AI spending is approaching $1 trillion per year, but will there be a return from that spending?
In this podcast, Motley Fool analyst Tim Beyers and contributors Travis Hoium and Lou Whiteman discuss:
- AI capex trends.
- Housing price declines.
- KPop Demon Hunters and other Netflix content.
To catch full episodes of all The Motley Fool’s free podcasts, check out our podcast center. When you’re ready to invest, check out this top 10 list of stocks to buy.
This podcast was recorded on August 29, 2025.
Travis Hoium: Could AI spending reach $1 trillion by 2030? Motley Fool Money starts now. Welcome to Motley Fool Money. I’m Travis Hoium, joined by Lou Whiteman and Tim Beyers. We’re going to talk today about housing. We’re going to have Lou and Tim cut some of their favorite stocks from a mini portfolio. But let’s start with artificial intelligence. Morgan Stanley recently said that they expect global Data Center spending to increase from $307 billion in 2024 to 920 billion in 2030. Data Center CapEx is driving companies like NVIDIA, Amazon, Alphabet. But they’re not making enough cash to make that investment themselves. Tim, when you see these huge projections, what do you hear, and do we actually have enough cash flow from these companies to spend almost $1 trillion per year on CapEx?
Tim Beyers: We do, and we don’t. Let’s start with the we do. They are committing quarter over quarter, Travis, and these are multiple companies, somewhere between $50 and $100 million every single quarter. That’s extraordinary. I’m sorry. I said million. I meant billion. This is just a few orders of bag which is bigger. This is extraordinary amounts of money. On the one hand, yes, but on the other, this is a market that is completely out of sync. The buildout of hardware is so extreme that the infrastructure to support all that hardware just isn’t in place yet. I know you’ve covered energy quite a bit over the years, Travis, and I don’t see how you don’t get to a point where there is a little bit of slowdown in the hardware buildout, and then you do some catch-up around energy infrastructure, around environmental infrastructure, around city planning, urban planning. How do you get all of this done in a way that actually creates sustainable growth? That doesn’t seem to be part of these projections, and I think that’s one of the flaws.
Travis Hoium: It does seem like the numbers are just we’re going to keep increasing this. Going from the ChatGPT moment, that was in late 2022, to where we are today, there has been a massive amount of growth in AI spending in spending on things like NVIDIA’s chips. But we’re still learning what these business models are too. We talk a lot about chips, but there’s more to it than this, than just spending money on these power-hungry chips. What else are they going to be spending money on?
Lou Whiteman: With all respect to Morgan Stanley, I do think that it’s smart research. But humans, we are terrible at recognizing cycles, recognizing pendulums for being pendulums. I feel like some of this is just taking what we’re doing today and assuming it into the future, and not considering a swing back. We’ll see about the actual number, but we’re going to spend a lot of money. I do think that at some point, Tim mentioned energy, we’re going to have to spend money on ways to be smarter about energy consumption or building out energy, so that’s part of it. I do think we’ve seen all of this hiring, this poaching, maybe a shift from building up this hardware to getting the brains that know how to use it. I wouldn’t be surprised if, at some point, just instead of just pure hardware and all this data center spending, that even if we still spend a significant amount of money, the spending gets spread out in ways that we’re just not seeing right now at the initial buildup.
Travis Hoium: Tim, the way that you’re talking about this reminds me a lot of the late 1990s and the telecom buildout. There was a dual bubble. We talk about the dotcom bubble. That was a dotcom stock bubble, but there was also a telecom bubble, which is basically these companies spending incredible amounts of money to build out the fiber that we still use today. Google bought up a bunch of that dark fiber. That’s one of the reasons that they have as good infrastructure as they have. What you’re saying reminds me a little bit about, you know what? The numbers are just going up so fast. Demand is going up so quick, and we hear this from every AI-related company that we’re just going to keep building and keep building at some point, there has to be a business model behind it. There has to be return on investment. If we’re talking about $1 trillion of investment, you’ve got to have some profits coming from that. Most of these companies aren’t profitable. Is that a good analogy for thinking about it, or is there a big difference in this buildout versus that telecom buildout?
Tim Beyers: You would hope there’s a difference. I don’t know that there’s a difference, and I think there is a genuine fight over the right economic model for AI, particularly any commercial AI. I think you have two ways to fight the portal fight. What I mean by the portal fight is, I think we are getting to a point where the next interface for computing is likely to be a chat interface. It’s likely to be some kind of, hey, ChatGPT, do this thing for me, or search for this thing, or whatever it is. Some chat interface. Now you’re going to have ChatGPT anthropic, companies like that that are native to the business of building up a chat portal. That’s their primary economic engine, and they want to build things that make that chat engine economically viable. Then you have a competing idea, which is the search model, the traditional web model, and then putting on top of that a chat portal, and that is Alphabet. That is Microsoft. That’s even Apple with Siri. Those two ideas are going to compete. There’s going to be a real fight to figure out who wins in that model. There’s going to be lots of trial and error here, Travis, between is it all going to be driven by advertising? Is it going to be driven by data access? Is it going to be driven by new tools to make different kinds of software that run from a chat interface? But those different types of companies converging to fight a battle to win the portal War, I think, is something we’re still in the infant stages of seeing that. It’s barely started. But that’s a big piece of this story that we aren’t talking about yet, but I promise you in the next 18 months, we’re going to be talking about it.
Travis Hoium: I saw somebody compare the moment that we’re in in artificial intelligence to the Motorola Razor moments. That really stuck with me. That was my favorite phone, I think 2005. But that was obsolete two years later.
Lou Whiteman: I love that phone. But this is such an important part of the AI conversation for us as investors to have because, Travis, to your point a lot of fortunes were lost on that infrastructure buildup, but it was still value adding over time. All of this money is being spent. I think it is adding value. There is a there there. This isn’t just crazy spending, but if history is a guide, that does not translate to every one of these investments will be a winner. As Tim says, there will be a period of figuring out who’s the winners, who’s the losers. I think there could be a lot of winners that aren’t spending the money, just using AI. Just to use one example, MongoDB was up 30% post earnings today on a huge surge of customers attributed to AI. There are going to be a ton of winners. They’re going to be losers, and we’re just so early. If nothing else, you just don’t put all your eggs in one basket here as an investor.
Tim Beyers: Can I add something there, Travis? Just quickly. One of the things that’s common about that. Now, we can’t be sure that MongoDB is going to be a durable winner here. But one of the things that’s true at least today about that MongoDB result is that they sit in one of the categories that historically, over time, you were likening it back to the dotcom bubble, the telecom bubble, the things that did endure from those periods, at least over time, it took some time to wash away all of the excess. But the companies that didn’t go away had real picks and shovels that they could rely on and build upon for the revolution to come. MongoDB is in that space. They’re not the only ones. If we’re an investor that’s looking to profit from the time that we’re in and the excess that we’re in, please, for the love of God, don’t just look at NVIDIA. Look for the picks and shovels. MongoDB might be one. We can’t say for sure that they will be, but they might be one. Any company that is in successful data management is one to at least consider.
Travis Hoium: One of the problems with the telecom buildout was the debt that those companies ended up taking on that increased their risk of their business. Right now, we have Amazon, Microsoft, Alphabet, and Meta generating almost $500 billion in operating cash flow. They’re spending about 365 on CapEx. That’s a projection for this year. They’re using almost all of their operating cash flow on CapEx, in other words. Are we going to get to a point where they’re going to be going into debt to build out more AI solutions?
Tim Beyers: They might, but they certainly have the cash flow to service that debt. You could see it. I’ll make a reckless prediction on this, Travis, you won’t see that. You will see a relentless focus on efficiency first because there is a software side of the AI equation that we haven’t figured out yet. Right now, most of these models are very dumb. They use a lot of tokens. They just burn through all kinds of energy. Almost indiscriminately, that can’t last. There is real engineering work that is happening and will continue to happen to make models, tools far more efficient. I expect, Travis, that you’re going to see a lot of focus on that at the labs at AWS and Google Cloud Platform, and at Microsoft. I just don’t see how they get around that because ultimately, those companies want to see more software built. If you want to build more software, you need better tools, and there just is no way to get around the need for clever, efficient software engineering. That’s just common.
See More
