Title: Jim Chanos: The AI Bubble Is "Much Worse" Than Dot-Com Show: Risk Reversal podcast (RiskReversal Media; hosts Dan Nathan & Guy Adami) Guest: Jim Chanos (Chanos & Co.) Date: 2026-07-17 URL: https://youtu.be/oM7VevsftrY Length: 46:18 Note: fillers (um/uh/you know/I mean) and stutters/false starts removed; wording otherwise verbatim. Every (mm:ss) line preserved.
00:00 In bull markets, you put a premium on promises, and bear markets, you put a discount on reality. And we're clearly in the former right now. Do we get to the latter? I don't know. There are a handful of companies that are gatekeepers and price makers in this boom. Nvidia's one of them. Not one company in the hardware space should trade at higher valuations than Nvidia.
00:24 That's one of the things we hold dear. >> Most of them are. >> Particularly some of the more specious data center companies that depend on getting Nvidia chips. No one that is dependent upon Nvidia to exist should trade at a higher valuation than Nvidia itself. And yet many, many companies do.
00:47 >> All right, welcome to the Risk Reversal podcast. Coming in hot here, guys. >> Well, I love this room, and I love the fact that Jim Chanos is here. What do the kids say? >> IRL. >> IRL, in real life. All right, Jim, you're the founder and president of Chanos & Co. You've been very generous with your time over the years.
01:05 We've been doing this how many years, guys? >> I think we're into our sixth year, if I'm not mistaken. >> Jim, you are a crowd favorite. We appreciate you being here. >> Oh, thank you. Thank you for having me on this 100-degree day with >> Oh my gosh. >> air quality index and gridlock alert.
01:19 >> But it's going to break, like many things. And with that segue, is there anything breaking right now? There are a lot of strange things happening in the world, Jim. So I know that's a 30,000-foot question. We'll get more granular, but are things breaking in our midst? >> The S&P is basically at all-time highs, within stone's throw of all-time highs.
01:46 And yet under the surface, and we may have mentioned this the last time I was on, under the surface, there's a lot going on. A lot of things are acting poorly, and a lot of things have just gone parabolic. It's a really interesting time for stock dispersion, for long-short guys, and for us this has been a lot of alpha on the short side since I last saw you.
02:08 So things that aren't loved by the crowd have basically been shunned. >> And just to take a step back for a second, because oftentimes in the press when you are referred to as famed short seller, right? I just want to take one step back and describe a little bit when you say we're short this, we're short of that.
02:26 It's not like you are a short fund, you have a model portfolio. Just explain a little bit for the audience. >> We put together a model portfolio for clients. We're no longer running third-party money directly. We're just running our own money and advising institutional clients. But we do have a model portfolio.
02:42 And that is structured with 40 ideas, but it's designed to be hedged, and we tell our clients how best to hedge it, whether through S&P or more custom-hedged. And we really don't want to take a view on the market overall. We want to take a view on our 40 stocks versus the market. >> And deep fundamental work is going on there.
03:07 Let's go back to the market for one second because you said stone's throw away from all-time highs in the S&P. Guy and I seem to talk about that every day. And there are parts of the market away from, let's say, semis and storage and memory, that sort of thing, that have been acting very well. >> Yeah.
03:24 So when you think about the dispersion that's going on. Today's a great example. You have the socks down 3%. We're recording this on Wednesday afternoon. You have that DRAM ETF, I think at one point it was down 10%. So that whole infrastructure is getting absolutely killed, but the hyperscalers are all up three, three and a half percent.
03:47 They almost feel defensive at this point after they've all sold off, let's say, at least 15% or so. How do you square that a little bit? Because we're going to go deep on hyperscalers versus the infrastructure that goes into it and the like, but that is one of the weirdest days that I've seen in tech in a very long time.
04:04 And the S&P is flat on the day because of this. >> Right. And that's what I'm talking about. So my hedges are probably broadly flat on the day. And that's fine. But the rotation that's happening under the surface is tremendous. I know others have measured it, but the willingness of these sectors to go in and out of favor and large-cap stocks that are well followed to move 5 to 10% a day >> Right.
04:33 >> on no real news other than flows and sentiment is remarkable, but I also think it's highly speculative. >> Right. Both up and down. Yeah, this has been going on now for a couple months where over the course of a week, you'll see this happen two or three times. >> To the upside and the downside.
04:51 And you have a volatility index that's 17 maybe. So you have single-stock volatility, but the broader-market volatility is not there. Is that just inevitable? >> I don't know, but all the signs that we see, and they've been building since for a year and a half now, of retail participation, retail speculation.
05:16 And now finally, we've talked about this before, the final thing that hadn't shown up for this cycle, at least since '21, supply has now started. And now we're starting to see record issuance both in IPOs and secondaries and insider selling. So you have now a third leg to the stool of generally rich valuations, retail speculation, and now heavy issuance.
05:45 Generally, that's not been a great sign for the market as a whole. So far, it's been okay. This happened in 2021, first half of 2021, we saw these characteristics where you had all kinds of issuance, you had a lot of disparity in returns. The general market went up, I think in 2021 the peak was on December 31.
06:07 But shorts began to work in mid-year after GameStop, and that's what 2026 feels like right now in our numbers and in our ideas that we share with clients. There's been a lot of volatility, most of it to the downside. >> Mhm. >> And this reminds me, Guy, you saw The Wire, right? Or were you more of a Sopranos guy back then? This was the days of the >> I love The Wire.
06:35 Bunny is one of the characters. >> All right. Well, listen, there was a quote from Slim Charles, he's saying to Cutty, Cutty just got out of jail, and he basically says to Slim Charles, he says, man, the game has changed. And Slim Charles says, no, no, the game is the same, it's just got more fierce, right? So when you think about 20 That's pretty good, right? I didn't come up with it.
06:51 But when you think about 2021, and you just think it was like death by a thousand cuts, and that was SPACs, it was just a bunch of that was coming to the market. And we also saw a bunch of stuff that people were not particularly interested in 2019 prior to COVID that just went parabolic. You remember, like a Snap went from like five to 80, and we saw it in PayPal, and the list goes on and on.
07:12 When you think about issuance, right? So SpaceX does a $75 billion IPO, and then they follow on with $25 billion in debt almost immediately. This is a week after Google, and I know you've had a lot to say about that, raised what was that number? 80 billion, and then it was upsized to 85, right? And then we know that Meta, and the list goes on and on.
07:32 So help us think about the combination of IPOs, secondaries, and then also debt, because this is the sort of thing where a lot of these companies who are raising debt, not the hyperscalers, but some of the smaller, like the neo clouds and Oracle for instance, they're like near junk where they're raising, and to me that's something that's very different given how important they are to the narrative or the sentiment about this AI infrastructure building.
07:56 >> There's a lot more debt that's being used than people think. A lot of it's off balance sheet as you guys know. Which makes it interesting, and the hyperscalers themselves are issuing now record amounts of equity and debt. So it's a torrent of credit needs in this very hot sector, the AI buildout, with very uncertain returns, and we'll probably talk about that a little bit later, but that's what's so interesting in that in 2005 and 2006, it was people taking out
08:33 mortgages to buy homes that were overpriced. But when all was said and done, the homes actually held their value. They went down for a couple of years and people that were over-levered or didn't have the income to support the mortgage defaulted, but the home prices pretty much held up.
08:51 The assets held up. This is much different. We are embarking on hundreds of billions of dollars in CapEx and projects that may be completely uneconomic from day one. And I'm not talking about the models. I'm talking about the physical structures, the data centers and whatever, that you just can't get the numbers to pencil out on in any really meaningful way, and yet more and more capital is going to that asset class with lower and lower returns, and that's going to be a really interesting
09:27 thing to play out. >> All right, shelve that for a second because then we'll come back. The credit markets show no sign whatsoever of being concerned at all. Now maybe, and you know this better than I do, but credit doesn't care until it does, and when it does it happens fast. Rising interest-rate environment is not we are not set up for interest rates to go higher in this country.
09:50 Whether it's the market or the economy. So thoughts on that. >> There's too many projects that get penciled out at mid-single-digit pre-tax rates of return. Whether it's office buildings, whether it's data centers, whether it's storage centers, whatever. Everybody is doing deals at five and six and seven caps and with the 10-year at four and a half.
10:12 >> Mhm. >> And so you just apply a lot of leverage and maybe some crazy mezzanine financing and you tell your equity investors I'm going to get you 15%, and that's how it's done. Rates going to six or seven just blows that all up. Blows up asset class after asset class. And I think that's something, that's a risk that is not being appreciated beyond just to talk about data centers or what have you.
10:47 SL Green here in New York City, where we're sitting, trades at a five cap for New York City office buildings. SL Green's stock's gone nowhere in 25 years. It's been a terrible investment. So their office buildings have been a terrible investment and we have no position in SL Green, but I'm just saying to me it's a bellwether of willingness of people to put really, really low cap rates on assets, and that applies to equities as well, but you move interest rates up to six or seven percent and the world falls
11:23 apart. >> Well, it's interesting. They can put the cap rates wherever they want. The market doesn't have to participate in those numbers. They can say this is foolish, but they're going down that road. So let me ask this question. You said at 6% things fall apart, but the market will sniff that out long before 6%.
11:40 >> But it will sniff that out long before. So where is that level? Where's that line of demarcation? >> I don't know. >> You have a thought, though. >> I think that if we saw acceleration toward 5% with feelings that it could go higher, I think spreads would begin to widen out.
11:57 And that's what you need to see. You referred to it earlier. And spreads in the lowest end of junk are starting to widen out, by the way, on triple C's. But in the triple B double B area, they're not. And I think that you would need to see rates moving toward 5% with some vigor where suddenly credit investors would begin to rethink, maybe I don't need to buy this office building at a 4 and 1/2% cap rate.
12:25 >> So I've heard you say that what's going on right now is like a redux of what we've seen in the late '90s, right? >> It's much worse. >> It's much worse. Okay, I don't want to get to why that is, but can you draw a line, and I think you did just a little bit, between the internet-infrastructure build and then as we get to the financial crisis, because one of the things that got my antennas up, this was going back 6 months or something like that, where Meta announces this
12:51 Louisiana data center and Blue Owl and SPV and KKR. And I had a hard time as just a dumb equities and options guy getting my arms around that. And then I started looking at some of the terms, Meta was only going to own 20% of it and they had a 3-year out.
13:08 And all this sort of thing, and I say to myself, how durable are these long-term contracts if you're one of the biggest spenders in the area, but you don't want to actually have this debt on your balance sheet? >> No one seems to suddenly want to have the hard assets. And we can get into that a little later because there's some developments this week about that.
13:27 >> Yeah, it's incredible. So in the dot-com buildout there were only two business models that failed from the get-go, right? And those were the CLECs, which were the competitive local exchange carriers, which were in effect putting a phone company in every building in Manhattan. Everyone was going to have their own phone exchange.
13:50 And the Lucents and Nortels spent lavishly to invest in them so they would buy their equipment. The other one was the fiber, the fiber light, which ultimately worked but took so long to pay out that the fiber companies pretty much all defaulted. Those two industries as a whole, and we went back and looked last year in our files, those two industries spent a total of a hundred billion dollars over five years, from '98 to '02.
14:23 Right? So while they were raising lots of money, it was basically 20 billion a year for both industries. >> What's that in today's dollars? >> Well, that's what I'm saying is it it >> It pales in comparison. >> in comparison to what individual companies are raising in this cycle on data centers and AI buildout that may or may not be profitable.
14:50 So most of the spending that occurred in the dot-com era was by enterprises. It was by Bank of America, it was by Merrill Lynch, it was by GE and Coca-Cola to basically network their computers together. And some of it was to replace PCs and because of worries about Y2K. By and large, those companies were profitable and remained profitable through the cycle.
15:18 What they did was they cut back their orders. And that's what I keep pointing out to people. If Coca-Cola needed 10,000 routers and it ordered from Cisco in 2000, by 2001 they cut it back to 2000 routers, right? And order books just collapsed and earnings collapsed at the same time. But by and large the entities that were spending the money, the Telcos, the enterprises, stayed profitable through that period.
15:45 In this go-around, right now the hyperscalers are the profitable companies that are doing the spending. So you have that bedrock, but almost everybody else in the ecosystem is not necessarily profitable right now and raising lots of VC money to build data centers or build out an AI model or what have you.
16:07 And so the numbers today dwarf what was happening in '99 and 2000 in terms of CapEx and relative to the size of the economy. >> Moses put the commandments in stone, right? Like a chisel probably, right? >> times supposedly. >> Yeah, 10. Well, with a lot of words. How etched in stone, I talk about the sanctity of CapEx.
16:31 People seem to think it's sort of etched in stone, it's going to happen regardless of environment we find ourselves in. Do you believe that to be the case? >> No, and history says that's not the >> So why are people so willing now to say >> Because marginal returns on that CapEx are so high right now. >> And that's the reason.
16:47 >> Yeah, that's the reason. It's simple. Right now if you dig a hole in the ground and you get a client, you can claim I'm going to be able to rent out capacity for X and that's a 20% return on my capital, 25% return on my capital. The bad news is I'm only going to pay you that for a year or two and then we'll see how things go, but the asset that you're building is a 20-year asset.
17:09 So people are making decisions on long-term projects based on spot prices. And that's a terrifying thing. We saw it in the shale business. We've seen it in lots of businesses. Railroads in the 19th century, rail freight lines, rates were enormous and were tremendous incentive to build double and triple capacity, from New York to Chicago or Chicago to St. Louis.
17:36 And then, of course, when demand fell back a little, freight rates collapsed and railroads went bankrupt. And so matching duration is a really important thing. It even happened in the global financial crisis on the other side of the balance sheet where people were financing the long-term unsaleable part of their derivatives book in the repo market, right? And suddenly they realized there was a mismatch and people who'd lent them money in the repo wanted their money back and the asset was still unsaleable.
18:10 So that's one of the biggest financial crimes I'm seeing right now in terms of finance 101 is that people are committing long-term capital projects based on near-term spot pricing. And given where AI models are, how they leapfrog each other, what demand may or may not be in 2 years, that's a big bet.
18:34 >> The best example of that is actually DRAM, right? And so this is obviously Micron. And they've owned this market and so now all of a sudden they figured out that you need high-bandwidth memory to attach to your high-end GPUs, right? To train the models, all that stuff, right? So when you think about Micron's a great example where after their last earnings report, some of the big reporters and whoever, big bulls were saying these are LTAs. They
19:01 were talking about the order book. They were talking about RPOs and LTAs, and Guy used the word sanctity. We've seen this before. Micron's a company that had negative gross margins 3 years ago, right? So how much of this future expectation for revenues based on an 80% gross margin are you willing to assign to a company that has gone from, let's say, what was it like 100 bucks like 18 months ago and it topped out at 1250 just recently and now it's trading at 950. That just
19:32 seems egregious to me. And the other point that I'd love to get your take on here is that here's a company when Nvidia over the last 3 years has basically been doubling their revenues, right? And if everyone knew that Blackwell and Vera Rubin and all this stuff was going to need increasingly expensive memory to do all this, why is it that Micron was not bringing on capacity? Like, what did they see or not see? What visibility
20:01 did they have? And so at this point, why would you believe that those LTAs, that huge order book that they are literally being awarded market cap right now for, that they are going to materialize over the next few years. >> History says you should be very cautious about that view.
20:21 I would turn the argument around from the other side of the telescope and say, okay, there are a handful of companies that are gatekeepers and price makers in this boom. Nvidia's one of them. Not one company in the hardware space should trade at higher valuations than Nvidia. That's one of the things we hold dear.
20:44 >> Most of them are. >> Most of them are. And particularly some of the more specious data center companies, no one that depends on getting Nvidia chips. No one that is dependent upon Nvidia to exist should trade at a higher valuation than Nvidia itself. And yet many, many companies do. And so even within the AI space, there's just insane dispersion between the way people are buying stocks and valuing them and not stepping back and saying, all right, let me try to look at this
21:20 ecosystem. What's sustainable? What's not? What's ephemeral? What's based on spot pricing, not long-term contract pricing? And should I be putting up the 45th data center in Abilene, Texas? >> All right. So with that premise that none of those companies should be trading at a premium to Nvidia, which that's fair.
21:47 >> Yep. >> Is that the market's way of saying that Nvidia is not going to enjoy 75% gross margins in perpetuity and when those margins start to decline, that valuation, which is reasonable now, will actually get more expensive. Is that sort of what's embedded in that? >> Right now, until Nvidia gets real competition, and it's starting.
22:09 I mean, the gross margins seem pretty solid for the next couple of years. After that, it's anyone's guess. But it's anyone's guess on all of this after a couple of years because basically the physical-plant cycle is a couple of years, right? To build a new DRAM plant, to put up more capacity, to invent a better GPU. Tech cycles are short.
22:33 People forget that. And then people are incentivized to move tech cycles very quickly. So basically, you can look out a year or two and that's kind of it. And yet, as I've said before today, people are committing massive capital projects well beyond two- and three-year horizons.
22:56 And that's what's different. >> Yeah, and it seems like where the valuations are getting a little out of whack, and if using the framework that you're talking about with Nvidia, those margins if it went to 71% which is where they were last year and then they kind of accelerated, it seems like the whole notion of bottlenecks.
23:13 This is where investors see opportunities, right? So it's a supply-demand situation, but the more I read about this, and I think it was Michael Burry wrote a piece about this maybe last week, is that Jensen wants to keep moving from Hopper to Blackwell to Rubin, that sort of thing.
23:34 But they're basically hundreds of thousands of these next-gen GPUs that are not plugged in and for a couple reasons, it's not that easy to retool a data center that was built around Hopper or something like that. >> Now I'm going to put my nerdy accountant hat on and say that's because it doesn't hurt those companies for the time being because those chips that are sitting in warehouses that aren't plugged in are not being depreciated.
23:59 >> Right. >> It's construction in progress. And I've been harping on this fact for a long time that if you look at the balance sheets of the guys spending big amounts of money, this subaccount of property, plant, equipment called construction in progress keeps growing. And that's basically stuff that isn't plugged in yet or capitalized interest or capitalized labor costs for projects that have not come online, and the accountants let you basically defer writing any of it down on a depreciable schedule until the
24:34 revenue starts. So they're not being hurt by it. And it's the other problem, of course, we've talked about is the dichotomy between massive CapEx which creates revenue and profits for the Nvidias of the world, the spenders of those same dollars who are basically not expensing those dollars. They're capitalizing them or deferring them.
24:59 So it's why you get these S&P 500 earnings estimates that the S&P EPS long-term trend is 6% a year. I think we're talking about 28 this year, another 20 something next year. And that happened, by the way, at the end of the dot-com boom in '99 and 2000. And you don't get above trend, 5x above trend or 4x above trend, unless something like that is happening where $1 of profit is not a cost for another company.
25:32 So that's what's happening now. >> Yeah. So the obsolescence of the asset, which is at terminal velocity now, is not being taken it's they're using historical depreciation models. Is that accurate? >> Well, people are arguing is it 5 years or 6 years? The hyperscalers, the neo clouds are at 5, 6 years, but it's worse than that because a lot of it isn't being the clock isn't starting the minute I get the GPU from Nvidia.
26:03 It might not start for 18 months. So it means in effect that chip that gets written off over 7 and 1/2 years or 6 and 1/2 years. We use 10 years in our models to be conservative, by the way. And even then we can't pencil out most of the data-center economics. But if you're using 5 and 6 years, it's terrible.
26:25 But again, as I indicated, it's worse than that because of this construction-in-progress accounting where a lot of this stuff is sitting not being used, and it's economically depreciating and technologically depreciating, but it's not going through the P&L. >> You use 10 years, they're using, let's say, six.
26:45 What should it be? >> Well, again, from what our research has told us is that if you run these chips 365 24/7, they burn out after 10 to 12 years. You have a physical life of 10 to 12 years on all these chips. >> Not taking into consideration technology gets better and the obsolescence of the >> and now I will play ball and say yes, but spot prices, Guy, are remaining high and have not deteriorated in a straight line.
27:20 And so therefore these chips are going to last forever. >> Right. >> And of course there's a fallacy again in using spot prices to do that. >> Well, it's interesting, and I saw one neo cloud. It was an article, I can't remember where I saw it today. They were talking about coming up with some sort of derivatives to hedge out pricing.
27:38 >> So allegedly Coreweave has been looking at a way to hedge downside in asset prices, which I thought was interesting given the story is >> Well they should be, right? >> Theoretically yes, but the narrative is this stuff is going up in value not down in value. >> what I thought. >> The bigger development was by Nebius today which came out and unveiled a new business model that's asset-light.
28:02 Which I had to just giggle on my way into New York this morning reading this because here you have the most asset-heavy neo cloud. They spent $4.40 in capital to generate $1 in revenue. Coming out and saying oh by the way we have a new business model. We will provide the software stack for you and we'll market your assets to our clients who need compute but you provide the data center, the GPUs, the CapEx, the maintenance CapEx, and we'll just kind of manage it for you.
28:40 Sort of like Marriott hotels with franchise properties, right? And I'm like wait a minute. For the last 2 years these neo clouds have told us that the capital intensity is where you want to be because this is just gold, right? And you just want to own the physical assets because they're going to mint money for years and years, and now one of the biggest upstart players has told you this morning, well, never mind.
29:08 We don't need to own the assets. We'll just manage them. I think that was a tremendous admission. >> Well, it's not too different in many ways from what Meta announced last week, right? Is that all of a sudden, wait, they have excess compute that they can rent out as a hyperscaler now? And this is a company that has tripped over themselves one time after another as it relates to Meta AI and all their models.
29:32 >> We'll go one step further. My old favorites, the legacy data centers, the cloud data centers, have all now suddenly put up their portfolios for sale. One of them's trying to come public that went bankrupt in 2023. The old Sixtera is coming public as C-Squared. And meanwhile, you have executives leaving these companies.
29:53 So the legacy data-center guys are all suddenly assets for sale. I have a theory about why all this is happening right now. And my theory is that the data-center guys, both the legacy and the new ones, are suddenly realizing as costs have escalated for new developments because of labor shortages, because of equipment shortages, all that stuff, they're realizing their maintenance CapEx numbers are going to be higher in the future than they've let their investors know. And thus returns
30:29 will be even lower than they already are. So they're trying quickly now to unload >> Well, let's talk about SpaceX for a second and really not Well, we're not talking about SpaceX. >> Data centers in space? >> No, I don't before we get to that because I want to go deeper into the data-center thing.
30:48 But so on the eve of the IPO, we all know that xAI was a money pit, right? And talk about overbuilding for their own needs and then they become essentially renting out the compute. >> were bought for SpaceX stock valued at 250 billion in February. >> Yeah. And so now they do these deals, one with Anthropic and one with Google and >> two spot deals with tremendous outs.
31:13 >> Yeah. So within 3 months out, right? So think about this. And they're each paying about a billion dollars. It's not going to come online until let's say the fall and at any point either one of those companies could get out with a gun. >> And Elon even tweeted out when that happened.
31:31 These are very short-term deals. >> Yeah. So it looked like it was a little stealing from Peter to pay Paul to get this thing on the tape, that sort of thing. >> I think they were highly promotional deals. >> Yeah. And to the point is Anthropic and Google also have a lot of incentive for all this excess compute to get stopped up to some degree.
31:52 >> Yeah. >> So this is going back to I think it was like last fall, Satya Nadella, they've been tripping all over themselves with their strategy, and Azure is not the cloud provider that a lot of folks, if you put it up against AWS and even the way GCP is rallying, but he said they were no longer compute constrained.
32:14 They're energy constrained. And this is going back I think it was like last October or November. All of a sudden though, are we starting to see the overbuild despite what we continue to hear about CapEx going higher, coming into the year Amazon raised their CapEx from 120 to like 200 billion or something like that.
32:32 >> Microsoft announced a deal recently where Chevron is going to provide the power directly. There's one thing we are not short of in this country, it's power. Now, there's transmission issues, there's regulatory issues, but we are not short of cheap power. And I keep pointing out to people who talk about data centers and space saving money.
32:53 Power costs for a data center are about 5 to 6% of revenues. They are the smallest cost component is power. And so power 2 years from now to 3 years from now, power is not going to be an issue. We're going to figure out these bottlenecks. We're going to figure out a way to get the various different data centers connected to the grid.
33:15 In Texas, they're already doing it in a myriad of ways, which is the biggest market for new construction. Power is not going to be a bottleneck. And a lot of data-center stories are based on the fact, well, we have permission to connect to the Texas grid. Great.
33:35 Texas grid is going to charge you wholesale prices. And read the fine print. And so it really is silly. I think there might be more political backlash to data centers, which we're starting to see, and that could be a bigger problem, but I don't think that power, access to actual electricity, is going to be the issue that some of the bulls think it is for scarcity reasons.
34:01 When does the market start to demand >> the return on invested capital from some of these companies because they haven't yet. >> Yeah, they haven't yet. The companies trade on announcements, right? They trade on so-and-so signs a 20-year deal with so-and-so to provide data-center services and power, and stocks go up or down.
34:26 More recently, they've been going down on these announcements more than up. So the market's already beginning to discern that not all these deals are great deals. And I think that's an important change. The other thing we're watching really closely for our clients is the return on incremental invested capital for the hyperscalers.
34:52 I think I've sent you the graph after the first quarter. And it's going down. It's going down rather rapidly. Every dollar that the hyperscalers is spending on the margin is on the margin creating less operating income. It's still healthy, but basically it's gone from as a group 40% a year and a half ago to about 20% today.
35:17 And if the spend keeps up at this kind of rate, it's going to be moving toward 10%. And then you're going to have real issues at the hyperscaler C-suite to say, okay, are we going to keep spending this or are we going to basically rent the capacity? >> So what do you think Andrew Jassy >> And by the way, if Google and Meta and Amazon and Microsoft and Oracle as a group are incrementally only earning 10% on their invested capital pre-tax, then the neo clouds are in a whole world of hurt.
35:54 >> What do you think Jassy, Sundar, and Satya, so they run the three biggest clouds, right? They're looking at Oracle and they see, going back to last September, OpenAI and AMD, these are hundreds of billions of dollars, right? And RPOs, backlog, the whole thing, LTA, all that sort of stuff.
36:16 And the stock immediately filled in that 30% gap that came the day after, right? And then it continued What is it down? 65, 70% or something from those levels? After a rally that it had, let's say a few months ago, it's been cut in half again. It's down 50%.
36:34 What do you think those CEOs are looking at Oracle? And is there anything that they can extrapolate a little bit because to me it makes >> It just Oracle has the worst of those ROIC metrics I just talked about. And so those other guys, if they're doing their jobs and I think they're doing their jobs, they're looking at Oracle saying, we don't want to be down at those levels.
37:02 We need returns higher than the lowest guy. And I think that's why we're watching that metric so closely for clients because the capex numbers keep going up. The estimates keep going up. Operating income estimates do not keep pace with that. So we're going to get a point in late '26, '27 where people are going to start putting pencil to paper besides us and say, wait a minute.
37:33 Does this next trillion dollars >> Make sense. >> make sense if you're only going to earn 50 billion on it? We could earn that on treasuries. >> Mhm. >> So we're going to get there some point in the next 12 months, I think. >> Market now, to your earlier point, the market is sniffing some things out and they're seemingly sniffing out the companies that have been on the sidelines and not spending and I'm pointing towards Apple which is where we're sitting here at a heroic run higher.
38:01 I think in large part due to the fact that they've been sort of watching everybody else. Thoughts on that? Not specifically the stock, but just in terms of maybe they made the right decision to sort of wait it out. >> They didn't take the risk that the other companies were taking that might prove to be unfounded.
38:19 And again, I don't know what the AI models will produce and I call that the magic, right? Amazing things might happen out of that. I'm really concerned about the capital-intensive side of the business. And that's where the money is going. That's where the credit's going. That's where the off-balance-sheet financing is going and that's where the blow-ups are going to happen if they happen.
38:44 >> On the Apple front, I thought that lawsuit, Apple suing OpenAI, that's kind of the run-of-the-mill stealing trade secrets. We see it a lot, but do we see so many researchers flipping around between Anthropic and OpenAI and Microsoft? It's been a huge brain drain, right? From some of the traditional like Microsoft.
39:02 >> Well, and the frictional cost to move around, not only employees, but users switching models is I think one of the defining things of this boom. >> Yeah, and I look at the Apple thing. I look at Sam Altman and Elon. There's so many they used to be frenemies, now they're becoming enemies, right? And I think that as we get through 2026, maybe into 2027, they're all going to be suing each other, right?
39:32 And let's be honest, there's an OpenAI Microsoft thing coming, right? We know that You get my point, right? At some point, I think the narratives get really sloppy and a lot of it has to do with a lot of this circular financing. Bloomberg has had this great chart that they keep updating and it's like let's just discuss that a little bit because that's also something that is reminiscent and you alluded to that of the late '90s because without Nvidia investing in all
40:02 of these companies that I don't even know money's changing hands. Is money just changing hands or is it just >> No, it comes back pretty quickly. >> Yeah, like just how could this go wrong because bulls have been so dismissive about this? >> Yeah, no one's too concerned about it because as a percent of Nvidia's balance sheet it's not huge.
40:24 It wasn't huge, by the way, as a percent of Lucent and Nortel's, either, if you go back and look. But it enables those equity investments enable the equity investees to then go raise more equity or debt. And there's a leverage effect on it, right? >> And they're using them as collateral when they're going to do that, right? Yeah.
40:44 >> So it does have an impact. The numbers are big nominally. They're still relatively small to the size of the AI buildout, which again, I can't stress enough is just enormous. But I suspect that it also reflects a little bit of the incestuous nature of the personalities who were there at creation, if you will, all helping each other out and investing in each other that are now basically competitors.
41:16 And it's going to be interesting to see how that plays out in some of those relationships. >> The epicenter isn't the banks. Historically, they have been, but they're tangential to this whole thing. And the banks have been as a group done very well. Thoughts on valuation, forget about price earnings.
41:33 Some of these historic metrics are getting a little expensive. JP Morgan, for example, north of three times now price to tangible book. >> Yeah. We have no involvement in the banks. We are short one of the big private-equity entities that I think is doing lots of stupid things.
41:51 >> done well, no doubt, because none of them trade particularly well. >> That's been our main financial short along with some of the >> Is that related to the AI spend? Blackstone and this is Apollo flew out. No, yeah. But Apollo and Blackstone did that deal, I want to say a month or so ago, really June, where they basically raised $35 billion to invest in TPUs or it was like the Broadcom >> what we're involved with is not only in AI, but it's an office building and it's in
42:20 that low cap >> Right. >> world. If cap rates go up or rates go up, this company is in big trouble. >> Yeah. And just as far as, I think Guy drawing that line, the banks now are doing a lot of lending, right? And so it seems like it's becoming more and more, but this buildout is becoming more and more financialized and it's finding its way into other parts of the economy and there's an article in the FT this morning that was actually saying that banks are now AI stocks because all of
42:54 the trading that's going on in all these names, that sort of thing. And we remember these sorts of narratives, and I just want to be really clear for the listener for the viewer right here. People are going to say you're leading the witness. We all know what Jim is. He's been very transparent.
43:09 He talks about it in podcasts and events and on Twitter, that sort of thing. Guy and I are just trying to figure it out. We watch the news as much as everyone. We're not that technical sort of people, but I don't think you actually have to be because some of the most technical people out there who are building the models or believe in spending hundreds of billions of dollars in the data center, there's some things that they are obviously missing. And so I'm just curious how you think about it. We can put
43:33 a kind of bow on this conversation, but how does this shake out? For instance, could the bulls actually be right about this because we talked about space centers and data centers in space, that sort of thing. Some of the smartest people I think I know, from investors and that sort of thing, they believe data centers in space is going to be a massive unlock, that sort of thing.
43:55 And I just don't know how we could have, and Sam Altman, by the way, is on the other side of that, right? And you think that he has a lot of incentives to believe that way. So it seems like there's a ton of daylight between the bulls and the bears on this. >> [laughter] >> Well, there always is, but there seems to be more right now than I've seen in a long career.
44:16 Look, this is going to be like the internet. There's going to be huge winners that come out of this and wealth created on the backside of AI. We're in the period now, however, where everything is being valued as if it has worked or it will work, right? And that was the problem with '99-2000 was that everything was valued as if it was a new-economy stock and demand was infinite and blah blah blah blah blah.
44:47 And the realization for a couple of years that wasn't the case and that the laws of economics and finance held and order books collapsed and profits went down was a real splash of cold water on a hot market. And right now we're in that same phase where everything, you can build castles in the air in terms of the narrative or data spenders in space to use the analogy, and it can't be disproven, right? And when you have stocks with valuations like Tesla and SpaceX that are based on promises, right? Who's to say that
45:27 my company can't trade like that, too? And so that's just a matter of investor psychology. It's the glass half full, I say it till I'm blue in the face, in bull markets you put a premium on promises and bear markets you put a discount on reality.
45:47 And we're clearly in the former right now. Do we get to the latter? I don't know. >> Well, with respect to Don McLean who you just invoked without even realizing it with your castles in the air, thank you once again, Jim Chanos. >> Anytime, guys. I'm always happy to join you. >> This podcast is for informational purposes only.
46:10 All opinions expressed by me, Dan Nathan, Guy Adami, and any other participants are solely our opinions and should not be relied upon for specific investment decisions. </content> </invoke>