Barron's Streetwise - Nuclear, Crypto, and the A.I. Power Shortage

Episode Date: September 13, 2024

Stephen Byrd, the Global Head of Sustainability Research at Morgan Stanley, talks about investing in the watts needed to run tomorrow’s supercomputers. Learn more about your ad choices. Visit megaphone.fm/adchoices

Calling all sellers, Salesforce is hiring account executives to join us on the cutting edge of technology. Here, innovation isn't a buzzword. It's a way of life. You'll be solving customer challenges faster with agents, winning with purpose, and showing the world what AI was meant to be. Let's create the agent-first future together. Head to salesforce.com slash careers to learn more. These Bitcoin sites can often offer a three or four year time advantage relative to connecting a data center to the grid the old fashioned way. Hello and welcome to the Barron Streetwise podcast. I'm Jack Howe.

And the voice you just heard, that's Steven Bird. He's the global head of sustainability research at Morgan Stanley, and he's talking about Bitcoin. Only this episode is not about Bitcoin. It's about power, data centers, and a power shortage, and how we're going to come up with that electricity, and some of the ways might surprise you. Ooh, there's a tease. and some of the ways might surprise you.

Ooh, there's a tease. Listening in is our audio producer, Jackson. Hi, Jackson. Hi, Jack. There's no time to waste, so don't try to bait me into some nonsense about, you know, bocce ball or 1980s movies or any of that stuff. We have a listener question. We've got to get right to it. Who do we have?

Yeah, we have Ronnie from San Diego. You mean San Diego is in the setting of the 1980s? Careful. Private investigator TV series. It's about polar opposite twins, Simon and Simon. You know what? You're right. Let's hear the question. Long time listener, first time voice member, or wanting to get your take on the energy trade, specifically for data centers. With the rise of AI, AWS, Google, Microsoft have submitted themselves as the data center providers, at least domestically. I'm curious, though, how does the market think on how we plan to power these data centers currently and into the future? I know as a country, we're looking at cleaner energy solutions

such as wind and solar, but would love to get your take if the green energy trade is looking to join the AI party. Thanks again, and keep up all the great work. Thank you, Ronnie. It's a great question. And in a moment, Stephen from Morgan Stanley is going to tell us all about it. Energy and electric power, that's one of the backdoor ways to invest in artificial intelligence and i'm hearing a lot about those recently b of a securities put out a report about sector rotation jackson do you uh do you shuffle around your sectors to try to get your exposure aligned with where we are in the stock market cycle or do you just leave it i just leave it i think i'm gonna just leave it right I just leave it. I think I'm going to screw it up.

That's what I do, too. My approach is tactical sloth, I call it. I let the index do whatever it's going to do, and that's my exposure. So recently, you are doing what B of A says to do because they say underweight tech and overweight utilities. And that's kind of just what the stock market has done for you automatically, because at the time of this recording, the iShares U.S. Utilities ETF has returned 12% since the end of June. That compares with a 4% decline for the iShares U.S.

Technology ETF. Utilities are beating tech. In its report, B of A said, our cycle model, the regime indicator, just shifted from upturn to downturn. I don't know if you felt that where you are, Jackson, out on the West Coast. I think that was an earthquake. The regime indicator has shifted. So what does this mean for investors? B of A says, prefer the tortoise to the hare. By tortoise, it means quality and income instead of the hare, which is growth and re-rating. In other words, stocks becoming more expensive. One of the things it predicts is that dividends will become a lot more attractive to investors and a much bigger part of investor returns.

If you look over the past decade, dividends have only made up bigger part of investor returns. If you look over the past decade, dividends have only made up 16% of total returns. The long-term historical average is closer to 40%. B of A writes that tech looks egregiously expensive, as it puts it. The earnings growth is impressive, but it's also decelerating. And the S&P 500 carries, quote, extreme concentration risk. Five stocks in there, Apple, Microsoft, Nvidia, Alphabet, and Amazon together make up more than a quarter of the index. BVA thinks that volatility is likely to be elevated

in the years ahead. So you'll want those stable stocks. Okay, that means utilities. As a sector, they pay close to 3% as a dividend yield, and they were a touch cheaper than average relative to the rest of the S&P 500. I was surprised at this fact. Listen to this. BVA points out, you think of utilities as being sleepy. They say since 1980, they've returned 11% a year. That compares with 12% for the NASDAQ, the tech-heavy NASDAQ. That's pretty good for utilities, 11% a year.

Jackson, is that surprising to you? That's super surprising. Here's the way I think about it. The height of technology in my home in 1980 was a powder blue wall-mounted phone with a rotary dial. There was no end-to-end encryption or anything like that. If you wanted to go into privacy mode, you had to crouch down on the other side of the kitchen counter and just talk real low. And to do that, you had to have an extra long cord that you got at RadioShack and one of those swively untangler things that goes on there. It keeps

it from tangling. It was constant strangulation risk. It's amazing that I'm even here to talk about it. Compare that with smartphones and artificial intelligence and the internet, and it tells you how much tech has changed. And yet the electric socket that I'm looking at right now is the same type that I was looking at when I was a kid. It seems like not much has changed. Slightly more three prongs instead of two prongs, maybe. That's a good point. Maybe there's some prong proliferation. I haven't been paying attention. But of course, as we know, utilities are kind of quiet tech beneficiaries. They benefit from all the increased demand for electricity over time. And that's

especially true now because as we've talked about before on this podcast, AI is particularly hungry for power. It takes a lot more electricity to run an AI query than it does just a standard Google search. So B of A says to favor utilities, and they say that they predict strong growth in demand for electricity from data centers and ai that's part of the thesis on utilities they also like reits real estate investment trusts the dividend yields there are even a little higher i know what you're thinking office reits they're in trouble right yes are. But they're just a sliver of that sector, not nearly as large as, let's say, telecom towers, retail, healthcare, and data centers. Yes, AI is part of the thesis for REITs too. It seems like everything is becoming an AI theme. Now, I know what you're thinking. If you want stability and yield, then B of A must also like

consumer staples. And it doesn't. It's underweight there. One of the things it mentions, I don't know if this is the main thrust of its argument, but it mentions obesity drugs. There used to be this perfect symphony of capitalism where the snack companies would sell us, you know, chips and cookies and we'd get fat and then that would make us sick. And then we would turn to the drug companies to buy our medicines for our diabetes and our heart disease. Now we've got these medicines for obesity and they make people less likely to overeat. So it's potentially problematic for the packaged food industry. I guess it's kind of a good news, bad news situation for the drug industry.

So B of A is underweight on both those sectors, healthcare and consumer staples. They're also positive on consumer discretionary. They think it's going to get a boost from wage growth and from falling interest rates. And they're positive on banks. They say the earnings growth there is stable and the balance sheets look good. Okay, that's enough sector stuff. We're going to come to my conversation with Stephen, and he's going to tell us why he thinks that there is a severe shortage of data centers and why the main constraint there is power and what can be done about it and who stands to benefit. He's going to give us some examples of industries and companies. I want to give you two examples that you're going to hear him mention,

just so you have a little extra background on what happened. Earlier this year, Amazon bought a 960 megawatt data center campus from a company called Talon Energy. Talon emerged from bankruptcy last year, and it's the owner of something nearby called the Susquehanna Steam Electric Station. It's a two and a half gigawatt nuclear plant. In other words, Amazon needed a lot of power in a hurry for a data center, and Talon Energy needed stable demand for its nuclear electricity. And the two came together to basically create a nuclear data center. And now Talon shares have tripled since last fall. And Stephen's gonna tell us why he thinks

some other utility companies with nuclear assets are gonna strike similar deals soon. That's one piece of background. The other is there's a company called Core Scientific and it is a crypto miner. And its stock has tripled, too, since last fall. And it's not because its crypto business is going gangbusters. In fact, there was an event in April in the world of Bitcoin called the Halving.

We've talked about that. It lowered the reward for Bitcoin miners. Many of them have been looking for alternate ways to make money. One of those ways, as we'll hear from Stephen, is that if you're a crypto miner, you happen to have a big facility that takes in a lot of electricity and turns it into computing power, which is just what an AI data center would need. And they're constrained now on how quickly they can get those. And so along comes a company called CoreWeave. That's an NVIDIA backed startup. And CoreWeave makes a deal with Core Scientific to turn some of its crypto mining infrastructure into AI data center infrastructure. Core Scientific is a company that also recently emerged from bankruptcy.

Now it's doing deals and taking in cash, and some other crypto miners might be looking to do the same. Stephen will tell us which ones. And that's enough, Jackson. I've teased and I've teased. I don't want to over-tease. Don't want to over-tease. It's time to get to the payoff, so let's play my conversation

with Stephen from Morgan Stanley after this quick break. Come on, you don't want me to wait until later and have to take a break during the conversation. Stephen will be saying something great.

You'll be on the edge of your seat and then I'll say, hang on, I got to take a break here. You don't want that, right? Clumsy and unprofessional, I say, Jackson. I say get the break out of the way now and then play the conversation uninterrupted. What say you? Let's go to the break.

We're not still recording, are we? Enough for a quick break. Welcome back. Stephen Bird is the Global Head of Sustainability Research at Morgan Stanley. We spoke recently about the looming shortage of data centers and the power constraints and what we're going to do about it and how it might benefit certain investors. Let's get to part of that conversation now. All I hear about are data centers going up. I'm almost a little surprised to hear about a shortage of capacity. Tell me about that shortage.

Sure thing. So we started at the chip level on up to develop a forecast of power demand. And I was shocked at the rate of growth in power demand because I heard the narrative that chips are getting more power efficient. You know, there's a lot of improvements. But what really shocked me was the volume of chips, the power intensity of the chips and the growth over time. So from that model, we developed a forecast around the world, including the U.S., of how much power we would need. It is a huge increase. And so, for example, next year,

we would see over 10 gigawatts of new data centers that need to get built. There's only a little over five gigawatts being constructed right now. So there's more demand for data centers than there is supply. And the main constraint on building new data centers is a finite amount of power that's available. Do I have that right? Well, that's correct. And, you know, there are two elements to think about on that. The biggest issue typically is interconnecting to the grid. To just suddenly connect a huge new power consumer onto the grid is not so simple.

You know, I used to be an executive utility myself, and the planning of that transmission system is an extremely long-dated plan. To suddenly go to that utility and say, hey, I'd like to connect a 100 megawatt, which is a huge data center to your grid at this point, we're seeing more and more delays to do that. So as an example, in Virginia, in Loudoun County, Virginia is the data center capital of the world, the utility there sent out notices to prospective customers, people who want to connect data centers, that you should be thinking about a timeline that could be as long as seven years. Now, seven years in the world of AI is a lifetime. So that's the real challenge.

We're essentially seeing delays in interconnect. A few markets actually have problems with electrons, not enough power being produced. So Texas is getting very, very tight, not just from the grid point of view in terms of the transmission system, but the actual power plants. There's a risk of an extremely tight supply demand balance in Texas. So this gets me wondering a couple of things now, these constraints for new data centers. Is this the kind of thing that can slow the growth of some of these tech darling stocks or people just who have an S&P 500 fund and they've seen the tremendous growth in AI themed stocks there? Are we going to hit a point where there's a roadblock and they say, well, we can't grow as fast as we otherwise would because we can't get these new

data centers up. And then on the flip side of that, the types of companies that you're going to need going forward to get that new capacity, how do you think that plays out over the years ahead? We have been wrestling with that question. My take from what I've seen is I think these technology companies will find creative solutions. And that's really where we see a lot of alpha among stocks. Let's put this in context. The numbers involved in AI are just so much bigger than the numbers in the power space. And I say that because I think the AI world will pay a premium. The fully loaded cost to build a data center with chips and servers and everything is about

$30 million per megawatt of capacity of a data center. The cost to build a power plant depends on the type, but call it $1 to $2 million a megawatt. So the folks who are spending the $30 can spend a bit more to the power sector for creative solutions. And to the power world, it looks like a huge premium because we're just used to much smaller numbers. Now, how does this tangibly show up? An example would be nuclear power plants.

You could build a very large data center at a nuclear plant fairly quickly, pay a premium to that plant owner for the power compared to a traditional just interconnect at some random location on the grid. So that's one approach. Another approach is to go to a Bitcoin company and say, look, you have something that's very valuable. What you have is connection to the grid. You get a lot of power from the grid at your Bitcoin sites. We would like to essentially convert that and stop using the power for Bitcoin, allow us to build a data center, and we will pay a large premium. That is going to be a big way to deal with this as well.

And the money involved is huge, especially through the lens of like a Bitcoin stock, for example. Are these publicly traded opportunities? Who owns these sites that create these Bitcoins? I mean, I just think of like, you know, parody cartoon Sniffer Dog 23 with a value of like seven one hundredths of a penny or something like that.

And you could go in and say, hey, we could do better stuff with this electricity. But are there any, are there stocks connected with that that would benefit, or is that just a private opportunity? Now, there are. In that note exhibit, four got quite a bit of attention. It looks at essentially the indicative upside if these Bitcoin companies converted their large Bitcoin sites to data centers. And several of them have huge sites. So for example, when you add these publicly traded companies together, they've got over 4,000 megawatts of really good

American sites and 3,000 or 4,000 megawatts under construction. That's huge. I'm going to read some names from your chart that have the longest bars, which I guess corresponds with the potential upside. You have Riot Blockchain, Cypher Mining, Bitdeer Technologies, Applied Digital. Those are some of the names that jump out at me. Go ahead. You were saying? Those names, and I'd also put Hut8 and Iron in that. Of those companies, the one that's not currently exploring that is Riot, but the others have all publicly stated that they're in the middle of exploring this exact opportunity. We did an indicative present value of a Bitcoin mining

site. You know, what are the cash flows? What's the kind of business opportunity? And I think in dollar value per watt of power, that is not how the Bitcoin world thinks about it, but that's how I think about it. And the indicative value of a Bitcoin site in the United States would be something like $2 to $3 a watt. That would be the value of a site. So we did a value of time and we found that every year of savings in terms of getting your chips powered would be over $3 a watt. And these Bitcoin sites can often offer a three or four year time advantage relative to connecting a data center to the grid the old fashioned way. To go to the grid and ask for that connection, if you're lucky, could be three years, but more likely in many markets in the US, it's four or five, six,

as long as 10 years to get connected. And that's worse than it used to be a couple of years ago, because we're having this sort of confluence of new users of power. If you go to a Bitcoin site and you sign a deal, you could start construction immediately. So you would save multiple years. It takes about a year, a little over, to build a data center. So you're saving a huge amount of time. The most well-known stock in this regard is a small Bitcoin stock called Core Scientific, CORZ. If you take a look at the price chart this year, you'll see what I mean.

They've announced a few of these types of deals. Earlier this year, it used to be $3 and change, and now it's $10 and change. It's up a lot this year. So I see what you're saying about the sort of lucrative opportunity there. So that's quite interesting. What else? Tell me about some other industries and types of companies that you would be looking at here. Yeah, another category, but on the same idea of time would be nuclear power plant owners in the US. So those names would be names like Vistra and Constellation Energy, ticker VST and ticker CEG.

If you want to build a truly massive data center, and this would essentially be, this is a supercomputer, that's what we're really talking about, that would be used to train the next generations of large language models. You're going to want a massive number of chips and a truly massive amount of power. a massive number of chips and a truly massive number of amount of power so for example a nuclear power plant can be over 2 000 megawatts of power output and that's rivaling cities in terms of the amount of power so for example the greater philadelphia area is a little over 3 000 megawatts of total power usage by everybody in the entire metropolitan area. So a single nuke is massive. The beauty from a hyperscale point of view, from a data center developer, is those sites have a lot of advantages. They have a tremendous amount of power equipment. They have a lot of acreage, a lot of security, a lot of water. But most importantly, the time

required to interconnect is much faster than a traditional grid approach. So we saw one of these transactions, and you can take a look at a stock called Talon Energy, TLN. If you want to see another fun stock show, Amazon signed a deal earlier this year with Talon to use one of Talon's nuclear power plants for just such a purpose. And Amazon paid Talon a premium for the power from that nuclear plant. We believe Constellation and Vistra will do the exact same thing.

They'll sign extremely large power deals with very large companies who want to build supercomputers. So that's a big opportunity. When you look ahead, years ahead, about what's going to happen with nuclear power in the U.S., we've had the CEO of Duke and some others talking about this subject on this podcast. What does it look like? Is it taking the existing big power facilities that we have and extending them and repurposing them or what have you?

Is there any effort or push on the way for new built nuclear facilities or smaller ones? What do you think the nuclear map looks like maybe a couple decades from now in the US? We do see a nuclear renaissance, both in the United States and around the world. And we recently did a big global note on this. And yeah, I've been in the power space a long time. It's very exciting. The reason it's exciting for nuclear power is we need more nuclear power because it just provides so many advantages. It's a stable source of power. So as we get more renewables, it's really nice to have a zero carbon producer that provides power around the clock. That's beautiful. It is scalable. And so what

we see happening is first, the existing nuclear plants will be. It is scalable. And so what we see happening is, first, the existing nuclear plants will be re-licensed. So they'll stay operational for a long time. And if you go visit a nuclear plant, I think everyone who visits it are amazed at how well-maintained these nuclear plants are. They're extremely well-maintained. They're often operated by former U.S. Navy nuclear submarine officers. It's not like they show on The Simpsons, isn't it? You're not going to see Homer over there with his feet up on the desk doing something zany. Yep, that's exactly right. The maintenance is unbelievably good.

Yep, and then from there, I think we will build quite a few new nuclear plants, but we should talk about the type of plant. I guess there are kind of broadly three options. You could build another of these giant units. You could build what are called these small modular reactors or SMRs. Or in the future, you could try nuclear fusion. The technology is not ready. The big units I'm concerned with because we have a long history around the world of those projects being massively over budget and way behind schedule. They're very complex. You could look at the Vogel plant in Georgia as an example,

but you could look at Flamanville in France, Okoloda in Finland, Sandman in China, like everywhere. So those big complex projects I worry about. What I love are these small modular reactors because they're simpler, more straightforward to manufacture. A lot of it can be prefabricated. So instead of a single unit being a thousand megawatts, these could be 200, 300 megawatts, and you just put unit by unit next to each other. Beautiful. And I would expect around the end of this decade and into the next decade in the U.S. and around the world, we will see a huge increase in nuclear power as a result of that technology really getting ready for prime time. It's getting close now, but a bunch of regulators are still reviewing. It's not yet scaled. And I wouldn't want to be necessarily

the first mover, but I wouldn't mind being the third or fourth mover in building out those SMR nukes. When people look at the rise of solar power and they start to envision, okay, how will that grow over time and which is going to be the dominant source of our power? I mean, I'll get that we'll get power from a lot of places, but what is going to be the surprising grower over the next couple of decades? Do you think it will be that small modular nuclear or what do you think is going to become a much more dominant source of power than it is today? Great question. When we do the hard math around the volume of, say, solar, wind, nuclear, natural gas, there's a bit of coal, but not that much anymore these days in terms of new construction. I would say solar is going to be the largest volumetrically. It's very cheap. It's

fairly easy to construct. Wind is growing nicely, but offshore wind has pretty well-documented challenges. Onshore wind, it's harder to find sites. So I would just say wind will be a bit below that. And then nuclear is uncertain. I would expect nuclear to be very important and maybe second only to solar in terms of the growth. important and maybe second only to solar in terms of the growth. Now, one key technology that we need to see evolve, and it is evolving, is energy storage. If you want to see solar and wind grow, especially, and even nuclear, it would help to have more storage. What I like is I'm seeing some amazing innovation that should drop the cost. An example is a private company called Form Energy that's using a technology called metal air storage.

Now, this product may not look awesome. It's big. It's clunky. It's also not terribly efficient in terms of power in versus power out. But the big advantage is it's really cheap. So that technology could allow us to have really long-duration storage. That would allow tremendous penetration levels for solar and wind, and it's going to become increasingly important. You're talking about something that's an alternative to lithium-ion batteries that

you use in laptops or that they use in electric vehicles, and that I've seen some cases where they've tried to put grid-scale batteries of that type in. You're talking about an alternative to that that's a lot cheaper. That's correct. Now, both are going to be needed. I'll give you an example. In California, the batteries are great for, call it, three to eight hours of storage. As solar increases in penetration in California, you're going to have an excess of solar power

in the middle of the day. It would be best to store that power. And then when people come home from work at 5 p.m. and solar power is declining, you release all that power. That's beautiful. But what we also need is sort of seasonal or interday storage. And that would be like literally 100 hours of storage. That's the missing link. But that's where we see some great technology innovation. So we really need both. What are some areas that I haven't asked you about or we haven't gotten to that might be exciting for investors looking forward as companies try to fill in this power gap? As much as we've talked about clean energy, natural gas fired power is going to be

a significant part of the solution as well. And one thing we highlighted in our note is we do think some of these tech companies will get directly involved in building out additional natural gas-fired power plants because they're relatively quick to build. They provide reliable power. So in places like Texas, for example, I think we will see some new natural gas-fired power plants providing power directly to data centers. So it's going to be a bit of a mix. I think natural gas has kind of been overlooked for some of these more exciting areas like nuclear, and I get it. But natural gas is not dead.

It's going to be used quite a bit in this space as well. What about an investor who owns a broad basket of utilities? What's that going to look like? They're getting some of these companies with nuclear exposure, some of these companies with data center exposure, but they're also getting some sleepy regulated utilities. In my area, north of New York City, it's Con Edison. I would never say a bad word about Con Edison. They might be listening. I don't want my lights to start flickering. I'm sure they would never do that, but I don't think of them as being on the forefront of the most exciting themes that I hear. So what if you have this broad

basket of utilities? How do you think that that's going to perform for investors going forward? You know, this may sound maybe a little surprising, but we've been talking all about power and the shortages, et cetera. The regulated utilities will experience some growth, but nothing dramatic in our view. So what we see from those utilities is, and you mentioned Duke Energy earlier, there are many regular utilities where that growth is strong. I mean, high single digit EPS growth for many of them is pretty good considering the low risk of those business levels. But think of those companies almost as battleships. It's hard to move the battleship very much. They've got long-term capital plans that are

pretty well set. And they're also, in many cases, they've not been able to connect data centers that quickly. So the solutions we've been talking about, like with competitive nuclear power plants, Bitcoin, et cetera, is to some extent sort of taking an alternative approach to the traditional grid interconnect that would boost utility growth conceptually. So I see, you know, that said, utilities are growing very nicely. I don't see these dynamics dramatically changing most of those companies, a few exceptions. But yeah, generally, everything we've been talking about will not have a huge impact on growth of those utilities.

I hear once in a while from students who listen to this podcast. I'm shocked by it. It's a college or a high school kid listening to an old fogey like me. But if there's someone out there like that and they're thinking, you know, I want to be involved in AI. I'm going to go and I'm going to study computer science and technology and this and that. involved in AI. I'm going to go and I'm going to study computer science and technology and this and that. What about this? It sounds like the field of power is going to be increasingly important. I mean, what do you think about somebody charting their school path and learning about, going to learn about power, but thinking about working someday for a utility or something like

that? Is that going to be a good field to be in? For a student thinking about a job where you can be creative and you can innovate, where there's going to be huge demand, and also in the process, hopefully you can try to decarbonize the planet, this is a great sector to think about. I'll give you a few examples. If you wanted to directly apply AI skills, one interesting way to do that would be material science for batteries. So just as we've seen AI breakthroughs in drug discovery, it's actually very similar in terms of battery chemistry. AI will help us think through a huge set of possible chemistries for batteries. That would be, and if you could crack that and improve batteries, huge benefit to the planet and you'd make a lot of money in the process. That'd be one direct approach. Another would be thinking about more versatile and resilient grids to handle everything from hurricanes to extreme heat to extreme power demand from data centers.

There's going to be a huge innovation necessary there. So yeah, it's a wide open field. It's not well known, especially among students, but it's worth spending some time thinking about that. Thank you, Steven. And thank you, Ronnie from San Diego for the excellent question. And thank you, Simon and Simon. Those are the eighties TV detective twins. Exactly. The ones Jackson Cantrell is our producer. You can subscribe to the podcast on Apple Podcasts, Spotify, wherever you listen. If you have a question that you'd like answered on the podcast, like Ronnie from San Diego, you can tape it on your phone.

You can use the voice memo app. You send it to jack.how. That's H-O-U-G-H at barons.com. See you next week.