Big Tech Keeps Pouring Money Into AI. Why Is This Boom Different? | Paul Meeks

Are we about to see some technology companies bankruptcies and investors losing significant amount of money? A lot of people are saying we're like 1999 right now. Or is this time different? Today we sit down with Paul Meek, who's an analyst, sell-side analyst. He's also been in the tech space for many decades. He was running the largest funds in technology during the 99 bubble. I think we will have a bubble pop. We will overspend. And there will be some weak companies that go away for good, but we'll have a correction. And I think it's going to be more benign than the correction we had after the internet bubble popped because. Paul, it's great to have you on. Welcome. Great to be here. I'm tickled pink to have this opportunity. It's great to talk to you, especially today, because there's a lot going on in the markets right now. You know how the bubble of 99 went, and now today we have this AI infrastructure built out, and a lot of people are comparing whether if this is like 99 or not. How are you seeing things? When we got to the internet bubble, it's inflation and then it's inevitable popping. I was in New York managing the largest tech fund in the world, actually six different mutual funds, assets under management, about $9 billion. I started them all from zero only two and a half years before that. So you can see the explosion in asset growth. And then, of course, when the bubble popped, you know, the NASDAQ over two years went down about 80%. But I will tell you what's different this time. What's different this time is that the folks that were spending on all the infrastructure, because they spent a lot then, they're spending a lot now. This time it's even more severe. But back then the folks that were spending on the infrastructure were quite often recently public companies that probably should not have ever gone public. They were not good companies. And so some of them went bankrupt within years of their IPO. The only reason they were able to do an IPO is the investment bankers were very hungry for their offering fees. And so they escorted every private company, even bad ones, to the door for IPOs. So you had a number of companies that were weak. Didn't really have established business plans, no way out of the hole. They take these IPO proceeds, they buy a lot of internet infrastructure, and then when the business starts to sank, they couldn't pay their bills, right? They had very weak balance sheets. And remember, some of these companies were bankrupt in a hurry, whereas today, Not over time, but at least so far, because we've only had the AI infrastructure building since ChatGPT was released in November of 22, so it's still an emerging industry. Spending trillions of dollars, we'll spend trillions more. However, at least to this point, 90% of AI infrastructure spending thus far has come from the hyperscalers. But these are companies that they may now be burning cash instead of generating free cash flow, but they come to the dance with very strong balance sheets. And even though they overspend for a period of years, which I expect, they'll have no problems ever going to the markets and raising either equity or debt capital. And so this time, no counterparty risk or less counterparty So I think we will have a bubble pop. We all do. We're all... as a species greedy people so we will overspend and we will have a bubble and it will pop and there will be some weak companies that go away for good but the bulk of them have very strong balance sheets they can weather the storm meta google and microsoft are going nowhere and so we'll have a correction but when we do have a correction i think it's going to be more benign than the correction we had after the internet bubble popped because less counterparty risk in 2026, 2027, even 2028. So how do you see the current decline? Do you see the current decline not even being, we're not even built up to the bubble part of? Yeah, I tell you the only thing The only thing I'm worried about is these companies have the demand and they'll have demand probably for the foreseeable future to put up these big data centers but sometimes these data centers are funded by equity and if you issue more equity you dilute shareholders more often than not they have to issue debt and what had happened with these companies that are serving the hyperscalers because remember these are fairly fresh companies. I'm talking what we call neoclouds like Nebius and CoreWeave Even though they're supporting the marquee hyperscalar customers, they used to be able to borrow at much lesser rates. But in the last month or two, particularly the last couple of weeks during all this panic. You know, the rates that they're borrowing at have skyrocketed. Now, I don't think it's fair what's happening. I think they're going to be fine in the end, but the market does and says what's going to happen, you know, without my input. And if they can't finance because of the worries about AI infrastructure, then we'll have some problem. It's kind of like a self-fulfilling prophecy. It's like a run on the bank. When there's worry about your local bank, everybody runs to the local branch, and of course, they don't have enough money in the vault if everybody who's a customer wanted to withdraw their money at once. And so unfortunately, even though that doesn't necessarily mean that your bank's a bad bank, when you have pervasive negative perceptions, you could have an issue where you have a run at the bank. And it looks like right now the neoclouds will probably be closed for debt funding for a while. That worries me. What doesn't worry me is their demand and their business models in the long term. Unfortunately, they have to depend on people buying their bonds. And right now, people have said, nah. So we'll see what happens. But I don't think it's a demand problem. I don't think it's a problem with their businesses. We have just such pervasive negative psychology. It's kind of like a run on the bank. Their debt right now is at the mid-teens, from what I know. Is there a level that doesn't make sense for them to issue more debt? Are we there yet for CoreWeave of the world? I think what will happen is, at least for now, It'll change. And I think the restriction on raising senior debt that's on the balance sheet, that window will open up again, but right now it's closed. But I will tell you, there still is probably demand for project level financing through these SPVs, special purpose vehicles, where they only raise money and borrow for a specific data center. of which the cash flows are essentially cash flows from a marquee company like Amazon Web Services so if you're CoreWeave for example and you can do an SPV only for your data center that you're building for Amazon Web Services uh JP Morgan may lend you money based on not your junk credit rating but Amazon Web Services triple A rating and so They'll continue to do project-based financing, data center by data center, but essentially using the cash flow support and the credit ratings of the marquee companies versus putting more senior debt on the company's own balance sheet. I'm talking about the neoclouds like Nebius and CoreWeave. What's the fear? What is causing this, in your opinion, that cutting off these companies like CoreWeave from? from debt borrowing? I think the greatest concern is the macro concern that all this AI infrastructure spending will never yield an adequate return. And then once folks get wind of that, including the hyperscalers themselves, they'll cut off the spending and no more spending on data centers. That's CoreWeave's business model. So I think it's totally overblown. Probably a great time to buy these neocloud stocks right here. I've said so on CNBC just earlier last week, but we'll see what happens. Again, I have no problem with demand. I think demand is going to be voracious for years. I used to say that any available data center capacity would be snapped up through 2028. Now I think it will be snapped up through the rest of the decade. So you have strong demand. You have counterparties. I just think a number of folks have gotten caught up in this negative feedback. And they're just worried about, not necessarily CoreWeave specifically or Nebius specifically about, but call it the 80-20 rule or the 90-10 rule with the bulk of the worries just, oh my goodness, they're spending too much and they're never going to make any dough. I disagree. You seem more optimistic than before. What are you seeing right now that you think this is going to last till 2030 at least? We've just gotten through the quarterly reports of all the hyperscalers, you know, the companies that I mentioned. And to a man, don't mean to sound sexist, but to a man or to a company, not only did they reaffirm their capital expenditures, which, let's face it, is all spent on AI infrastructure, for 26. But they also said, and here is Google talking, we're going to spend significantly, quote unquote, don't know what the number is, more in 27 than 26. And so the spending from the guys that are strong counterparties. It's not moderating. It's accelerating. The only company that actually trimmed their guidance a little bit for 26 was Microsoft. But the only reason was it wasn't demand. They wanted more capacity, but it was a supply issue. Some of its vendors, and it might have been Nebius or CoreWeave or both, couldn't put up enough capacity to fill their budget requirement for CapEx for 26. So they cut it back for a supply reason, not a demand reason. And so think about it, all these companies are spending heavily, they're good for it. They continue to spend because they think that for their businesses, AI will be a big winner. We'll see what happens. But if the customers are saying that, and these are the most sophisticated folks on the planet, I tend to believe them. What about Anthropic and OpenAI? Anthropic is about to go public, and there's some information available from them. What do you think of their information, their financials, their numbers? So, you know, today they issued their second version of their S1. I delved very deeply a week ago into their initial version of their S1, and I actually was pleasantly surprised. Of course, in an S-1 or an SEC document, your lawyers will require you to list every risk known to man. And of course, everybody is laughing now that one of the risks is, oh, there's a certain percentage chance that our technology is going to blow up the world. I'm paraphrasing. But when I take a look at that business on a non-GAAP basis. Even in the midst of all this heavy spending and all these expenses that are elevated as you go from private to a public company with the IPO, Anthropic has about a 10% operating margin now, which I think is a nice baseline that actually pleasantly surprised me. I mean, their strong revenue growth was no surprise because they've been reporting that every month all the way along. But their operating margin baseline, and I expected to go up from here, I actually thought was pretty impressive. And the inference business, so the inference is the stage in the AI food chain where we are now, right? For a couple of years after ChatGPT, we spent all of our time and money training models. Now we're in the inference phase. We're going to use the models to actually deliver some goodies. It looks to me, from going through Anthropic's data, That their inference business will have a contribution gross margin of about 85% and a contribution operating margin of about 40%. If that does indeed happen, that is an attractive business. That's not an average business. That's not a business you worry about. That's actually a super attractive business. And so I'm encouraged. And, you know, they have made commitments of almost $900 billion. To pay all of the companies to put up data centers for them. Again, oh my God, that's a lot of money. But I think they're good for it. And again, they're telling you, we are willing to spend $900 billion over time because we see. As we take a look at our inference model, that this is going to be money well spent. Right now, it's very easy when you see revenues are small and capital expenditures are many, many fold more. People are saying, oh, it's never going to work out. Well, let's see what happens over time. So I was actually incrementally positive, not hugely positive, but incrementally positive about the Anthropic's business and the overall industry when I read that S-1. What about OpenAI? Do you have any thoughts on them? Yeah, I mean, OpenAI will be there too. I'm a little bit worried that Anthropic, which will go public, you know, probably in the next couple of weeks or months, you know, I would expect before Christmas. What happens if they take all the oxygen out of the room for these institutional portfolio managers that say, you know what, I have so much money to invest in an IPO this year. I already spent a bunch of money on SpaceX within the... Models. I think I like Anthropic more than OpenAI. Plus, Anthropic comes first in 26. I don't have to wait until 27. So I'm a little bit worried about OpenAI, not necessarily operationally, but the fact that the oxygen and the dollars might be taken away from them. Because people can invest in everything, even the big funds. And coming public so much later, I don't know when in 27 it's going to be. It might not even be early in the year. That lag from Anthropic could be a problem. And Paul, you mentioned that we will have a bubble at some point. Yeah. You don't think we're there. And some people have said we might overbuild the capacity. And that's what happened in the 99 and the railroad bubbles. How do you think we could end up there? What are the signs that you're looking for? Yeah, I would tell you. Right now, I think it's premature to judge the success or failure of the spending. Even when ChatGPT was released in November of 22, back then I was thinking about who we are sitting in November of 26. I didn't expect, even back then, that we would be generating money or really have any killer apps yet in which to monetize the AI spend. So because it hasn't happened yet, hasn't been a disappointment to me because I hadn't expected it yet. And I actually think 27 will be another year of spending and don't be disappointed if we're still not showing the revenues. But what I see happening is I think capacity will be built hot and heavy. At least through 28, maybe even longer, but not anytime soon because I don't expect it anytime soon. But yeah, let's say as we get closer to the end of this decade, we're going to have to prove like Anthropic is going to have to prove that instead of hoping their inference model has contribution gross margins of 85% and contribution operating margin of 40%, it happens, right? We see it in the results. So, I think at some point, I will agree with everybody else that you have to show me the money. I just don't see it necessary yet. And so that'll crack this thing. And you know what? If and when it happens, now I think it's going to happen further away than people think. Yes, we will be caught with our pants down. There will be overspending, and some companies will go away. But it won't be the hyperscalers, and it won't be OpenAI or Anthropic, or probably SpaceX. So I feel comfortable with those because the technology typically, the strong gets stronger and these guys have the financial wherewithal to stay in the game, keep funding their investments until they see the goodies on the other side. Now, the rates are going up and as you mentioned, there's a lot of money going and a lot of money getting invested into this AI infrastructure. Are the rates impacting the investments? Yeah. So the biggest impact on increasing the cost of borrow for these hyperscalers and the neoclouds and the AI colos, as I call them, everybody in the AI infrastructure food chain, is mostly a widening of credit spreads. That's the biggest culprit. But when the baseline interest rate, of which all debt is priced, is rising, like you see with a 10-year treasury over at 5%, at least recently it got up to 5.2, now it's back below 5, yeah, that's a problem. And it does look to me that even though Powell, probably because he didn't want to be seen as political right before tomorrow's elections, kept rates flat after raising them last time, kept them flat this week. If you take a look at Fed Funds Futures, it looks like we'll be back to raising rates again at the next meeting, right? Because the coast will be cleared after the elections. I'm still hoping that if and when, God be with us, the war ends in Iran, we will have a lessening of obviously the price of crude oil, and that's going to bring down inflation. And maybe we can get back to what our original plan was. Let's lower rates, not raise them. But I tell you, if you're a tech And specifically, an investor in companies like AI infrastructure, where they're in the heavy building phase before they get the revenues, yeah, if rates keep going up, that's a real headwind. Now, how do you see the macro economy? There We've had a lot of growth, but some people say that they're not feeling this growth. is a How are you seeing things? you So we've reported pretty good growth, right? GDP 3% plus, man, that's fantastic, very despite all of the madness worldwide. But I will tell you, I'm a little bit worried, particularly if near-term trends, including much. cessation to the end of the war, happens, because if you peel back the onion, the bulk of the growth in the U.S. economy has come from what we're talking about, AI infrastructure spending. It is really, really, really dependent on that. And so the rest of the economy is not as impressive. And so, wow, you're asking all the right questions because we better hope this thing keeps going or else, think about it, AI is driving technology. Technology is 40 to 50% of the weight in the S&P, so it's driving the markets. And if we have a problem with AI infrastructure spend, It hits the... Wallet of investors. It impacts the economy. We're not growing at 3% or 4%. We're probably growing half of that rate. And then people say, oh, my goodness, that ain't good. And so, yeah, I'm a little bit worried because the numbers look robust. We haven't reined in inflation. Rates are still too high. The war is still going on. And too much of all this, at least superficial look at the economy about how wonderful it is, is coming from... One source, AI infrastructure building. So keep your fingers crossed, not only for AI infrastructure investors like me, but for all Americans, because it's driving the economy. How do you see the inflation? You know, the rates are going up and it's a global phenomenon. It's not just the US rates. Like, what are your thoughts on inflation? No, it's too high. But I would say that because oil is a major component of any inflation gauge, domestic or international, and because it's a global commodity, it is impacting the inflation rates of everybody, right, as a global commodity. I would prefer... The Fed to set a target higher for inflation, you know, because for years they've been saying they want to get either CPI or PPI or the expenditures indicator that they like to use down to two percent growth. Well, despite all the levers they're pulling, I don't know if we ever get there, because when you say that we're going to get down there. And years and years and years go by, and you never do, you lose your credibility. I'm hoping maybe 25 basis points more to the Fed funds rate at the next meeting, and then hopefully we're steady state, but that's gonna depend on the ending of the war in Iran. So you think that 2% is not realistic in the long run? No, because you think about it, they've talked about it for years and they've failed to meet it for years under different Fed chairs, under different presidents, under different fiscal plans and monetary policies. I just think we should retire that benchmark because we're just going to set ourselves up for failure. Now Paul, you have a lot of experience in this field and you have some lessons for us that you must have learned managing all these funds over many years and decades. Tell us about some of the lessons you've learned and right now where it seems like we're going through something you went through in the 90s. Yeah, so when I first got in the business it was in July of 1987. And only a couple of months later, in October of that year, we had the day where the Dow went down over 20% in one day, right? So I was there as a youngster in my 20s, only months on the job when we had the first fiasco. And of course, we've had many since then. I much prefer when things are going great because I'm on TV so much that quite often people associate me with any tech success, where sometimes I get asked for autographs, which is stupid. But then when things go bad, not often, but a couple of times in my life, I've gotten death threats. I prefer the former than the latter. So anyway, I see both ends of the spectrum. When I'm a hero or I'm a goat, neither one of those is real. I'm somewhere in between. But I will tell you, I think over time, the only way you can make money is even as a tech investor, you follow what Buffett says. You have to be a contrarian, right? Buffett was the one that told us, I am greedy when others are fearful. Right now, CoreWeave, I think the company is going to make it. The stock's trading at about 50. I am going to be greedy when others are fearful. So one of the things I learned is stay strong in your convictions. The only way you win is being a contrarian and know that when you see somebody like me on TV or you read these reports from the investment banks, Wall Street research is a money loser. Investment banks only publish it. So they can get in good with the managements of the companies they cover so they can get investment banking business because that's the big money driver. So I will tell you there's a reason that of all the thousands of companies that are covered and analysts like me publish on, you very rarely get a sell rating. Everybody's got a strong buy or buy because they don't want to piss off the CEO or the CFO. Because writing research is not a thing. It's a money loser. All the money comes from deal making. So make sure that you realize when you hear me on TV or you read my stuff that, you know, I might have an ax to grind. I might have a bias. It may not be me talking. It might be the investment bankers. And so what happens on Wall Street is when stocks are running. The analysts want to look good, so they keep on increasing their price targets so they can stay with the name because nobody wants to sell a winner. And they start raising price targets to a level that is just nuts based on fundamentals because they're just doing it after the fact. They're following the momentum. They're afraid to sell. And then the opposite happens, and it's happening right now in some stocks when the momentum is going in the other direction. Somebody that loved a stock at 200, and now it's 20, but the stock's going in the wrong direction. They should be telling you to, I love it at 200. I really love it at 20. But you know what they do? They don't support it, or they even go to a sell rating when they should be buying. So unfortunately, our entire research industry is geared upon following momentum, particularly tech investors. And unfortunately, that gets you buying high and selling low So I try to distinguish myself on being the tech contrarian. And even though I'm a growth investor and Buffett was a value investor, I try to be greedy when others are fearful and vice versa. Because unfortunately, our whole media and Wall Street is set up to chase momentum and actually do the opposite. And of course, when I was a kid, I did the same thing. Now I try to be like CNBC. When they introduce my weekly segment, they say, Okay, here comes Paul Meeks, the tech contrarian. That's my gig. And so I say, be contrarian. And you know what? Identify a good thing. Identify the leaders, the leadership companies in that thing. And you know what? All companies that are publicly traded have to spill their guts every 90 days. It's not really fair. I mean, if you're a CEO, can you really plan for the long term when you have to report to analysts every 90 days? So anyway, of course. Companies are run by human beings and we're all fallible. And so even great companies will have a bad day. They will miss a quarter. And when you miss a quarter in tech, as far as the analyst expectations for earnings per share or revenues, your stock does not go down 2%, it goes down 20%. And if you like the company and you can prove to yourself that it's a transitory problem, not a permanent problem. You take advantage of those. Of course, what will happen is Wall Street says, oh, bad news, bad news, sell rating, and they perpetuate the downside. Sometimes you do want to sell because if a company is permanently impaired, it could be worth zero. But what happens is great theme, great company, had a bad day, it's inevitable, you can buy them. And so I actually look for companies that I like and I wait for them to have one of those quarterly misses because it's coming for everybody. And sometimes it's not them. Sometimes it's a macro variable. But I try to swoop in. And so that's exactly how I play the Wall Street game, which is opposite of everybody else. So anyway, that's kind of my gig. So that's probably the biggest lesson I've learned. Thank you so much, Paul, for being here, sharing your story and giving us some thoughts on what you're seeing out there with the AI infrastructure build-up. It was great to have you on Marketing Insider. I'm looking forward to coming back. We'll see if I'm right with this contrarian call. I'll either be the GOAT, which is lowercase, which means I'm a bump, or I'll be the GOAT, all uppercase, Tom Brady quarterback style. And we would love to have you back on to see, to tell us what you're seeing next. Time will tell, my friend. Hey, thanks for doing this, man. Tell us what you think about the markets right now. Do you think this is like 99 right now? Or do you think this market rally and market expansion has many years to go? Make sure to tell us in the comments. I'm Siyama Khorrami, this is Marketing Insider, we'll see you next time.