34:18:04

Chris Keefer

So we need to modify. You should stop advocating for this. And I go, will you will concede that the economics of large are better than small? Yes, but it's just not going to happen. No one's going to take the financial risk or maybe the demand won't emerge. And I was always saying, well, I'm going to kind of, you know, hold the candle on this this belief partially out of being a climate hawk and being like, well, the demand should show up because we're trying to decarbonize.

00:34:18:08 - 00:34:43:08

Chris Keefer

And of course, that's been a little stunted. But anyway, here we are. I've kept the faith for a few years and it's just interesting to see that. But let's let's talk about smaller micro reactors and what these gigawatt scale server farms individually and just the collective demand of all that infrastructure. Does that tip or change the balance of, you know, what makes the most sense to build?

00:34:43:09 - 00:35:07:20

James Krellenstein

So I think the question that that we're going to have, I think the smaller I think Smarties are going to be now in the same camp as the large light water reactors for this, the Smarties that we've been talking about, as we've talked about in this clubhouse, the problem is the stars right now are taking the exact same amount of silicon, not the exact same, but actually proportionally more civil works, more concrete per megawatt than a large plant is.

00:35:07:20 - 00:35:30:11

James Krellenstein

And the problem is, is that these civil works are not things that are easy to have learning curves and still are going to require large workforces to agglomerate onto a site. And the classical education you get into in the summer. Right, Because a lot of places can take a gigawatt or two gigawatts of what's new capacity. Now the whole issue is, is that we're building gigawatt scale offtake right at the site.

00:35:30:13 - 00:36:03:19

James Krellenstein

So it's becoming a less compelling argument considering the civil works that we're seeing. And I think as the BWR x 300, which is Hitachi, is megawatt boiling water reactor, some are. What we're seeing out of that design is that these are very you know, these are it's not a factory manufactured module that's been put into a pole bar and it's being put into a massive subterranean and, you know, structure that's more than 30 meters in diameter and more than 20 to 30 and more than 20 meters deep.

00:36:03:19 - 00:36:33:00

James Krellenstein

Now, even maybe even more deep then that subsurface with a lot of steel, concrete, composites. And what what the nuclear engineering experts I speak to are increasingly concerned about that intensity is civil works. We would also say that new scale kind of died because of the civil works, right? The while the actual nuclear power module, of course, is transportable and factory fabricated, that massive common reactor building is larger than the nuclear island of an AP 1000.

00:36:33:00 - 00:37:06:03

James Krellenstein

You get half the power of it, right? So so in terms of like project length and timeline, considering it's going to be more labor hours per megawatt electrical, more concrete, more steel per electrical, I'm less and less able to see the case for these small for meeting this sort of demand. Now, the other side of that pendulum, the micro reactors, the idea, the micro reactors, no, we're actually going to make the reactors so small, you know, 50 megawatts electrical or less that we don't actually need that much civil works on the site.

00:37:06:05 - 00:37:22:14

James Krellenstein

It really is mostly the factory fabricated module and maybe it's sat on some sort of cocoon or shell that you're putting there. If you have to build like 50 cocoons or shells, doesn't that sort of add up in terms of what works? But it could be you could imagine a world in which it's actually much, but you could deploy it much quicker, right?

00:37:22:17 - 00:37:49:06

James Krellenstein

It's much simpler to do a basic, you know, concrete base math that's very tiny that comes in a couple of factory fabricated prefab or even precast concrete. They're transporting from a factory and placing it on Russian. The problem, I think, is that these economies of scale with micro reactors, really, you're really tiny then and you're really having a hard time getting the economies of scale to work.

00:37:49:08 - 00:38:19:18

James Krellenstein

And I think that just some of the basic core physics of neutron economy are very, very challenging as you get to these very tiny reactors and you look at like, say, the Westinghouse, The Vinci reactor. And if you try to find burn up numbers, that is burn up relates to in nuclear engineering, it's the amount per kilogram of fuel that's the enriched uranium you have kilogram heavy metal, we call it how much thermal energy in megawatt days or gigawatt days you get out.

00:38:19:19 - 00:38:43:00

James Krellenstein

Problem is, is is that because these cores are so tiny, it looks like ISIS hasn't published a number. But researchers at Brookhaven National Lab, their own neutron analysis and tried to calculate this, you know, just from the basic core physics, you really lose on just fuel utilization. And when we think about nuclear power, we don't think that fuel is a general driver.

00:38:43:02 - 00:39:15:18

James Krellenstein

But, you know what? The numbers are showing on the up on the edge case. We're going to be using 16 times more uranium metal per unit megawatt, then a light water reactor. We're a little short on Liu, not to mention how high. AC It's even it's even more profound than that, because it's not only just the uranium ore, which is also, you know, we're seeing increased market pressure for more uranium ore then we're also seeing in the whole front end of the fuel cycle, we're seeing tightness in enrichment, in conversion and conversion.

00:39:15:20 - 00:39:35:03

James Krellenstein

But then the other problem is that all of these, you know, micro reactors, the way they get around having a full containment building is by using what's called a functional containment, which is generally by using the fuel type is more like chisel, a uranium oxide carbide, tri structural isotopic fuel and that is a really expensive fuel to make.

00:39:35:03 - 00:39:59:14

James Krellenstein

Right now, this taking is costing about $10,000 a kilogram right now to fabricate traces of fuel. And we're usually utilizing it a lot less efficiently. We're utilizing a lot more uranium. All of a sudden, it's becoming more and more challenging to figure out how do we actually get the economics of a micro reactor to work, not to mention the capital cost issues that we're having from lack of economies of scale.

00:39:59:16 - 00:40:23:23

Chris Keefer

So I think I got the stat from you regarding the eventually micro reactor requiring 900 kilos of halo, which incidentally is what sentries which is what sentries committed to making this I think expanded that excuse me, 1616 centrifuge cascade. In the end, they didn't even meet that goal. I understand. But it's only because of this, because of the movie that's not their fault.

00:40:23:23 - 00:40:57:19

Chris Keefer

It was of the energy could provide them like, okay, well I ultimately the goal the goal was not met. So, no, I'm not slamming centers here. It was. But what you're saying is that nine degrees might be an underestimate given these these poor neutron acts and the lack of a published number from Westinghouse. And yeah, I mean, I some people ask me in the tons of of Hailu per 15 megawatt or five megawatt electrical reactor micro reactor and that's going to be the ultimate question I think that's going to play out in the next couple of years.

00:40:57:19 - 00:41:19:21

James Krellenstein

I think you probably know where my biases are, but I think it's going to win. But the real question is if there actually is a micro reactor that can be sort of factory deployed, sited and set in a couple of months, that could be incredibly attractive to doing this. The question is, can we make it economical? And I think it's attractive for a small server firm, but a gigawatt server farms do the math.

00:41:19:21 - 00:41:38:06

Chris Keefer

I mean that's a lot of micro. It's going to be hundreds of micro reactors per plant. Are you convinced by that? Does that sound economic? Well, what's it if it turns out that actually these things I just don't the first question I have is the fuel and how do you get over this, the basic neutron mix of having a tiny, you know, neutral economy?

00:41:38:06 - 00:41:55:14

James Krellenstein

What we're talking about is it's like how many, you know, in a nuclear reactor, as we know, like you basically you physically split an atom and that atom releases neutrons. And those machines don't have like a G.P.S. that tells it, this is where the next physical atom is. But it sort of is taking like a drunken walk.

00:41:55:16 - 00:42:15:13

James Krellenstein

And it might hit another physical to 35 or 39 nuclei and split that, but it might also get absorbed by something else. They by water are made by some of the elements of the fuel, or it might leak out of the core and activate something well, or hopefully not and hopefully just be absorbed into water or something and shielding.

00:42:15:15 - 00:42:31:09

James Krellenstein

Right. But the problem is, is that as you think about just its geometry, right? If you think about the surface area, the volume of a you know, generally we are cylindrical, right? A big core is going to have the probability that a neutron in the middle of it is going to leak out of the core. It's pretty low.

00:42:31:09 - 00:42:55:06

James Krellenstein

It's going to get something before it leaks out. Obviously, you get very, very tiny. You're leaking a bunch of neutrons out of the walls of of the reactor core. The boundaries essentially are that you can put a reflector around it, the bounces them back, the new reflector going to be that good. So there's just an inherent tronic disadvantage of going that tiny from for fuel utilization.

00:42:55:08 - 00:43:21:12

James Krellenstein

And so what I'm trying to say is that the laws of physics are the laws of physics. I like it. Less fuel is going to come much, much cheaper than it already is. We're going to have a pretty hard time getting the economic world that's ignoring all the other drivers. We don't generally think of fuel as as a major driver of the cost of nuclear power, but those ten times the amount of fuel or 16 times the amount of fuel, all of a sudden that becomes a little bit more significant.

00:43:21:12 - 00:43:48:16

James Krellenstein

But now we're talking about also all the capital costs associated with building five, ten or 15 or 20 megawatt electrical reactors over and over again. So, you know, there was a much maligned article by the nuclear community. Allison Macfarlane was one of the principal authors, and they talked about the waste volume that was created. And I think a lot of nuclear advocates said, yeah, maybe a little exaggerated and picking, you know, worst case scenarios.

00:43:48:18 - 00:44:12:04

Chris Keefer

And a lot of the advanced companies in particular were saying, listen, you didn't consult us, we had better numbers for you, but it sounds like you will make more nuclear waste. I'm not afraid of that waste. But with how we currently manage waste and the plans of doing things like deep geologic repositories, which in the Canadian example are going to be north of $20 billion if we're going to manage it that way, the volumes might matter or I don't know.

00:44:12:06 - 00:44:32:00

James Krellenstein

I mean, I just like I just don't think about nuclear waste and I that's like my own pro-nuclear nuke, bro. But I mean again, $28 billion or do procrastinate and put them in casks, I don't know. I just like I just don't even think about. Okay. Yes. I mean, I mean, I haven't done the economics. I imagine that there is an economic case that this is a problem, but that is not what I'm particularly sure I really am not.

00:44:32:00 - 00:44:53:01

Chris Keefer

And just like maybe that's my own biases. I just like kind of saying, you know, we have safely stored nuclear waste pretty cheaply. I'm just yet to clarify that the only reason I'm worried about increased waste volume is just if we're crazy enough to handle the waste being planned with degrees which cost tens of billions of dollars, then that that starts to matter to me.

00:44:53:01 - 00:45:10:10

Chris Keefer

Not not the sort of danger from the waste and our ability to manage it an intermediate basis. Okay. I want to get controversial on the Canadian side. So we have this plan, a Darlington to build 4x3 hundreds. And again, to me this is like a past dependency from a time where we really didn't see demand kicking off and it might have been foolish.

00:45:10:10 - 00:45:33:02

Chris Keefer

And when I'm able to finance a gigawatt scale reactor or a series of site is licensed for 4800 megawatts, we are treading water by tinkering, taking Pickering off line. Luckily, we succeeded in saving Pickering. B but the addition of 1200 megawatts, we're turning water, we're not adding capacity. We're relying on Bruce to start building additional nuclear capacity and frankly, at risk.

00:45:33:02 - 00:45:54:09

Chris Keefer

But frankly, that's not a lot compared to the pathways to decarbonization projections of 18 gigawatts and probably what I think is actually more pragmatic, rational for deploying, maybe setting a whole bunch of data centers. And so, you know, the x 300, as you're saying, I mean, it's it brands itself as EMP of a kind. But the this plant architecture is certainly first of a kind.

00:45:54:09 - 00:46:17:10

Chris Keefer

It's pretty novel. There's some unforeseen risks there. But fundamentally the rationale is to build it in order to be able to deploy in smaller provinces. And to me, that's a noble goal. I personally think it should be a federal government responsibility. If you want to have a nuclear strategy that's deploying out of province, not a provincial utility, I'm going to get in shit for saying some of this stuff, but and maybe you won't reply.

00:46:17:10 - 00:46:33:16

Chris Keefer

And that's that's totally fine. But to me, it's like there's a new rationale now to do large at Darlington. You have these data centers and also and also, you know, contrary to what all the engineers are telling us, that we could, you know, shut our nuclear fleet down and go efficiency and bring water in from Quebec. Quebec is hurting.

00:46:33:16 - 00:46:54:10

Chris Keefer

We've been burning gas all year to export to them so they can hold back water in the reservoirs for winter like it was a mild winter. But they lost $1,000,000,000 in exports. So I'm back. So, I mean, if we produce big power in Darlington, if we consume the 4000 megawatts on nuclear sites, which are very hard to come by now, I think there's the security.

00:46:54:12 - 00:47:22:05

Chris Keefer

Are the people secure enough? Are the kind of a 510 year PPA might make a nuclear developer a bit hesitant how locked in or these is how much of an economic rationale is there in terms of de-risking large nuclear at a site cycle. Just say like Darlington so you don't have to wade into the politics. Yeah. No, I think we've been instructed by certain members of the Canadian nuclear community not to comment about, you know, God forbid, public members of the public comment.

00:47:22:05 - 00:47:46:01

James Krellenstein

And maybe that was more directed at you than me. But, you know, I think without talking about. So here's a couple of things that are good. What is that? What are good about a lot of northeastern nuclear, I guess southeastern Canada, but northeastern United States, southeast like Ontario, Quebec, New England, New York, Pennsylvania. What's good about these places?

00:47:46:01 - 00:48:05:14

James Krellenstein

Are there places in the world like, you know, we're just weird? Bruce on Lake Huron. Lake Huron is really, really big. Like, there's no way we're going to have to worry about consumptive water consumption. And one of the things that we also are doing at Datacentres is cooling all of this thermal load or putting this you basically all get converted into heat.

00:48:05:18 - 00:48:25:22

James Krellenstein

So as I said, the actual amount of the actual amount of thermodynamic computation that you're doing is very, very little. So almost all of it just lost resistive load on the heat. So when you put 700 watts into an H 100, right, which is a Nvidia GPU, a hopper Microarchitecture. Right. That's almost all coming back. Right. Right. That needs to be taken away.

00:48:26:00 - 00:48:50:18

James Krellenstein

And so one of the good things that we're looking at nuclear sites is, is that they're all they're generally also co-located with large cooling water resources. So we're really seeing this nuclear marriage between nuclear and data centers work for a bunch of reasons. Siting is really good. We have that direct power. We can just take off the bus bar and then also we offer it generally co-located on Lake Huron or Lake Ontario.

00:48:50:23 - 00:49:24:12

James Krellenstein

We have Darlington and Bruce, which is great for that and we are starting to see with Hydro-Québec right in this region of the United States and of Canada. You know, I live in New York State and we were relying on the Quebec law to basically export power down through chippy and other feeds to us. And as you just said, Quebec right now is into importing power from Ontario because, I mean, there's a combination of things, but I think it's also I mean, Ontario has always had a large amount of electrified heat, but we are generally starting to see that electric power use is also increasing as well.

00:49:24:12 - 00:49:45:13

James Krellenstein

And so we are in this space where it's, you know, we're going to need a lot more power. And the question I'm going to opine on it, that if I'm OPG, I would be asking at the Darlington site if I have 4800 megawatts of electrical sited licensed or having an impact statement for that, does it make sense to use that site for 1200 megawatts?

00:49:45:15 - 00:50:07:07

James Krellenstein

Right. Or are we going into a world where we can be confident enough securely that there's going to be enough demand for low power, low carbon power, that you might as well take that site and build for if you want, thousands or monarchs or even your 1400s or. BW There of course, won't opine on selection for and of course I will.

00:50:07:07 - 00:50:32:18

James Krellenstein

But like, you know, that's the question and it's a real different this is a game changer if this whole if this pattern holds, we are looking at where there's a premium for four gigawatts and gigawatts and gigawatt hours of low carbon firm electricity, and we don't have that much supply coming on to the market, which really goes back.

00:50:32:19 - 00:50:52:08

Chris Keefer

It doesn't make sense to build for BW 100. I mean, I think there's again past dependencies and sort of some cost fallacies that play into these decisions. Yes, my paradigm, again, SLR is in grids where you can't deploy big reactor because it's more than 10% of the grid. And if it has an outage unplanned, it's a big deal to compensate for.

00:50:52:09 - 00:51:11:13

Chris Keefer

That's an absolute case for us and Mars. And we have a number of provinces in the country that really don't have hydro and require us Mars to decarbonize. I don't think they're going to be as economic as large. But if we're willing to pay a cost premium, if the federal government's willing to step in and backstop that and support that for climate reasons, I think that's fabulous for me.

00:51:11:13 - 00:51:40:01

Chris Keefer

Like the Bulls case for nuclear in Canada, accepting that I think is a is a pretty obvious premise that nuclear power has tended to scale up in terms of chasing economics is building nukes and on big nukes in Ontario, in Quebec and NBC. And if the Quebec powers won't do it, why don't we build them close to the Quebec border and strengthen our transmission lines and or paradoxically, use the existing lines to be sending power back to them if if their power demands are dropping.

00:51:40:03 - 00:52:03:02

Chris Keefer

But we're kind of veering off of off of the question. The question, though, is, is that I think for. Right. You know, going back is is that these are great sites, though. Canada's cold as as we know, it's cold. It's got a lot of water Right. It's great for doing passive cooling and getting that power utilization efficiency really, really low on having to build cooling tower.

00:52:03:04 - 00:52:23:02

James Krellenstein

Well, it's expensive. Hypothetically, right? Well, yeah. And even if you do have to build cooling towers, it's so much easier to have so much make up water for the evaporative losses and for the blow down of your cooling tower that you know, this is a great these are great sites. Toronto is a megalopolis, right? You know, the Northeast has Boston and New York City, etc..

00:52:23:08 - 00:52:43:20

James Krellenstein

We really are in this place where this corner of the United States and Canada has a combination of high populations. Relatively good economy is a real going to be a real air loads center and they don't have a clear path to make that, you know, cheap power. Except for that the long history of nuclear power in these regions.

00:52:43:20 - 00:53:04:21

James Krellenstein

So I'm really I'm seeing for the northern part of the United States, particularly in the eastern side, is really going to be an opportunity. And in Ontario for new nuclear. And I think that is going to be the sort of paradigm shift that we all have to embrace is a good one, which is that maybe it's going to be we're going to be demand constrained.

00:53:04:21 - 00:53:31:15

James Krellenstein

I mean, sorry, supply constrained, not demand constrained. Right. And that's the greatest thing that we've ever had for nuclear. The one thing that happened, I think in the not naughties and and to use your term in the first renaissance is we had a combination of increased efficiency. You're decreasing electric power, decreasing natural gas prices, and we had decreased electric power demand, especially as the global financial crisis hit, because, you know, economic activity sort of decreased.

00:53:31:15 - 00:53:55:06

James Krellenstein

And so we just didn't see if not decreases, we didn't see almost any load growth and it may become harder and harder to justify such a large capital investment to build new load, even if you you know, and people didn't value carbon that much and probably don't do it now. But now we're in a situation where the real question I have to ask is, are we going to be building gas or are we going to be building new nuclear capacity?

00:53:55:06 - 00:54:33:04

James Krellenstein

And the answer is probably both, right? Plus renewables to serve this increased demand and the real question that is, is can the nuclear industry present a credible enough path? So it's not just my hand on one and two, you know, it's building Susquehanna three. I mean, Bell Bend, right? Like, you know, that question about can we actually get in strong enough demand signal and have as a counterparty if you're Microsoft, Google or Amazon, can you actually have as a counterparty, a nuclear provider that can actually that they can be comfortable they're going to power to power their data in by the time it's done?

00:54:33:04 - 00:54:50:04

Chris Keefer

It's interesting you talking about Susquehanna three. If people talk about Bubble five, I mean, bogus may be an exception because there's now four reactors on it and it's the largest single nuclear site in the US. But in Canada and Ontario, we filled out our sites, right? I mean, you were just at a four reactor station on a eight reactor site.

00:54:50:04 - 00:55:06:08

Chris Keefer

There's room at Bruce. That's a huge site, but we are a little bit constrained now in terms of Darlington. Apparently, if you do the for us, there's not really room, certainly not without opening up an environmental assessment process, which is the surrounding of Dodges. It is not a huge site, Darlington B but I mean the terms of lakefront.

00:55:06:10 - 00:55:25:19

Chris Keefer

All right. And then we have we have Nanticoke, which is an old coal site. We have Western Ontario Hydro back in the sixties and seventies, started working and planning out and acquiring properties that were favorable for basically large coal or nuclear plants. And they built some nuclear plants on them and they built some coal plants on others. And there's a couple empty sites, maybe only one West level.

00:55:25:21 - 00:55:40:19

Chris Keefer

But I guess we're at we're potentially running out of nuclear sites in in the US in a lot of single or double station plants. You can add a ton of nuclear without having to go through all of the issues with permitting and the social license around a new site like it sounds, not to mention the sort of brownfield coal sites.

00:55:40:23 - 00:55:56:16

James Krellenstein

Right. It sounds like there's a lot of places to build. And you're saying there's not a lot of places. I mean, sites like this place could have 30 reactors. Right? Right. Like if you think about the water, the lake frontage, you don't have to have the reactors on the lakefront. Right. You just need to have the make up.

00:55:56:18 - 00:56:19:23

James Krellenstein

You have to the circulating water intake, an outflow. Yeah, right. And so you can basically pipe in the water, you know, 500 to 1000 feet inland. Right. Right. If you have a big enough pipeline. Right. If you think about like a nuclear plant like. SEABROOK Right. SEABROOK is not in New Hampshire, right? SEABROOK It has a once, you know, an open cycle, one through, you know, takes seawater in and squirts it back out through a deep, long outfall.

00:56:20:00 - 00:56:40:14

James Krellenstein

But that's like, you know, SEABROOK is on the beach. Yeah, right. It's like in the marsh. But if there's a big underground 15 foot diameter tunnel, all I know is that there's preoccupations around not having enough sites or making sure we preserve sites because there's competitions for these sites. You know, farmers want to farm them, housing developers want to develop them, manufacturers, real estates.

00:56:40:16 - 00:57:14:13

James Krellenstein

So maybe this is something where where regulatory reform really is important. And I think one of the things that we we do need, you know, I'm one voice that defend the NRC, defend the regulators and the Canadian system, because I just you know, I it was only until I just drove up today that I figured out that it's not you guys have your own government know law of not that the United States itself is a diverse state, but but one of the things that is going to be a major problem in the United States regulatory environment for deploying nuclear as fast as we're going to need to do to meet this demand is going to

00:57:14:13 - 00:57:34:07

James Krellenstein

be not necessarily the nuclear side of regulation. I think we could speed that up pretty quickly. In some ways. There's some statutory changes that we need to have for the mandatory hearings to be abolished, but especially now that we have such a amount of certified designs, we know how to site those designs. Do the you know, the geotechnical boring and seismic analysis has been cited.

00:57:34:09 - 00:58:07:12

James Krellenstein

They promise the environmental impact statements, which are historically the rate limiting the slowest part of the nuclear for the 50 part 52 process it's getting this is issued and that's not in nuclear legislation in the United States. It's the National Environmental Policy Act in the late 60 NEPA. And the question that I have to ask that we're going to have to I think nuclear advocates are going to have to do is how do we get the NEPA reform for EIA is for for environmental impact statements to be issued fast enough that we can really actually start building nuclear quickly enough.

00:58:07:12 - 00:58:42:05

James Krellenstein

And that's I think we as advocates are going to need then the entire community is going to have to raise your focus down on how do we get EIA issued fast enough to be able to enable nuclear to be deployed. And the on the industry side, we need to start seeing a real battle plan being put together. But how if we go to a large light water reactor out, how do we get, you know, a BWR level, you know, build times of 48 months and below for a large light water reactor to be put together at that.

00:58:42:10 - 00:59:10:18

James Krellenstein

So we do have a regulatory issue because we're it's not just that previously was like, well, you're doing that. Yes. You're not actually you don't have a construction site going through capital costs or interest costs are low, right? You're actually building the plant. You had the. Yes. But now if we're talking about we actually need to meet load quickly, we need to actually just think about the total schedule time from I decide to build a nuclear plant, nuclear plant coming online every bit of time needs to be compressed.

00:59:10:20 - 00:59:35:03

James Krellenstein

We've got to figure that out. And there is a regulatory real regulatory barriers to getting this deployed quickly. Okay. Which which I think is going to move nicely into air applications because I've got a couple ideas here. And I don't know, I think you'll find this actually interesting. So Sam Altman talking about needing a energy revolution. And of course, you know, Oklahoma Silicon Bro is going straight to Fusion and then and Oklahoma.

00:59:35:05 - 00:59:54:11

Chris Keefer

Okay, pretty reactors. That being said, this is classic, right? When the nuclear issue starts talking about innovation, they innovate the stuff that probably doesn't matter the less you know. From Jakob's work, from an MIT work, we know that the majority of the plant costs in terms of overnight construction costs is actually boring. Old stuff like balance of plants, civil works, etc..

00:59:54:12 - 01:00:16:02

Chris Keefer

So and fundamentally that's a question of project management. We were just a bit bruised meeting some of the most competent project managers in the certainly the Northwestern hemisphere, probably the Western Hemisphere. Really, really amazing crew up there. Exciting to see the way in which steady we know project after project refurbishment. Refurbishment is training up and dialing in those human factors.

01:00:16:04 - 01:00:46:01

Chris Keefer

Is there a role contrary to just this idea of maybe cracking cold fusion or something like that, which I'm skeptical of? What about I using and assisting project management? Is there a role for that course? I mean, I it's going to be transformational. Almost every aspect of the economy and of course is going to help. And I you know, I'm not so sure that's going to be the we can't just I mean, we need to be AI enabled and we need to have developer organizations that are folks that you were just at, Bruce.

01:00:46:01 - 01:01:08:23

Chris Keefer

I mean, we just we're at the the the sort of outage control center right that was doing on the NCR control center. And, you know, there's a lot of this is a lot of real human labor issues, right. I mean, we were just talking about at Bruce the shortage of housing right now in that area because there's so much because the nuclear is giving so much blue collar labor, so many jobs.

01:01:09:01 - 01:01:47:07

Chris Keefer

Right. You know how we do project management, You I can help I don't think is going to be a panacea for this. And I would just say is we know that the Japanese Ave project there was there was no neural networks powering that there were Palm pilots and a lot of software that was being developed. But if you look at some of the more successful nuclear projects in the United States, as you know before that a lot of times there was computer based sort scheduling and work package management, but in some cases it was taking weeks to run those compute computer programs because, you know, the mainframe may have been at a Basco or Bechtel's

01:01:47:07 - 01:02:23:10

Chris Keefer

headquarters and you look at time for the punch cards to get up to from the site to the spot. So I don't I think it's going to be helping. I don't think it's going to be transformation at all. I think we do need to focus on what Bruce and what the amazing team at Bruce Power and at some other organizations has been doing, which is really getting that secret sauce of how do we do this project management on these crazy complicated candy refurbishments right down to an actual science and that, I think, is is where we're going to need to if large light water reactors can meet this demand.

01:02:23:12 - 01:02:41:17

Chris Keefer

And the question that it's going to ask is, do we just want to totally eliminate that away? I put most of it in a factory, right? I'm doing the micro reactor out and fuck it. What does that mean? A lot more uranium ore. And, you know, this was this is kind of news to me because modularity has been this buzzword.

01:02:41:17 - 01:03:10:05

Chris Keefer

It's really sort of been associated psychically with small modular reactors in particular. But I was watching some videos of a BWR construction in Japan and seeing how modular they were through talking with you, understanding you see a0200 story which is made up of 70 or 75, 8772 sub modules that the module assembly plant that was built in Vogel One of we've talked about the applications for AI in project management.

01:03:10:05 - 01:03:29:15

Chris Keefer

What about the applications of AI and robotic labor replacing that stick build construction labor which is craft labor, which seems very hard to romanticize, but if you have a module construction factory, it would seem that that's a more compact space in which maybe you could replace some of that craft labor I don't like as a kind of bleeding heart.

01:03:29:20 - 01:03:49:09

Chris Keefer

I'm not sure I'm a socialist anymore, but that's kind of the school I come from. I love the human story of nuclear. I love the human flourishing that comes along with the incredible skill set that's required. You know, again, I'm very kind of hesitant, but just to explore this through, not to make a value judgment. What do you think about sort of again, in relation to nuclear construction productivity?

01:03:49:09 - 01:04:18:02

Chris Keefer

So if you look at like that and I imagine it's it's that unit which has a weird parallel between Japanese porcelain that you're talking that's like, you know, this like, you know, craftsman, craftsman, but like you know what it is like, for example, the BWR. I mean, that's why they want a reactor, the, the containment liner, right? They, they actually designed Taichi and Toshiba designed it such that it's in TEPCO, such that it can be done by robotic welding.

01:04:18:02 - 01:04:42:00

Chris Keefer

Right. And you saw the robotic welding in the nineties. That's what I base it just literally going through a circle right through the diameter of the circumference excuse me, of the of the the containment liner. And it's just doing it as robotic welding, right. Using classical CMC based welding techniques. And there is going to be a lot of role for automation.

01:04:42:00 - 01:05:17:12

Chris Keefer

But once I do not see it, I have a pretty I accelerate a I acceleration is and if you look at like some of the projections, one of the things that is going to be least impacted by A.I. is stuff like manual labor, like craft labor, not because because the main thing that's preventing us from doing like, you know, robotic welding is not that we don't have the you know, we don't have computer numerical control to basically figure out how to do a well, it's that like literally if you think about what's in the nuclear plant and you know you were just in we're just sitting inside can do and the heat transport pumps are

01:05:17:12 - 01:05:38:16

Chris Keefer

doing a lot of those wells. It's really hard for a robot to be able to do the dexterity of a human being, to be able to actually do the welding in these close sort of proximity space spaces. Right. So not all the wells can be automated. I think not all the electrical work is going to be able to be automated, although the h vac and plumbing is going to be able to be automated.

01:05:38:16 - 01:06:00:11

Chris Keefer

So one of the things that we do if we're serious about using a modulate in to a robot size in a module factory, yes, I think we're going to see a lot of robot station in nuclear. I think that's already happened. I'm not so sure. I think the AI revolution is going to enable that particular type of, but I think a lot of that could be actually already done better.

01:06:00:13 - 01:06:25:01

Chris Keefer

Yeah, the robot. And there will be a robotics revolution that may be actually secondary to AI that maybe allows us to do a lot of this through manual, through robots rather than through here. Robot Welder Yeah, but, but I think the question is, is that I don't see a world in which we're one of the few things that I'm a pretty pessimistic you know, I acceleration is I still see a large role if we're going a large reactor out for skilled trades.

01:06:25:01 - 01:06:42:00

Chris Keefer

And I think one of the things we were just talking about it up at Bruce, obviously there's a skilled wage shortage here in in Canada and in the United States, that skilled wage and definitely some people are saying there's a skill shortage and there are some people say, that's crazy trade. Yeah, right. Skilled trade, sorry. And that I think we have both.

01:06:42:00 - 01:07:21:10

Chris Keefer

We need to figure out a way to get the administration. We're talking the apprenticeship periods that we're talking about for welders and electricians of four years, bachelor's degree, length of training that is necessary. We need to start thinking about the innovative ways that we're going to increase the supply of skilled labor and working with the unions and the union halls to make sure that actually we have the skilled labor workforce that maybe has atrophied over the last 30 or 40 years to be able to build these plants, because everything that you're hearing is one skilled, skilled trades are shortages and all you know, it's not just nuclear that's being built, battery plants, electric vehicle with

01:07:21:10 - 01:07:44:06

Chris Keefer

the Inflation Reduction Act and the I don't know what the Canadian equivalent of this is, we really are seeing a return to industrialization, which is a fantastic thing. And I think, you know, despite the almost melee and fight we created at Bruce at one point talking about what the wages are like, there is good you know, obviously the work for the workers were what were pushing us together the better.

01:07:44:08 - 01:08:08:01

Chris Keefer

But but people are getting paid a very decent wage with benefits, with good stuff, too. Being a skilled welder, a skilled electrician, a guy who's good at each back and doing cable trades, these are skills that are going to be needed as we move forward in this economy. And assuming robots don't take over the world. Right, this is something that we need to figure out as an industry.

01:08:08:06 - 01:08:40:21

Chris Keefer

How do we incentivize to build the workforce that we need? If we're talking about building 30 gigawatts or 50 gigawatts, let's in Canada, in the US, let's figure out how do we create the labor force that can actually build those plants. We don't have it right now, right? I mean, which brings me, you know, in a more sort of broad political sense, you know, the ways in which our economies have become in the West, kind of hollowed out from a productivity not not productivity, but from the production of goods and industry, a kind of fire economies, finance, insurance, real estate.

01:08:40:22 - 01:09:01:12

Chris Keefer

And the E is for you think about the way that the UK, you know, probably partially due to Thatcher, but also just kind of neoliberal globalization. A lot of the economic activity is retreat to sort of the city of London. And it's it's stuff that, you know, derivative trading and stuff that's not related to actually building things. And you know, it's interesting how, you know, whether, you know, everyone's been learning to code.

01:09:01:12 - 01:09:26:18

Chris Keefer

I think we can all agree that coding is now not something that kids should be going into. They should be maybe going into this skilled trades, but there's going to be like a culture shift, I think in the US like this could a relatively elevate some of the red states or like flyover country, like where people know how to build ship, like there's a potential for this to sort of radically change things as well in the context as well of globalization or certainly French shoring.

01:09:26:20 - 01:09:45:00

Chris Keefer

You know, I see major societal shifts that are not just kind of AI overlords and robots killing us all Terminator style, but the ways it's going to help them too. But like, you know, you could okay, if you need more energy, kill off the humans, right? Especially the poor ones or whatever. And then you could use the existing nuclear fleet and run future.

01:09:45:00 - 01:10:10:22

Chris Keefer

Yeah. I don't, I don't know that I might be doom scenarios about that know we don't have to go to do PRISM but in assuming that ignoring humanism for one second you know on the flip side this is kind of I think, a rebalancing that's going to be societally good. Right. And I I'm a pessimist as well, but I think this there is a world in which when we start seeing the return it I you know, you go a blue state red state but in some it's it's urban rural.

01:10:10:22 - 01:10:39:13

Chris Keefer

Yes in some ways. Yeah right. That we're talking about and the idea that it's not going to be consultants and coders at some, you know five letter startup with no vowels. Right. Like it could also be the fact that we're also giving jobs to the men and women graduating high school. They don't have to necessarily go to a a college will be paid to learn in an apprenticeship, but really having that solid in industrial workforce coming back together to build the electricity that we're going to need.

01:10:39:13 - 01:11:01:05

Chris Keefer

And it's not just nuclear where the transmission infrastructure being built. We're to need manufacturing of all sorts of different things and you already are starting to see it in, as we tried to French or to use your term in on the chips on semiconductor fabrication stuff, right. Very closely related, of course, to AI. Right. Because we don't have the United States really nowhere has except Taiwan and a little bit of South Korea.

01:11:01:07 - 01:11:32:01

Chris Keefer

These very, very low nanometer node, you know, four nanometer fabrication technology that is necessary to actually make these H one hundreds now V one hundreds that are coming online, these large GPUs that are powering the the server farms that are actually running the yellow labs. So we are starting to see that there's a shortage of people just have the trades throughout the economy, both energy and actually building the chips, building the fabs that make the chips.

01:11:32:01 - 01:11:57:15

Chris Keefer

This is sort of I don't want to I hate to sound so optimistic because it sounds I mean, I'm sure a lot of hate about being so Pollyanna ish, but there is a world in which this is a very good thing for American and Canadian dynamism. The idea that we're actually going to go back to some of the communities that had been devastated by neoliberal policy from 1980, 19 92,000.

01:11:57:20 - 01:12:16:10

Chris Keefer

Right. And bringing back to those communities real industrial jobs and also giving them a path way for people who like to work with their hands. Right. Rather than just sitting at a stooped over a laptop like maybe you are or you are physician. But, you know, I spend most of my time about my laptop and I can work wherever I want.

01:12:16:12 - 01:12:45:12

Chris Keefer

Whereas if you're actually doing, you know, craft or manual labor, that's something that I think could be societally good, Right? Right. Yeah. I mean, carry on, carry on. But I'm just I'm just thinking about our trip to Bruce, right? You've just seen how crowded that plant cafeteria was, right? I mean, the plant was just bustling with people like there's part of it, you know, everyone's complaining about the lack of housing.

01:12:45:14 - 01:13:10:02

Chris Keefer

But on the flip side, this is fantastic for these communities to have that economic opportunity that's coming in to have new growth. And it's not just college educated consultants at McKinsey and BCG, but to have people able to access six figure incomes without, you know, pretty quickly out of coming out of school. That's a good thing for the economy.

01:13:10:05 - 01:13:32:04

Chris Keefer

Right. And I think what the street needs to do is, is let's figure out how we can make sure that we're getting the amount of trades people that we need for the building, the large reactors, if we go down that road. And that's I don't know, I'm optimist. This is what I'm excited about, is that we're actually starting to see in real life like what we were talking about, about why we like nuclear so much.

01:13:32:06 - 01:13:56:07

Chris Keefer

It's interesting. I mean, there's certainly different flavors within the nuclear obviously community that I would describe some people as, frankly, being religious. We often criticize environmentalists for sort of, you know, this niche in is dead, But human beings are fundamentally kind of religious beings. And we you know, the environmentalists have found a new god to worship, and it's got very religious undertones or even overtones of it.

01:13:56:08 - 01:14:16:14

Chris Keefer

You see that in the nuclear world as well, and a kind of worship of energy, abundance and kind of just a love for nuclear. You know, nuclear for nuclear sake is not really my motivation. Certainly any more like nuclear as a means to an end makes make sense to me. You know, again, I am concerned in terms of, you know, ultimately who can pay for the energy.

01:14:16:16 - 01:14:35:11

Chris Keefer

And I do think it's going to sort of pull away and distract from climate objectives or even just everyday people's needs in terms of, you know, who's got the dollars to pay. You know, I can't give you why I'm I'm unclear just to turn first. And I and people know I'm a Duma gloom man on the air necessarily.

01:14:35:11 - 01:14:57:21

Chris Keefer

But, you know, I'm just general you know, I'm a Jewish, you know, hypochondriac. And like, so the the the issues that I see what's great about this and let me just let me just be optimistic for one time is and make a really bullish case for nuclear. And why I'm excited is that here we have the ability for the first time that we're we're not demand constrained, we're supply constrained.

01:14:57:21 - 01:15:19:09

Chris Keefer

Yeah. So number one. So that means that people are going to be less price sensitive than they normally would. And all of a sudden, if they're able to say, you know, I'm going to pay $115 a megawatt hour for carbon free power, or maybe even north of that. Right. All of a we talk about coming down the learning curve in nuclear or in any clean energy technology.

01:15:19:09 - 01:15:40:05

Chris Keefer

By the way, this is not just nuclear, but if you're interested in CCS or geothermal, you're just another renewables, battery storage, whatever. All the sudden we have people who will buy those first plants to start us going down that learning curve. Right. And that is a difference from even a couple of years ago about where we were. Right.

01:15:40:08 - 01:16:10:19

Chris Keefer

That we are going to be able to start drive down this learning curve. And that is a very different place to be then not being able to even sell that first couple of projects. Right if actually these power prices hold, if we continue having a reasonable, you know, carbon policy environment or even a better carbon policy environment, then we really could start actually going down and deploying in in a way that we haven't seen before.

01:16:11:00 - 01:16:26:18

Chris Keefer

Right, Right. We got to wrap it up somewhere, James. This has been absolutely, as usual, fascinating. I mean, even more fascinating than, you know, part three of Vogel 1052 right.

01:16:26:20 - 01:16:47:22

Chris Keefer

Because again, I think it's just such a evolving, exciting issue, which is blowing up. I think some of the past dependencies we've been had on just the new paradigm is here and it, I think, radically alters where we're going in terms of energy. Again, if you appreciate this podcast, I mean, I feel like I'm doing things that will get me canceled all over the place.

01:16:48:00 - 01:17:13:19

Chris Keefer

Free speech I think is really awesome. It allows ideas to circulate. You know, ideas are based on paradigms of ten years ago and the paradigms change if we exist within a culture, potentially a nuclear communications culture, which is not particularly tolerant of, again, public criticism of public discourse, a relatively new phenomenon in the industry that people that are maybe more pro-nuclear than they are, or people who are outside of the industry commenting, I think it's good.

01:17:13:22 - 01:17:29:21

Chris Keefer

Support us if you like the work that we do Patreon or you can just reach out on the website and get in touch if you want to make a bigger animation. We have more efficient ways to make that happen. Shameless plug at the end there. James, any any closing final words, The final thought goes to you, my friend.

01:17:29:21 - 01:17:54:07

Chris Keefer

Well, I just this is a moment. So ignoring the other maybe societal impacts of AI which are going to be profound and maybe negative, maybe positive, or maybe both likely both. This is the moment that that that we've been waiting for in some ways, because the power demand signals are there. It's no longer a hypothetical that we're going to see load growth really happen.

01:17:54:07 - 01:18:24:15

Chris Keefer

Load growth is happening and Virginia is a harbinger. The fact that we have multiple states are cutting basically say we can't meet load if we build a big data center here. We have well-financed, you know, customers who are able put up equity, put up cash, sign long PPA for high power prices for low carbon firm power. The challenge now once again falls to us, can we deliver and is going to be micro reactors going to large reactors.

01:18:24:15 - 01:18:44:03

Chris Keefer

I don't know the answer. That is going to be smart. I don't know. But whatever it is, we need to be able to demonstrate that we're actually going to be able to meet that. Otherwise people are going to go to other forms of energy and they may be carbon genic, they might be not carbon genic, but we have an opportunity here and now the challenge is ours.

01:18:44:03 - 01:19:07:15

Chris Keefer

And this is going to require a full court press going to require us dealing with regulatory issues, and it requires dealing with technology issues, with supply chain issues, with labor issues, with all of these things. We we really do need to rise to the moment. And it's incredibly exciting, I think, because I think it's the I'm the most optimistic that I ever have, and I'm terribly optimistic.

01:19:07:21 - 01:19:28:05

Chris Keefer

All right. All right. I don't know about the future of humanity, but but but nuclear power, it seems like it's it's having its day. And in the midst of it, I'm just hoping that the intermittency of renewable energy may save us if things get out of hand with. I told that joke too many times. Well, this will happen in the Matrix, of course.

01:19:28:05 - 01:19:48:13

Chris Keefer

Right. And again, I haven't watched it in 20 years. Well, actually, I, I, I don't, I'm not like a big Matrix fan, but like, apparently it's all solar powered. Then they all the humans and destroy the eye basically blotted out the sun right to basically block the solar PVS and then they enslaved all of humanity to use the human as their sole power.

01:19:48:13 - 01:20:09:02

Chris Keefer

It is, you know. Yeah. In 80 watts or whatever we have per person. And so, like, you know, I hope that part of the negative cycle, they should have been like feeding the, the levels of nuclear power. Okay. Do they? We shouldn't do nuclear power. They either should all be solar and wind, and then we can stop the wind and stop the sun to blot out lot up the machines from killing us all.

01:20:09:07 - 01:20:20:19

Chris Keefer

Love it like James. Great saying in verse. perfect. And we will connect again on the JP 1000 can do show otherwise known as decouple. This is our special Syria.