Artificial Intelligence (AI) Takes a Lot of Energy

In this podcast, Motley Fool host Ricky Mulvey and analyst Asit Sharma discuss:

• How big tech is dealing with the electricity demands of AI systems.
• Companies that could benefit from more energy usage.
• If Under Armour can turn things around.

Then Brendan Hughes, the author of Markets in Chaos, joins Ricky to discuss why he’s concerned about the level of money printing in the United States, and one area for investors to watch.

To catch full episodes of all The Motley Fool’s free podcasts, check out our podcast center. To get started investing, check out our beginner’s guide to investing in stocks. A full transcript follows the video.

This video was recorded on Sept. 16, 2024.

Ricky Mulvey: Chatbots sure are thirsty. You’re listening to Motley Fool Money. I’m Ricky Mulvey. Joined today by Asit Sharma. Asit, how’s your weekend?

Asit Sharma: It was long and not long enough. Aren’t all weekends like that, Ricky?

Ricky Mulvey: I really like getting topic suggestions from listeners, giving us something to talk about on the show and one of our listeners, Previn wrote in to us, recommending that we chat about the massive power demand and grid strain that’s coming from these AI companies. This is apropo since last week, Jensen Huang and Sam Altman met senior White House officials to discuss infrastructure needs for AI projects. Asit, that means power. Setting the table just so we get a baseline understanding is that according to a Goldman Sachs report, a ChatGPT query needs 10x the amount of electricity that a Google search does. First question, Asit, why are these AI systems so thirsty?

Asit Sharma: Hey, Ricky, first, I want to say that I used much less energy to answer this question than you probably did, because I think you told me you consulted ChatGPT on this question. You generated some energy. Think about a Google search, when you or I type something into that Google Search bar, we’re using some computation. We’re sending some algorithms into action. But basically what we’re doing is consulting this index of knowledge. While that’s a little power hungry, it’s nothing like the experience of you asking ChatGPT a question. What’s going on there?

We’ve all heard that large language models are these big neural networks that are self training, but also human-trained. That’s one part of what’s going on. You have a model that’s being consulted by a program. But the larger part of that energy usage is in something called inference. Inference is simply asking the large language model to make an iteration of its prediction in order to answer your question. Crazily enough, Ricky, it comes down to math. When you ask ChatGPT to make a prediction, it takes a lot more precise math and a lot more complex math for the model to consult itself, to weave between these layers of a neural network and then put all the statistical inference together. To spit out this question that makes sense in a sentence, and even contextually sounds like something that is a decent answer to your question. That is such a more complex problem to solve. It’s like the difference between me asking you to turn around and tell me what’s in that shop window, or to sit there and I describe this great toy for you and you try to explain back to me what that toy does. You consume so much more power if we are forcing ourselves to think to use our own neural networks, and that’s what’s going on here.

Ricky Mulvey: I think you pretty much have it right. You’re either sending someone down a path of search engine algorithms or you’re building something from scratch based on previous questions that you’ve asked the model and also it’s inferences based on what you’re saying in the billions of combinations it has to do to get that code. Now we get to the big meeting at the White House. Why are Sam Altman and Jensen Huang there? Well, it’s because, in some ways, they’re trying to get maybe a little bit of cheddar get more energy subsidies for these AI companies. They want help on public and private investment. Of course they do. I’m not going to comment on that. But also, the White House is trying to get these AI companies to use “clean and reliable sources.” Another set the table thing is that electricity demand is expected to rise 15-20% over the next decade. Data centers asset could consume almost 10% of all US electricity demand. Who wins from this? Is it the renewable companies that the White House would like to win from this, you think?

Asit Sharma: I think some of the renewable companies will win. I don’t know if that’s going to be the hugest win out there, Ricky. Renewable energy certainly has a role in reducing the consumption peak demand. But it’s more indirect to me. I don’t think, for example, solar panels are going to be able to supply in this crazy need that data centers have. However, think about a company like Enphase which has a technology called the micro inverter. What that helps Enphase do for customers is to have modular electricity configurations. In a world, in the future, where power is becoming decentralized, because our centralized power is dedicated to serving all these data center needs, for example, then you’re going to have smaller and smaller grids. Companies that can help facilitate the building of smaller decentralized grids where you or I might power our homes because it just costs more from the big providers. Companies like that, I think, could be surprising winners here. In that sense, renewable has its role to play, but it’s not going to be the end all and be all of this investment theme.

Ricky Mulvey: Maybe this is just because I’ve been watching Industry Season 3. I don’t think you’re not watching Industry right now, are you?

Asit Sharma: No, man. You’ve got me with three or four series still on my to do list to watch. I can’t keep up with you, but I’m going to add it to the list. I will add it to the list.

Ricky Mulvey: You got to put Industry. I know we were talking with Mary earlier about some Netflix stuff. But Industry goes to the top of the list, the show rules. A lot of the theme of this season is just skepticism around these environmentally friendly companies and the whole ESG play and how many of those companies turned out to be not what investors hoped and they weren’t the real fundamentals behind it. I have that in my brain right now as we talk about this.

Then also, there’s this core part of me Asit, which is, I would love to use all renewable energy. But if you ask me at my core, what do you want? I want cheap electricity. I think that’s going to come from a lot of the legacy energy providers, especially is you need that baseline infrastructure to keep up with all of the demand of these data centers. Who are you watching that may benefit is a legacy player from the increased demand in AI electricity use?

Asit Sharma: Well, Ricky, I’ll go with what I know best. Duke Energy is a company that’s in my backyard. They’re in Charlotte, North Carolina. They’re one of the largest energy producers in the United States. I think to me, what’s interesting is just the unprecedented power demand they see coming down the road from the forces that you’ve described. As we’ve had this conversation, I think the CEO recently in a conference call, use the word unprecedented to describe what those demand loads are going to look like. They’re working with major hyperscalers like Microsoft, Amazon web services to figure out how they can efficiently serve all this data center need in the future. But there’s something else really interesting that I think is going to benefit these bigger regional providers, and that is the Inflation Reduction Act. There is a ton of investment in the United States going to the semiconductor industry, and that itself is super power hungry. As we try to figure out how to make more efficient semiconductors and build some of that stuff here, guess what? That needs power. I like some of these legacy providers for members who are listening. If you’ve got one in your backyard, that’s a major provider to several states. There’s a good chance they’re going to benefit from this tailwind.

Ricky Mulvey: We’re not just doing a list of companies here. I think there’s something for stock investors of all types to be optimistic about here. I’ve been reading this book called Question of Power by Robert Bryce. It’s actually a suggestion from a listener named Dave. Dave, thanks for that suggestion, too. There’s a chart, and there’s a very close correlation between power usage and a company’s GDP. In some ways, I see this is something that’s good. We have some headwinds in terms of population growth, aging populations that could hurt the economy. But this is something where there’s going to be a lot more power demand, and there’s a pretty close relationship between power demand and economic growth in a country. Maybe ultimately, through the strains, through this innovation, through what companies are looking for, this is a good thing for the economy if we’re looking to use more energy.

Asit Sharma: I agree. It’s a good problem to have. For developed countries like the United States, which lead in terms of so many different services and manufacturing sectors, although I’m sure we’re going to get some mail now about how the decline of US manufacturing is also a big theme. That aside, it does show what a company’s development stage can be in the future. We’re an advanced economy going to something more advanced. We’re talking about semiconductors, ChatGPT, large language models. We haven’t even mentioned so much investment in things like electric vehicle takeoff, lift off. I’m getting the acronym all mixed up for eval because it’s a Monday morning. Having said that, I think that’s a good point. Also we can look at the relationship of a company’s GDP to its sources. Are you a small country that’s using more energy? You have rising GDP, but it’s all coal powered? Are you an advanced economy like the US who’s trying to really diversify those sources, which goes back to the reason you want to talk about this. The US is trying to incentivize these companies not to have the least efficient forms of energy and to try to have some cleaner sources mixed in there.

Ricky Mulvey: The wild card that we haven’t talked about yet is nuclear power. Earlier this year, Amazon bought a 960 megawatt data center for 650 million in Pennsylvania that runs on nuclear power. That amount of power, 960 megawatts, that’s enough to electrify 800,000 households simultaneously. You also have Sam Altman, who doesn’t just do OpenAI. He’s also the chair of a company called Oklo. We had the CEO on the show, but they’re building small nuclear reactors that could very well power data centers all set. You’ve also got Microsoft maybe separately investing $10 billion in renewable energy capacity. Are you looking at nuclear? This is something on my wildcard list that I haven’t put any money into yet.

Asit Sharma: I think it’s on my wildcard list too, Ricky. I’m not looking at it closely, so I’m not scouring the investment globe for opportunities here. I think it’s an industry that needs to mature and there’s all types of flavors. You talked about the company that Sam Altman is working with. Microsoft itself, which is allocating so many billions, has a different approach to any of the companies you mentioned instead of small modular reactors. There are signs, they’re looking at something called mini modular reactors. It’s a nascent sub industry right now, but I’m fascinated by it. I think this quantum computing, these are all markers of where the future is headed and the enormous capital for any one of these ideas that’s required to get it out of that NAS stage, something we can invest in and compare companies to another. There are not a lot of companies right now, pure play that you can compare the financials to, let alone any revenue yet because so much is still development stage.

Ricky Mulvey: Last week, we talked about a turnaround at Meta. There’s another company trying to brew one up with a co founder, and that’s Under Armour. I was supposed to talk about this with David Meier last week, but we’ve been talking turnarounds lately, and there’s a good Wall Street Journal feature on Under Armour, especially as Kevin Plank is back in the CEO seat. This story Asit is not rosy. “More than half a dozen former executives said Plank was responsible for much of the turmoil and complexity that he is now promising to clean up in recent years. They say he filed marketing plans put in place by others. He pushed product ideas that flopped in blurred lines between his brand role and that of the CEOs who succeeded him.” We’re talking about a CEO who has had trouble letting go. Maybe some shades of Disney in there, Asit. But after reading the story, do you think Plank can turn it around at Under Armour?

Asit Sharma: I have no idea, but I’m going to air on the side of an opinion here and say it’s going to be a very difficult road. This is a company that plays in a market, which is exceedingly complex. We were chatting you me, Mary, and our producer Dan Boyd, just before we started taping, and I was like, talked about metal last week, that’s a much easier turnaround to turn around a tech company versus a global brand company in the athletic industry. The reason is you need a few things to succeed here, besides a lot of capital in the big marketing budget. You need a product which has athletic credentials, technical credentials, there’s a demand there people want to use it because they perceive it’s better, it gives them an edge. But you also have to have that brand love. You need great endorsements or brand that really resonates with folks. Those two things are the hardest to put together and do it scale for an extended period of time. Companies like Nike and Adidas, we’ve talked about in the past, Nike, they can stumble and still be around. Smaller brands can be upstarts like Lululemon and On shoes. Those right now have a lot of mojo and momentum. But if you fall off the horse in this industry and you’re not either of those two categories, you’re in the middle, it’s very tough and there’s a lot of history here that I’m sure will make some listeners skeptical that Kevin Plank can turn his baby around.

Ricky Mulvey: I wonder if this only works one way for a consumer brand Asit, which is that you can go from premium to non premium, but I don’t know if you can walk the other way. According to the Wall Street Journal, or this is from Piper Sandler, excuse me, Under Armour, since 2020 has consistently been the Number 1 brand that upper income male teens say they no longer wear. That’s tough to get that back once you lose that customer base and it’s tough to go from a non premium brand, even if you’re cutting your product line to becoming a premium brand, if people associate you with discounting to begin with.

Asit Sharma: It’s very true. Then you get into this death spiral of trying to manage inventory. To work your inventory down, to reduce the size of your workforce to close distribution facilities, all of which Under Armour has done over the past several years. To try to get to a point where your costs are low enough that you can start offering price points that are a little higher. It’s really hard to do that because few people will buy the story of a tarnished brand if it jumps up to a high price point again, without innovation, without something that makes those young, let’s call them semi-affluent buyers, spend those dollars on that brand versus the alternatives they have in the marketplace, the companies with the deep pockets, who’ve been hitting them with those great marketing messages. Just so hard to compete if you don’t have all those elements together, Ricky.

Ricky Mulvey: Now, in a couple of years, watch us eat our words as Under Armour becomes a wonderful premium brand that’s doing gang busters numbers, and we’re associating with Lululemon. I think the theme to tie this together from the story from last week, though is that Meta was more able to turn around because it had a business and messaging problem that it could fix due to that tailwind of artificial intelligence. Under Armour has more of a culture problem in a brand problem that might be a lot harder to fix by Kevin Plank.

Asit Sharma: I agree. I actually think the CEO who was only there for maybe a year, Stephanie Linnartz who came from a completely different industry, was doing some of the right things, the hard stuff to affect a turnaround and now you have all of her work that’s getting scrapped. One thing that Kevin Plank does have a good nose for, though, is what the consumer wants. He’s had a couple of misses. As the Wall Street Journal article mentioned, he’s never really kept his influence out of brand decisions. But he might be able to come up with a winner and at least give them some short term momentum, but longer term. I would be skeptical of the ability to have this become a phoenix that rises up from the brand dashes and challenges both the Giants and Yep starts again. It’s going to be tough. The financials are OK I mean they’re not terrible, but this is going to be a long journey.

Ricky Mulvey: Good, please send it. As Chairman.

Asit Sharma: Thanks for being here. Appreciate your time here in sight. Thank you, Ricky. This was a lot of fun.

Advertisement: Ricky Mulvey: Managing your finances is complex. I get it. Off the top of my head, I got five different accounts with financial stuff. Monarch money keeps things simple by letting you see all of your accounts in one single place. Monarch is the top rated all in one personal finance app. You get a comprehensive view of all of your accounts, investments, transactions, and more. Create custom budgets, track progress toward financial goals, and collaborate with your partner. Plus, Monarch automate Smart Money moves to get you closer to your goals. Now get an extended 30 day free trial when you go to monarchmoney.com/fool. Monarch is as powerful as the tools used by certified financial planners. You can customize dashboards and set up automatic rules for transactions. After trying out Monarch for myself, I understand why it’s the top rated Personal Finance App. Right now, listeners of this show will get an extended 30 day free trial when you go to monarchmoney.com/fool, that is M-O-N-A-R-C-H-M-O-N-E-Y.com/fool for your extended 30 day free trial.

Before our next segment, wanted to let you know that if you’re in Denver, Colorado, we have an event coming up this Wednesday. I’m going to have a link to more details in the show notes. It’s with our friends at bigger pockets. Should be a lot of fun.

Ricky Mulvey: Lower interest rates mean that money moves more freely. Brendan Hughes, the author of Markets and Chaos joins me to discuss the risks of having unbacked currency in one area for investors to watch if you’re a little concerned about the level of money printing in the United States.

A lot of the case studies and chapters you talk about have to do with money printing. We have that going on in the United States right now quite a bit. One of the big themes is that when countries print a lot of money and that money is not attached to anything, that can create catastrophic consequences. Let’s focus on the US, because since the great financial crisis, according to the Fred data, the US has more than 2x its money supply. We’ve got inflation, but we don’t have hyper inflation that’s detailed in your book in a place like Zimbabwe. How have we avoided hyper inflation after printing trillions and trillions of dollars?

Brendan Hughes: I want to take a step back for a minute because I think we need to revisit a really titanic moment in history. Everything or a lot of where we are today goes back to this critical moment in 1971, where the United States severed the link between the US dollar and gold. A lot of the loss of monetary accountability, not just in the United States, but in a lot of countries around the world has to do with this because whenever in history, countries have tried to use urban paper money, there’s always been a loss of accountability, and this time has not been any different. I just wanted to provide that as a preface for this discussion. In the 50 year period, after 1971, when the US dollar was severed from gold, the US national debt soared 70 fold and similar things have been going on around the world.

The next big moment was after the global financial crisis, and central banks around the world experimented with this ultra easy monetary policy, and they said it would stimulate growth. But that never happened. The pre global financial crisis period, the average productivity per year was 2.3%. With the ultra easy period, that diminished to 1.1%, and similar things were playing out in other countries. But really, what these ultra easy monetary policies did, was they boosted the markets for assets such as housing, stocks, cryptocurrencies, instead of stimulating productivity, which was the goal. I think that those are just some interesting things to think about. After the global financial crisis, which is getting to your initial question between 2007 and 2017, the Federal Reserve printed about five times as many dollars as had been printed in the previous 500 years. To your question, why have we not had hyper inflation like we had with Zimbabwe or some other countries.

I think it comes back to the confidence in the government, to put it simply, people have more confidence in what the US government is doing than what Zimbabwe did. We can get a little bit more as to what happened in Zimbabwe once we get there. But a lot of it has to do with with the confidence. I referenced in my book, Germany in the 1920s when they had hyper inflation, and a difference between why they were ultimately able to quell their hyper inflation, whereas Zimbabwe has never been able to, Zimbabwe has periodically introduced new currencies and it’s never worked as people had more confidence in the German government, and they ultimately accepted the new currency. My view is, that’s what’s happened here and that’s what’s happened in some other more developed countries.

Ricky Mulvey: You mentioned higher asset prices, higher home prices. Still, we can go to the grocery store and buy loaves of bread with our dollars and not worry about that price changing by the time we exit the grocery store. Even though we haven’t stimulated productivity, we’ve gotten good stock market returns. Why is it a bad thing that we’ve severed this link to gold and just gone to the full faith and credit of the United States government?

Brendan Hughes: Ultimately, when there is a loss of monetary accountability, it usually does not end well. No. We haven’t seen here or in some other developed countries, like a complete loss of faith in the currency or anything like that. But it can’t be ruled out. If a country is constantly spending more than they’re making, ultimately, that’s not sustainable. Now, what is the point where it becomes unsustainable, and people ultimately say, I don’t have confidence in what’s going on. I don’t know. But if you do have at least some link to an asset such as gold, it does provide for some degree of accountability, and throughout history, the classic playbook has always been to return to gold. I think that that’s something that a lot of people with recency bias, don’t really understand or think about. This period since 1971 is actually a highly unusual period. It’s actually, there’s never been over a 50 year period where every country in the world is operating on an unbacked currency. This is really the unusual test period. No, I think the jury is still out as to what will happen, now. This is a long test period, but sometimes it can just take a long time to see how things fully flesh out.

Ricky Mulvey: You really like gold as a store of value. You’re an investment advisor. Is that something you use, whether it’s physical gold or gold ETFs, is a store of value?

Brendan Hughes: Yes. I think gold and some gold linked assets. Gold miners and things like that can be a store of value to an extent. There’s also some gold royalty companies. I would view that as a more attractive store of value than cryptocurrencies. But I don’t spend any time thinking about cryptocurrencies. No, I will provide some, I guess, food for thought on this topic because it is something that is covered in my book to a decent extent. Regarding a store of value, gold has been, whether it’s been used as a form of currency or as a form of jewelry or something like that, it’s been perceived to have value for thousands of years.

Now, the first people who are thought to have used gold in their monetary system we the Lydians around 600-700 BC, and they resided in what is now Turkey. Rome implemented gold into their monetary system around 300 BC. For thousands of years, whether it’s been jewelry or people actually using it in their monetary system, gold has had value. Now, this is very different from cryptocurrencies like Bitcoin is believed to have been invented around 2009. Well, that isn’t even blip on the radar in the grand scheme of history. I just personally don’t know how that’s all going to shake out. But what I do know is that every time or most times that countries have had issues with confidence in the currency and things like that, they’ve always turned to gold, and I don’t think that this time will be any different.

Ricky Mulvey: Brendan, you mentioned gold royalty in that answer. What is that? What are you talking about there?

Brendan Hughes: There are some gold royalty companies like a company like Franco-Nevada. They’ll pay an upfront amount of money, and they’ll get access to a stream, like say they’ll pay an upfront cash amount or in stock, and they’ll get access to 3% of the ongoing royalties from a given mine. They’re not as susceptible to some of the things that gold miners are susceptible to. The gold miners, they take on the risks of actually operating the mine and such. But the gold royalty companies also do have their own set of risks, but it’s a much less capital-intensive business, and it’s an interesting concept. It’s a similar concept, I don’t know if you’re familiar with Texas Pacific Land in the oil space, but Texas Pacific Land is a company. They own acres out in the Midwest, and they get a royalty anytime oil is drilled on their property. Now, companies like Franco-Nevada are different in that they don’t own this land, and it’s not just an ongoing royalty for nothing. They’re providing, like cash and things like Texas Pacific is a bizarre situation where they acquired all this land for like 150 years ago, and then they’re just getting a check in the mail every time someone drills on their land. It’s a very unusual situation.

Ricky Mulvey: As always, people on the program may have interests in the stocks they talk about, and the Motley Fool may have formal recommendations for or against, so don’t buy or sell anything based solely on what you hear. I’m Ricky Mulvey. Thanks for listening. We’ll be back tomorrow.