Editor’s Note: Today, we’re continuing our prediction series with Brad Thomas. And for today’s conversation, we’ll discuss the trade that has moved markets all year – artificial intelligence (“AI”). Is there an “AI bubble”? Can Nvidia hold onto its AI chip dominance? And where does Brad see the best opportunities in AI? Read on for all that and more.
Van Bryan (VB): Brad, let’s discuss AI. It’s the trend that everybody is following it seems. When you think back on 2025, what stands out to you as it relates to AI?
Brad Thomas (BT): The first thing that comes to mind is the ungodly amounts of money the Big Tech firms are throwing at this technology. Just looking at some of the biggest spenders like Alphabet, Amazon, Microsoft, and Meta Platforms, estimates are that total capital expenditures (“capex”) for 2025 will be in the ballpark of $300 billion. That’s roughly equivalent to the GDP of Portugal. And it’s from just four companies.
This doesn’t look to be slowing down. Estimates I’ve looked at show that capital expenditures from those four should be north of $500 billion by 2030. Most of that is going into the AI ambitions of these firms. These numbers are so large that they’re estimated to account for about 1% of American GDP.
VB: Numbers like that have led some to worry we’re in an “AI bubble.” Do you think it’s a bubble?
BT: Wide Moat analyst Nick Ward made an interesting point on that a few months back. In essence, he said that bubbles don’t form when people are worried about bubbles. They form when investors are too distracted to think about bubbles. I think Nick is probably right about that.
The trouble with spotting bubbles is that there’s no one definition for a bubble. The clearest example is the dot-com bubble. And if that’s the measuring stick, I think there are some important differences.
First, some of the biggest movers during the dot-com era were small, unprofitable firms. Companies that reported zero earnings were soaring. That’s just not the case today.
Look at something like Nvidia (NVDA). Its graphics processing units are the workhorses in AI data centers. This is not some startup. The company reported about $32 billion in net income in the third quarter. Here’s what’s interesting: The company’s revenue in the third quarter of last year was $35 billion, roughly equal. So, after one year, Nvidia’s profits were roughly equal to its sales one year before. That’s almost unbelievable growth. You could wait a hundred years, and you might never see something like it again.
Another reason why I’m skeptical of the bubble claim is that the infrastructure to support AI is seeing almost immediate use. You’re probably too young to remember, but in the 1990s, the telecoms spent billions building out fiber networks to support internet usage. The idea was the world would need that infrastructure eventually as the internet gained adoption.
And that was correct… eventually.
But in the near term, a lot of that infrastructure went unused. It just sat there for years. They called it “dark fiber.” I don’t see that today.
The closest parallel would be data centers. I follow the data center real estate investment trusts (“REITs”) very closely. And I can tell you that demand for new data-center capacity is almost insatiable. As soon as one comes online, it’s up and operating, and the tech firms already have their hand out for the next one.
Now, having said all that, that doesn’t mean there isn’t the potential for irrational exuberance from investors. That happens all the time. And it’s why we’ve seen so many stock market pullbacks around this “AI bubble” narrative in 2025.
And that doesn’t mean there can’t be some overbuilding in the data-center space. Because that is happening… at least in the near term.
But just by looking at the fundamentals of the players in this ecosystem, and by gauging the demand for AI infrastructure, I think you’d have to conclude that it’s very different than the dot-com era.
VB: If that’s the case, the investment thesis seems simple. Buy Nvidia. Should investors be considering that?
BT: Nvidia is a wonderful business. I’ve never seen anything quite like it. Could the stock continue higher? Of course. But it’s just not my preferred way to gain exposure to AI.
For one, I like to find opportunities when I can give readers an edge – some dislocation, some overreaction from Wall Street, something that stacks the deck in our favor. There’s none of that with Nvidia. It’s the most-followed stock on the market. There’s nothing I can possibly say about Nvidia that isn’t already being covered by an army of analysts that track it. So, there’s no edge.
For another thing, I always want to find a nice margin of safety when recommending a stock. Looking at Nvidia, the stock trades for about 38 times current earnings. That’s cheaper than Nvidia’s historical multiple, but it’s still about twice as expensive as the S&P 500.
Now, I think you could argue Nvidia has earned that premium. As I just mentioned, the fundamentals are growing like crazy. It makes sense you’d pay up for that growth.
But the share price for Nvidia assumes the company will continue to grow this way almost indefinitely. And maybe it will. But semiconductors are a ruthless industry. I think it’s possible somebody will find a way to chip away at Nvidia’s dominance. If that ever happens, or if Big Tech capex ever pulls back, then the stock gets repriced lower.
VB: If not the obvious names like Nvidia, what is your preferred way to gain exposure to AI?
BT: It probably won’t surprise you to learn that I’m looking at real estate and data centers. An interesting fact is that data centers have been the best-performing property sector over the past 10 years. They’ve returned about 10.7%, on average, over that time. But importantly, that figure was mostly before we saw this huge land rush as a result of AI.
Having said all that, you might be surprised to learn that the data center property sector underperformed this year, down about 14%.
VB: What do you attribute that underperformance too?
BT: Some of it could be sentiment. As you said, there’s been this “AI bubble” narrative all year.
Another is likely rates, which impact all property sectors. While rates have come down this year, I don’t think it was as much as the market expected. Remember, the Fed held its rate steady for months before once again continuing to cut.
And, yes, there is some overbuilding in the sector. It’s not existential like we saw with the telecoms in the dot-com period, but the data-center space might have gotten a little over its skis. And that’s how you arrive at underperformance for data centers.
VB: But you’re still bullish on the sector?
BT: If I’ve learned anything as a real estate analysts it’s that, barring something calamitous, periods of underperformance are typically followed by periods over outperformance. And when I look at blue-chip data-center REITs like Digital Realty (DLR) and Equinix (EQIX), I see world-class businesses, with world-class portfolios, paying world-class dividends, trading with attractive margins of safety… all with a yearslong tailwind thanks to this AI capex boom.
And, so far, we’ve just discussed the private sector. Some very interesting things are also happening with the government’s involvement in AI.
VB: Like what?
BT: Not many people noticed, but the White House released a new executive order in July. At a high level, Executive Order 14318 outlines how the Trump administration is throwing its weight behind the AI build-out. Trump frames the conversation around AI, correctly, I think, as a new Manhattan project, something the U.S. must be dominant in.
In fact, the federal government is going so far as to accept lease agreements with data-center operators to build and maintain these facilities on public land. The government would be, in essence, the landlord for these data center operators.
What this shows to me is that the administration is all-in on the AI build-out. It would be one thing if private industry alone was committed to this project. But the fact that you have the federal government also rowing in the same direction… that seems like a very profitable, yearslong investing trend to me.
Editor’s Note: Tune in Monday for our final conversation with Brad before the new year. We’ll discuss the ongoing mergers and acquisitions boom, new opportunities taking shape in real estate, and what worries Brad most as we enter 2026.
And if you haven’t already, you might find this interesting. As Brad said, the Trump administration is throwing its weight behind the AI build-out, and a handful of real estate stocks are at the center of Trump’s “final deal.” Get all the details right here.
|