![Large Size Player Large Size Player](https://www.moneytalkgo.com/wp-content/themes/electron/views/tdtheme/img//icon-large-player.png)
The rise of artificial intelligence has led to a boom in stocks related to data centres and chipmakers. Juliana Faircloth, VP for Portfolio Research with TD Asset Management, and Evan Chen, Associate for Portfolio Research at TD Asset Management, make the bull and bear case on whether this surge in demand is creating a market bubble in the tech sector.
Print Transcript
[MUSIC PLAYING]
The rise in computing demand brought on by artificial intelligence has led to a boom in stocks that are tied to data centers and semiconductors, but some investors may be wondering if there's more room to run or if we're in a bubble.
Joining us now to have that bull-bear debate on the topic is Juliana Faircloth, VP for Portfolio Research with TD Asset Management, and Evan Chen, associate for Portfolio Research at TD Asset Management. Welcome to the program. This is our first two-guest program on Money Talk Live. And we're going to have a debate. So I think we're going have a bit of fun.
Thanks for having us.
We're going to start with Juliana right now. Let's talk about the lay of the land when we're talking about data centers and AI demand.
Sure, it's very timely. NVIDIA, just this week, overtook Microsoft as the most valuable company in the US. So certainly a lot of focus and excitement around AI and around this trend. I'm sure all of your viewers have heard quite a bit about it.
If we roll back the clock, this really started in 2022, when ChatGPT came about. This, I think, really opened our eyes to what AI could potentially be and brought it closer to the future. What's changed over the last year and a half since late 2022 is that this theme has really broadened out.
And the market has realized that in order to get to an artificially intelligent future, there's a lot of physical infrastructure that's needed. And that's data centers. Data centers are the buildings that house all of the computing power that makes AI possible. We've seen a lot of stock movement around this trend and theme.
Something that really stood out to us was just how much companies, really across the board, are talking about data center as a theme. So we've brought in a chart that shows a count of mentions of the term "data center" on S&P 500 earnings calls. And we've taken out technology, and we've taken out real estate.
This theme has been topical in those sectors for a long time, but you're left with consumer, industrial, utilities, energy, every other sector. And if we annualize the rate of data center mentions that we had in Q1 2024 for the rest of the year, you'll see there that bar on the right-hand side is just an explosion. So everyone's talking about it, companies are talking about it, and so we should be talking about it, too.
So we are going to talk about it, but of course, when you see this kind of explosive growth-- and we've seen it reflected in the share prices of companies, NVIDIA taking the number-one spot as the most valuable company in the world-- some people, Evan, will say, is it a bubble or is it not a bubble? I think you're going to make the case for this not being a bubble. So why should we not think it's a bubble?
Yeah. So I'm going to make the bull case. And there's really three reasons why. So the first is that we're spending productively. The second is that we often underestimate the potential of new technologies. And the third is that there's barriers to growth, and there's bottlenecks there.
So let's look at the first one. We're spending productively. So I don't mean you or I. I don't have that type of money. But in CapEx terms-- so for the big three hyperscalers-- that's Microsoft, Amazon, and Google-- they're going to spend around $150 billion this year for a 10-year CAGR of about 25%.
So what does $150 billion buy you? It buys you 13 aircraft carriers, 17 nuclear power plants, or a lot of data centers. So at the same time, revenues are $200 billion for this year. And that's pretty impressive. That's a 40% 10-year CAGR only on the cloud division. So this spending has actually come with revenue, and is a really good return on capital at scale.
So the second reason is that we often underestimate the potential of new technologies. And I don't mean technologies like the Segway or the selfie stick. I mean big, transformational, infrastructure technology-- so internet, rails, electricity, things like that.
So I want to take us back to the 1800s in the UK. This is when the rails started crisscrossing the United Kingdom. And it really brought with it an economic boom. So in the 50 years from 1800 to 1850, global manufacturing and exporting doubled in the UK worldwide, cementing them as a superpower at the time.
As well, GDP growth was 1% higher in those 50 years than it was in the past 50 years. So why is this the case? Well, it was really three reasons. The manufacturing hubs could become bigger and more connected. The second is towns could come up in places that just weren't viable before without this fast travel. And the third is that communication and coordination became easier as messages could crisscross the country.
Lastly, there's limitations to this growth. So the demand is far outstripping supply, but we can't build these things as fast as we want. I'm going to have two examples. So the first is Eaton. They manufacture electrical equipment. And their backlog has doubled since 2020.
As well, another company, Quanta Services-- so this is an electrical engineering firm that specializes in power supply. They're just gated by skilled labor. They've mentioned it's a scarce resource, and you can't just go out and buy experienced engineers who want to work in the field to plug everything together.
Now, this is just an example of some of the bottlenecks that exist. There's others, such as permitting, energy availability, real estate availability. And all of this slows down the buildout. So overall, we're spending productively. We're underestimating the potential of AI and the cloud. And there's bottlenecks to growth that prevent a bubble from forming.
You've nicely laid out the bull case. Juliana has been listening to you lay out the bull case. So Juliana, you're going to make the case for why we should view this as a bubble.
I've been listening intently. Evan did a great job. There is always a bear case, so let's walk through that. I also have three reasons why we may be in a bubble. The first is we tend to overbuild. The second is that we're already overvaluing some of these assets. And the third is that we might be overhyping AI.
So to start with, we've overbuilt. Evan mentioned we find it informative to reference prior bubbles that we can look at in history-- so the railway bubble, the internet bubble, and now we're questioning whether we're in a data-center bubble.
With railway and internet, what we saw was a huge and very rapid increase in capital spending by British railway companies and by US telecommunication companies. This created a huge overbuild of infrastructure, whether it was rail or fiber optic networks or internet capacity.
There was an overbuild. And not to say that this didn't ultimately end up being productively used, but there was a bit of an air pocket. And so that's the risk that we can be worried about from a data center and AI perspective is that there's a little bit of an air pocket to some of this spending.
The second is that we may already be overvaluing this theme. We talked at the beginning of this segment about the number of companies that are mentioning data center on their earnings calls. I think it's worthwhile to question whether all of these companies are really that involved in data centers.
We have seen instances of companies where they generate a low single digit, so very small percentage of revenue, from a data-center-exposed type of machinery or equipment. And we've seen those stocks get a lot of support on the back of this theme. So that type of excitement and overvaluation is something to be cautious about.
And thirdly, we might be overhyping AI. We don't really know what the future holds. For now, AI has no killer app. ChatGPT was really popular. It exploded in late 2022 and early 2023, but it's really plateaued in terms of usage. I don't think it's necessarily massively changed our day-to-day lives.
And the other element that might keep a bit of a lid on the growth potential for AI is regulation, whether that's ESG-related climate regulation-- data centers are quite energy-intensive-- whether it's regulation around job security or managing employment levels. We don't know what the future holds. And this does seem like a technology ripe for some sort of regulation.
I love how you brought up the Segway earlier. As a young journalist, I remember all the hype leading into the announcement. And then the day of the Segway, you're like, that's it? Just a little thing you can drive around? Great layout of the bull and the bear cases. Evan, what are the investment implications here?
Yeah, so whether we're in a bubble or not, we've laid out some of the risks, some of the mitigants to data centers and AI. And really, this is a broad theme. So across sectors, across business models, there's a lot of opportunity there for things that are adjacent to data centers.
It's not just tech. It goes further. And what we've done is bucketed stocks into three different groups with data centers. So the first is tech and tech-adjacent. The second is infrastructure. So this is utilities, industrials, and real estate. And the third is energy materials. So let me go through them.
First, the tech and tech-adjacent-- this is the most obvious play that you can think of in data centers. So this is NVIDIA, it's Microsoft, it's Amazon, it's Google, all the things that go into a data center and the people who build them.
The second is that infrastructure layer. So in industrials, people who sell the power, people who sell the cooling racks, so Vertiv. There's the Quanta Services I mentioned earlier. Utilities that have power demands are going to grow over time. And they're going to benefit.
And lastly, there's real estate. You've got to put all these things somewhere. So these companies should benefit. Lastly is energy and materials. They need a lot of energy. They need a lot of power. So natural gas has done well, and nuclear has done well on the potential of we're going to attach a nuclear power plant to one of these data centers.
Materials-- we need copper to wire everything up and create better connections with the power grid. So while there's a lot of debate around if we're in a bubble or not, we think this is an important theme to consider for all investors because it affects everything.
[MUSIC PLAYING]
The rise in computing demand brought on by artificial intelligence has led to a boom in stocks that are tied to data centers and semiconductors, but some investors may be wondering if there's more room to run or if we're in a bubble.
Joining us now to have that bull-bear debate on the topic is Juliana Faircloth, VP for Portfolio Research with TD Asset Management, and Evan Chen, associate for Portfolio Research at TD Asset Management. Welcome to the program. This is our first two-guest program on Money Talk Live. And we're going to have a debate. So I think we're going have a bit of fun.
Thanks for having us.
We're going to start with Juliana right now. Let's talk about the lay of the land when we're talking about data centers and AI demand.
Sure, it's very timely. NVIDIA, just this week, overtook Microsoft as the most valuable company in the US. So certainly a lot of focus and excitement around AI and around this trend. I'm sure all of your viewers have heard quite a bit about it.
If we roll back the clock, this really started in 2022, when ChatGPT came about. This, I think, really opened our eyes to what AI could potentially be and brought it closer to the future. What's changed over the last year and a half since late 2022 is that this theme has really broadened out.
And the market has realized that in order to get to an artificially intelligent future, there's a lot of physical infrastructure that's needed. And that's data centers. Data centers are the buildings that house all of the computing power that makes AI possible. We've seen a lot of stock movement around this trend and theme.
Something that really stood out to us was just how much companies, really across the board, are talking about data center as a theme. So we've brought in a chart that shows a count of mentions of the term "data center" on S&P 500 earnings calls. And we've taken out technology, and we've taken out real estate.
This theme has been topical in those sectors for a long time, but you're left with consumer, industrial, utilities, energy, every other sector. And if we annualize the rate of data center mentions that we had in Q1 2024 for the rest of the year, you'll see there that bar on the right-hand side is just an explosion. So everyone's talking about it, companies are talking about it, and so we should be talking about it, too.
So we are going to talk about it, but of course, when you see this kind of explosive growth-- and we've seen it reflected in the share prices of companies, NVIDIA taking the number-one spot as the most valuable company in the world-- some people, Evan, will say, is it a bubble or is it not a bubble? I think you're going to make the case for this not being a bubble. So why should we not think it's a bubble?
Yeah. So I'm going to make the bull case. And there's really three reasons why. So the first is that we're spending productively. The second is that we often underestimate the potential of new technologies. And the third is that there's barriers to growth, and there's bottlenecks there.
So let's look at the first one. We're spending productively. So I don't mean you or I. I don't have that type of money. But in CapEx terms-- so for the big three hyperscalers-- that's Microsoft, Amazon, and Google-- they're going to spend around $150 billion this year for a 10-year CAGR of about 25%.
So what does $150 billion buy you? It buys you 13 aircraft carriers, 17 nuclear power plants, or a lot of data centers. So at the same time, revenues are $200 billion for this year. And that's pretty impressive. That's a 40% 10-year CAGR only on the cloud division. So this spending has actually come with revenue, and is a really good return on capital at scale.
So the second reason is that we often underestimate the potential of new technologies. And I don't mean technologies like the Segway or the selfie stick. I mean big, transformational, infrastructure technology-- so internet, rails, electricity, things like that.
So I want to take us back to the 1800s in the UK. This is when the rails started crisscrossing the United Kingdom. And it really brought with it an economic boom. So in the 50 years from 1800 to 1850, global manufacturing and exporting doubled in the UK worldwide, cementing them as a superpower at the time.
As well, GDP growth was 1% higher in those 50 years than it was in the past 50 years. So why is this the case? Well, it was really three reasons. The manufacturing hubs could become bigger and more connected. The second is towns could come up in places that just weren't viable before without this fast travel. And the third is that communication and coordination became easier as messages could crisscross the country.
Lastly, there's limitations to this growth. So the demand is far outstripping supply, but we can't build these things as fast as we want. I'm going to have two examples. So the first is Eaton. They manufacture electrical equipment. And their backlog has doubled since 2020.
As well, another company, Quanta Services-- so this is an electrical engineering firm that specializes in power supply. They're just gated by skilled labor. They've mentioned it's a scarce resource, and you can't just go out and buy experienced engineers who want to work in the field to plug everything together.
Now, this is just an example of some of the bottlenecks that exist. There's others, such as permitting, energy availability, real estate availability. And all of this slows down the buildout. So overall, we're spending productively. We're underestimating the potential of AI and the cloud. And there's bottlenecks to growth that prevent a bubble from forming.
You've nicely laid out the bull case. Juliana has been listening to you lay out the bull case. So Juliana, you're going to make the case for why we should view this as a bubble.
I've been listening intently. Evan did a great job. There is always a bear case, so let's walk through that. I also have three reasons why we may be in a bubble. The first is we tend to overbuild. The second is that we're already overvaluing some of these assets. And the third is that we might be overhyping AI.
So to start with, we've overbuilt. Evan mentioned we find it informative to reference prior bubbles that we can look at in history-- so the railway bubble, the internet bubble, and now we're questioning whether we're in a data-center bubble.
With railway and internet, what we saw was a huge and very rapid increase in capital spending by British railway companies and by US telecommunication companies. This created a huge overbuild of infrastructure, whether it was rail or fiber optic networks or internet capacity.
There was an overbuild. And not to say that this didn't ultimately end up being productively used, but there was a bit of an air pocket. And so that's the risk that we can be worried about from a data center and AI perspective is that there's a little bit of an air pocket to some of this spending.
The second is that we may already be overvaluing this theme. We talked at the beginning of this segment about the number of companies that are mentioning data center on their earnings calls. I think it's worthwhile to question whether all of these companies are really that involved in data centers.
We have seen instances of companies where they generate a low single digit, so very small percentage of revenue, from a data-center-exposed type of machinery or equipment. And we've seen those stocks get a lot of support on the back of this theme. So that type of excitement and overvaluation is something to be cautious about.
And thirdly, we might be overhyping AI. We don't really know what the future holds. For now, AI has no killer app. ChatGPT was really popular. It exploded in late 2022 and early 2023, but it's really plateaued in terms of usage. I don't think it's necessarily massively changed our day-to-day lives.
And the other element that might keep a bit of a lid on the growth potential for AI is regulation, whether that's ESG-related climate regulation-- data centers are quite energy-intensive-- whether it's regulation around job security or managing employment levels. We don't know what the future holds. And this does seem like a technology ripe for some sort of regulation.
I love how you brought up the Segway earlier. As a young journalist, I remember all the hype leading into the announcement. And then the day of the Segway, you're like, that's it? Just a little thing you can drive around? Great layout of the bull and the bear cases. Evan, what are the investment implications here?
Yeah, so whether we're in a bubble or not, we've laid out some of the risks, some of the mitigants to data centers and AI. And really, this is a broad theme. So across sectors, across business models, there's a lot of opportunity there for things that are adjacent to data centers.
It's not just tech. It goes further. And what we've done is bucketed stocks into three different groups with data centers. So the first is tech and tech-adjacent. The second is infrastructure. So this is utilities, industrials, and real estate. And the third is energy materials. So let me go through them.
First, the tech and tech-adjacent-- this is the most obvious play that you can think of in data centers. So this is NVIDIA, it's Microsoft, it's Amazon, it's Google, all the things that go into a data center and the people who build them.
The second is that infrastructure layer. So in industrials, people who sell the power, people who sell the cooling racks, so Vertiv. There's the Quanta Services I mentioned earlier. Utilities that have power demands are going to grow over time. And they're going to benefit.
And lastly, there's real estate. You've got to put all these things somewhere. So these companies should benefit. Lastly is energy and materials. They need a lot of energy. They need a lot of power. So natural gas has done well, and nuclear has done well on the potential of we're going to attach a nuclear power plant to one of these data centers.
Materials-- we need copper to wire everything up and create better connections with the power grid. So while there's a lot of debate around if we're in a bubble or not, we think this is an important theme to consider for all investors because it affects everything.
[MUSIC PLAYING]