Title: We're In The Greatest Stock & Earnings Bubble In US History | Fred Hickey Show: Thoughtful Money with Adam Taggart Guest: Fred Hickey (editor, The High-Tech Strategist newsletter; 47 years an active tech investor) Date: 2026-07-14 URL: https://youtu.be/mhPwRH3uq1Y Length: ~1:16:30 Note: Fillers (um/uh/you know) and stutters removed; wording otherwise verbatim. Auto-transcript name garbles corrected (Zhipu, DeepSeek, Qwen, SK Hynix, Chamath Palihapitiya, Kevin Warsh, Jensen Huang, Vinod Khosla, BPGDM, Uzbekistan, McKinsey, zero-DTE, Grantham-plus). The startup that shrank Qwen 3.6 to an iPhone is attributed by Fred to Vinod Khosla but the company is not named in the transcript.
00:00 the economic modeling if there was any [laughter] to begin with is disintegrating as we go forward here. It's seeming, and I think it means that we have massive oversupply, overcapacity, massive malinvestment here of data centers that are being built. The problem is that when you get great bubbles like this, as we saw in 2000, the earnings will collapse as they did in 2000, and then expectations will collapse and then the stocks will fall apart.
00:39 [music] Welcome to Thoughtful Money. I'm its founder and your host, Adam Taggart. AI-linked stocks account for a record 45% of the S&P 500's total market capitalization, and they drive nearly all of the gains it's had for the year so far. And note that I said that's the S&P, the general market.
01:04 These companies make up an even higher percentage, nearly 70% of the NASDAQ 100's market cap. Will this new AI-powered tech renaissance continue powering the indices higher for years to come? Or is this extreme dependence of the markets on a single sector a major vulnerability, putting all of our hopes into a handful of companies that may not be able to keep growing at the meteoric rates that Wall Street is expecting? For answers, we're fortunate today to speak with Fred Hickey, editor of the highly respected newsletter, The High-Tech Strategist, which Fred has been
01:37 publishing since 1987. Fred, thanks so much for joining us today. >> A pleasure to be here. >> Pleasure to have you back here, Fred. You're one of those guys I get asked pretty much every week by the audience, hey, is Fred Hickey ever coming back? In fact, somebody asked me about 10 minutes before we hopped on this recording here, and I was really happy to tell them that I was about to talk to you.
02:00 So, Fred, I love talking to you because you're such a smart guy in general, but you've been covering the tech industry for decades, going on what, 40 years at this time. >> Actually I started, let's see, 47 years now I've been following it, but I've been writing about it for 39. So >> Oh my goodness.
02:24 >> 47 years I've been an active investor in tech. >> Okay. So you've had a front-row seat to the show for almost half a century. And I know from watching your work and having talked to you in the past, as transformative as AI may be for the future, you've got a pretty healthy skepticism about at least current expectations for the sector.
02:52 And while I've been waiting to get you back on the program, the expectations have just been ratcheting higher and higher and higher. And as I mentioned, this sector is almost like you've got an Arctic sled with 10 dogs on it and one dog weighs about 1,000 lbs and all the other dogs are Chihuahuas, right? Which might be great in terms of your rate of speed this dog sled can take you, but if anything happens to that big dog, you're in a lot of trouble out there. So I want to talk a lot about
03:24 your latest thoughts about AI and where things are going there. If we can, maybe just before then, just start with a more general question. When you look at the general markets right now, what is your current assessment of what's driving them and any other key defining themes? >> Well, I'm in complete agreement with Jeremy Grantham that we're in the greatest stock bubble in US history.
03:52 And he thinks you're going to have a great crash. The stocks are going to decline 70% or more. I'm not sure that will happen, but I agree with him that this is the greatest bubble, but it's more than just the greatest stock bubble. There's also an earnings bubble. And that's something I've never seen before.
04:14 And when I say an earnings bubble, right now the price-to-sales ratios, the major market ratios, show that this market is far greater in valuation than it was even in 2000 at the top. So the market cap to GDP ratio, the Buffett indicator, is 241% of GDP versus 160% back in 2000.
04:45 The price-to-sales ratio is 3.7 times versus 2.3 times in 2000. The price-to-earnings ratio is the only one, on just a regular PE ratio, the S&P is about 25 times. But the problem is the E, and that's where the bubble is. All of the earnings are coming from the hyperscaler data center spending that's going on.
05:14 And there's a huge gap between what companies are reporting right now for revenues and earnings and what they really are expending. So what happens is that when the data centers are being built, and even before, when they're ordering parts, you have the Nvidias, the Microns, the Western Digitals, all the component makers get their orders and they recognize the revenues right away, pretty much right away, and the earnings. And so we've seen these massive earnings increases. We've seen Micron, Micron which is a cyclical company as cyclical
05:51 as they come. It had negative gross margins back in 2023. Negative gross margins, never mind losing money. They had negative gross margins. So it's a hugely cyclical business, but now they have 85% gross margins. And because of the incredible amount of spending that's going on by the hyperscalers, it's $750 billion by the five biggest US guys and then there's another hundred billion or so from outside.
06:20 And so what's happening is all these orders are going in and Nvidia's numbers are soaring and Micron's numbers are soaring and Western Digital's numbers are soaring, and so as their earnings soar you're seeing Micron stock go up hundreds of percentage, up 700% this year. But the hyperscaler costs don't show up right away.
06:42 First the data center has to be built and then it has to be turned on. Once it's turned on, then the expenses, the costs, then start to tally up, but only a small amount each year. So it's the depreciation. The GPU parts, for example, from Nvidia, and the DRAM parts, those aren't the servers, those are depreciated.
07:04 Those expenses are depreciated over five to six years. So you only see a small part of the expenditures. And so that has created this gigantic gap of inflated earnings from the companies that are providing the components to the data centers, and the expenses that are going to show up, and they're starting to show up now at the hyperscaler level.
07:32 So the other part is that the hyperscalers themselves have been showing good earnings because they're not showing the depreciation expenses right away. They'll accelerate over time. But also they have invested, most of them, like Microsoft and Amazon and Alphabet, have all invested in the frontier model companies, Anthropic and OpenAI, as well as SpaceX, and they're generating huge mark-to-market gains, one-time gains in stock prices. That accounted for 40% of the growth in earnings in the first quarter.
08:10 And the other thing I did was I took the first quarter numbers for the market overall and I asked AI. It was Gemini that I did it. I asked them to strip out these one-time gains and also the gains that are coming from all of these suppliers like Nvidia and Western Digital and Seagate in the S&P 500.
08:36 And the S&P 500 earnings growth rate was single-digit growth. So when we talk about an earnings bubble, we have these massive earnings showing, I think it was 28% in Q1, but you strip out the 40% from the one-time markup gains and it's 16%. And then if you strip out the gains from all these companies, which is just the hyperscaler spending, then you're down to single digits.
09:02 So I saw a calculation that the CAPE number, the Shiller adjusted PE ratio, which is 40 right now and it's only been 40 before in 2000, >> geez >> is actually 67, which >> which would be in line with the price-to-sales and market cap ratio being about 50% higher than what the records were in 2000.
09:33 So we have this huge spending going on. It started in 2022. Google and Microsoft were competing with each other and they started this spending binge, and all of these other billionaires had to get involved or be left behind. And so now we've gone up, I think this year's spending growth rate is 75% over last year's, and we're talking about $750 billion of money. And the problem is that all of this spending has been done on the
10:14 anticipation of success of this generative AI, these LLMs, >> right? Future productivity gains >> by corporations. >> Yep. >> and returns on investment. And what we're not seeing right now are the returns on investment. So there's a chart that we'll show here that shows profit margins outside of the tech sector and also the profit margins of the MAG7, and it shows huge gains in the MAG7 which are mostly the hyperscalers, and then virtually no growth at all in
10:59 the non-hyperscaler companies, and that's a problem. I just saw yesterday that Chamath, I have a terrible time pronouncing his name, everybody knows >> Chamath Palihapitiya. Yeah, >> Palihapitiya. I have a terrible time pronouncing it, but he's ex- >> Facebook corporate guy at Facebook, billionaire, now a VC guy.
11:30 He asked in an All-In podcast, he asked his chief technology officer, how is my token spending going? And the tokens are basically what are being generated here. And his CTO said, well, the token costs are doubling every 45 days. And he asked him, well, what's my return on this? And he said, well, 5% max. Mhm.
11:57 >> And Chamath said, well, so my costs are doubling every 45 days, yet my upside is essentially flat. We're going to have to take a look at this and do something about it. And he's not alone. So even the hyperscalers themselves are starting to question the amount of token spending. They call it, we had this surge of what they call token maxing that began when we started with agentic AI really at the beginning of this year with cloud code and those things, and
12:33 there was an explosion of tokens, token use and costs, and companies like Amazon and Meta, they were encouraging their employees to do as much as possible because they thought it would lead to great things. And now they're finding out, not so much. >> Amazon reportedly got stuck with a $500 million one-month bill >> that they had to deal with.
12:58 >> And sorry to interrupt, Fred, but it's all about incentives, right? I've heard stories of engineers just basically writing looped code just to generate token requests, right? >> right >> just to rise up the boards because they had leaderboards to who was using the most, who was using the most tokens.
13:17 So that's where we were, and that was generating a lot of revenue for the OpenAIs and Anthropics. But even then OpenAI was still losing money in the first quarter even with that accelerated spending that was going on. But now we're seeing the reaction. So Microsoft stopped most cloud, using cloud code licenses.
13:41 Microsoft, so that's one of the hyper- >> You had Tesla, Elon, putting $200 weekly limits on the amount of token spending. Coinbase is cutting their AI spending in half now because they were suffering. And what they're doing in all of these cases is they're going to using lower-cost models.
14:08 So instead of using the frontier models, the OpenAI and Anthropics which are more expensive, although even then they're being forced to slash their pricing, they're using these Chinese models. And there's a whole bunch of them from DeepSeek, which if you remember a little over a year ago, or March of 2025, >> there was a great concern about DeepSeek and what the impact would be.
14:30 Well, now we're seeing the impact of DeepSeek. So there's all these Chinese models and their usage is soaring. So there's a chart that I have here that shows the amount of token usage by the top 20 models broken out by the Chinese and the US models. And you can see on that chart how, just from one month, the amount of tokens doubled.
15:05 from the Chinese, and Open Router, which is a company that measures this stuff, said that now they see 46% of market share from the Chinese models. >> Wow. >> Right. 46%. Now the average over the prior 12 months was only 11%. So the usage has exploded here. And there's another chart that shows the rankings from artificial analysis.
15:34 It shows the rankings of these models, how useful they are. And you can see the lighter green are the US-based models and then the darker ones are the Chinese models. But Zhipu now, their latest, and these are all the latest models that have been issued, Zhipu's model now exceeds the capabilities of Gemini 3.5 and 3.1.
16:03 You can see DeepSeek a little bit lower, but as I had just read very recently, with DeepSeek you can get 90% of your tasks done at 1.5% of the cost, which is insane. So all of these companies, we're talking about these are the tech companies that are finding out that they really have to cut back on their spending and to use these other alternative open-source models.
16:40 So Microsoft is also considering using DeepSeek. That was a report that just came out a week or so ago. >> Yeah, they asked for permission I think of the US government >> as well as their Yes, as well as their own lower-cost models. So the economic modeling, if there was any [laughter] to begin with, is disintegrating as we go forward here, it's seeming. And I really never thought that they had any economic models. It really was FOMO.
17:09 We have these very rich billionaires who, if they weren't fighting each other, challenging each other in cage matches which Elon and Zuckerberg did, or fighting in court which Altman and Musk were doing, they're always battling because they have gigantic egos, and that was what was driving us.
17:32 Larry Page of Google said that I'm willing to go bankrupt rather than lose this race. Well, that's the mentality. It's more of a FOMO and I can't lose this. Not, we've made a lot of calculations here and we know this is the amount of data center capacity we need. Nobody thought about that. It was all, I've got to be, I've got to have the fastest models.
17:56 I've got to be in the lead. And now this is what it brought us to. And I think it means that we have massive oversupply, overcapacity, massive malinvestment here of data centers that are being built, and much more to come because they've only turned on 5% of what they expect to build.
18:19 And yet, economically it's not working. Oracle are spending 100% of their revenues right now on their data centers. Amazon and Meta are spending 100% of their cash flows on these data centers. So if pricing is collapsing here and you're not going to get the revenues, and I worry very much about what OpenAI, I mean Anthropic, will end up doing here because they require lots of cash, that's why OpenAI and Anthropic want to have an IPO,
19:07 they require lots of cash, yet the hyperscalers' backlogs, they represent 50% of the hyperscalers' backlogs, if they end up with a lot of trouble, if they go bankrupt. And I believe some of these companies will end up going bankrupt. I mean, Oracle has, their debt is two and a half times their sales from last year.
19:30 It's just enormous amounts of debt that has been poured on into this market. If some of them have financial difficulties, then it's going to be reflected back into the hyperscalers as well. That's what the backlogs are. And you can see in another chart that I have here, the expectations by the market are that the cash flows have collapsed, you can see where they were from the hyperscalers in 2024 and 2025.
20:01 And then only now the only ones that are positive cash flows are Google and Microsoft. The other ones all cash-flow negative. How they expect by 2028, 29 and 2030 those cash flows to soar, their free cash flows, but also their earnings. And the problem is, with the depreciation expenses going up over time, if the revenues don't flow, then, for example, Google's depreciation expenses are going to rise from 17% up to 35% by 2028.
20:42 So they're going to see the depreciation expenses double. So this is the problem, that you have a gigantic market cap that's being fed by these MAG7 hyperscalers, and yet it doesn't look like right now, the way things are trending, that they're going to get the cash flows that they thought they were going to get.
21:08 They're not going to get the earnings that they thought they were going to get, and they may get eaten alive, the P&Ls will get eaten alive by higher depreciation expenses, and some of that's been reflected in some of the stocks. So I think that's one of the reasons why the Magnificent 7, there's weakness there.
21:33 Microsoft stock, as of the end of this weekend, was down 20% on the year. Oracle's was down 28%. And off their highs, Microsoft's up a third and Oracle is down 60%. So there is some concern that's seeping into the market right now. >> So Fred, the Mag7 has sort of been referred to recently as the Lag7 because they have cooled off, as you've said, while semiconductors have taken over the torch and gone white hot.
22:08 Right. >> When you look at how much those stocks have cooled off, do you feel like, oh okay, they're pricing in what you're talking about now, or is this just early innings? Like, do you expect them, if the business models deteriorate the way that you fear, do they still have a lot further to fall? >> Wall Street is saying that they're the cheapest they've seen them in a decade or more.
22:34 >> And I say, well no [laughter]. The PEs have come down. So if you look at a Microsoft PE, it's 23. Meta and Google Alphabet, those PEs are 27 times, 24 times levels, that area, and so they look like they're more reasonably priced, except that, as I said, 50% of their backlog, and they're getting a lot of revenues now from these circular deals they all have, is from these OpenAIs and Anthropics that are at great risk right now.
23:25 That backlog is at great risk, and then you're going to have the depreciation expenses hit them as I said, with in Google's case doubling over the next couple of years. So what's the PE ratio if the data center buildout doesn't work out as planned? What's their real earnings going to be? Well, they're going to be a lot lower.
23:49 So the PE adjusted for non-success would be significantly higher. >> Would be a lot higher, >> right? And the problem is, I think they would be the ones that you might want to buy when we get to a big selloff, which I expect we'll have. I don't know if it'll be Grantham-plus. I don't know that. I don't know if the people at the Fed would allow that. I don't know.
24:14 But at the end of all of this, I do think that Microsoft, see, Microsoft's taking steps. They're taking the right steps. They're offering their customer base the low-cost options now. They're going to survive here. If they have to, they'll offer DeepSeek. It's the way to survive.
24:35 It's the way to continue to control your platform. And I think Microsoft obviously will be a survivor. The problem is that when you get great bubbles like this, as we saw in 2000, the earnings will collapse as they did in 2000, and then expectations will collapse and then the stocks will fall apart and nothing was safe.
24:58 So we saw the NASDAQ fall 83% in 2000, and Microsoft at the time, Microsoft was the best performer but it fell 60%. So I think they'll all get smashed in a market downturn because they represent half of the market, and you also have a bunch of investors, for example, all these ETFs have these huge percentages of these stocks, and if they're selling, they're going to have to sell them, >> right? And some >> saying, in other words, these things go down, they're taking everything with them.
25:36 >> They're taking everything down. And so I don't expect anything different than what I saw in 2000, which is if the market goes down hard, all of them are going to go down hard. Then it will open up an opportunity to buy, right? And I did that in October 2002 and I did it in October of 2008 near the bottom.
26:02 I was five months too early on that one, but I was on time on the October 2002. The survivors of this will probably be some of the same names. Microsoft has their installed base, the enterprise world. They're trying to take the steps they need to ensure they'll still be there.
26:26 Google and Meta have massive cash flows from their non-hyperscaler business that they've been building here. And so the Facebooks and the search engines and all of that at Google and YouTubes, they're still going to be generating a lot of cash. >> We saw with Meta, Meta stock was getting whacked when they burned $85 billion on the metaverse.
27:01 They stopped spending, then the stock rallied quite a bit. >> And I think that will happen again. And really that's going to be the trigger point for this. We're seeing the economics fall apart here. But no one has said I'm going to cut back my spending yet. I expect that to happen at some point.
27:24 You just can't keep going spending at this rate and not getting the returns on investment, and seeing the pricing undermined, and not say we're spending at too high a rate, we have to cut back. That could be the trigger to send these things falling. >> Okay. So I got a couple of questions for you.
27:47 Maybe we can do a little rapid fire in just a second, but just to make sure I've taken good notes. You feel like these things, my words not yours, are sort of perched on a precipice that when the market wakes up to the fact that really the business models here, the economic models as you said, are deteriorating, there's going to be a drastic repricing downwards.
28:09 And that will really hurt. There's a long way to fall for all these companies including the hyperscalers, after which you think there will be a really interesting buying opportunity. So I got that right. >> Correct. >> Okay. Yes. >> Do you have any, I totally understand if you don't, but do you have any sort of timing estimate on that? Like is this something you think is going to happen in 2026, or it could happen over the next two years, you don't know.
28:39 >> Well, I knew in 2000 that we had a gigantic bubble. I knew the fiber optic capacity was way over. I didn't know it was 10 times more than they needed, exactly. But I knew that all of those kinds of things were happening. I knew that the markets were at dicey points and that it would fall apart.
29:03 I didn't know what happened would start on March 10th. No one knew. And in fact, leading up to that, the few days leading up to March 10th, 2000, the NASDAQ was soaring, I think about 9% in a matter of days. There was no sense that that was the top. None. And what you look for is exhaustion.
29:24 Now, that's not something you can pinpoint. You look for signs. You can see signs of exhaustion. I think we're starting to see some of that, with some of the hyperscalers pulling back. Also, the amount of money that's being, we talk about exhaustion, it's like more money being, for a long time we had a lot of buybacks from companies and they were pushing stock prices up.
29:59 We no longer have those buybacks. We have massive equity issues now. So we had SpaceX, right, with, it was the largest IPO ever. Two point two trillion dollar, was an $85 billion IPO. That's huge. We had SK Hynix just come last week, $27 billion of equity. We've had almost $200 billion of debt from the hyperscalers.
30:22 >> So, sorry to interrupt, but do you see this as late stage? Is this the people saying let's get out while the getting's still good? >> Well, for sure in SK Hynix's case and certainly in SpaceX's case, the timing was good for them. They don't generate any money, any earnings.
30:39 And it's all on future, we're going to >> it's all story. Yeah. >> But Amazon, when Amazon came out with their more recent $25 billion bond, it added eight basis points to the 10-year yield. It affected the Treasury market. That's a sign that you might be nearing exhaustion.
31:05 And Amazon's had to borrow now in euros, in yen, Swiss francs, British pounds as well as US dollars. Why? Because they basically saturated the market in the US. And if you look at the bond market too, overall the bond market, even though oil prices have come down from the highs of what, $120 or something per barrel down to the 70s level, I know it's up today, but it's in the 70s level.
31:32 Despite that, we've seen rates stay high. We have a 30-year at 5%. We have a 10-year at 4.6%. We have a 2-year at 4.2%. So rates have been creeping higher even as oil prices came down. And part of it is I just think there's just too much supply of everything here. And that's how you get to exhaustion.
31:58 So all of the spending going on, all of the IPOs, the fact that the hyperscalers' capex is so high and their cash flow has been evaporated essentially, they no longer can do the buybacks, and they were the leading buyback companies that had been reducing the number of shares that were in the market.
32:19 So there's no way I can pinpoint a date. I know it's coming. There are all sorts of other signs that you would see. There's a chart here I have on tops. The chart of margin debt is another example, where you have margin debt now up to over $1.4 trillion. It's up 55% year-over-year.
32:48 If you look at the chart, you'll see, prior to the bear markets in 2000 and in 2008 and also in 2021, you saw spikes in margin debt. But this one is like none other. It's straight up and it's enormous. And as a percentage of GDP, it's 4 and a half percent of GDP, so it's not just the dollar number that's inflation-affected.
33:15 It's four and a half percent versus 3% back in 2000. So we have the margin debt. We have the huge amount of speculation that's going on, whether it's in SpaceX, or even overseas we're seeing it in Korea and in Tokyo and in Taiwan. Taiwan margin debt's up 160%. So all of this is tied into the spending, the money pouring into these US markets to participate in the AI bubble. I'll call it the AI bubble, but they're all chasing the great growth and the great successes, and that's what you see at tops. You see that kind of thing. So all
33:54 the bells are there, all the speculative activity, the zero, we have new ones too, we have the zero-DTE stuff. We have the levered ETFs that we didn't have anywhere near in past bubbles. Most of that is new. It's new forms of speculation.
34:16 So I don't know exactly when. I just know it's coming, and so I just wait, and I've been waiting, and I'm perfectly happy to continue waiting. I have the benefit of not having to perform on a monthly basis. And I've had alternative investments that have done very well over the last few years so that I haven't missed anything by not being involved here.
34:41 It was much harder for me in 1999 to have not been participating at the very end of that. It all worked out then, but it was very difficult. This hasn't been difficult for me. So I'll just sit here and I'll wait. I think it's coming sooner than later, but I can't put a date on it. >> Okay.
35:00 And I want to talk to you about some of your alternative investments before we wind up here, and particularly the precious metals. One of the takeaways I get from you, Fred, and you tell me if I'm taking the wrong message here, but is, through your long career you have appreciated that it is far better to sit on the sidelines and miss out on the last gains of a bubble but be perched there with dry powder to be able to buy in when you get generational bargain opportunities.
35:35 And so the whole FOMO, you just let that wash over you and you let other, maybe perhaps greater fools, deal with that. >> I've been a value investor all my life [laughter] and also a contrarian, and usually those go hand in hand and it's been very successful.
35:57 Even in the precious metals world here, we had what I thought was very hot, very speculative behavior, a crazy mini speculative bubble, or short-lived. And so I had backed off that as well. And so I'm sitting in the most cash I can remember. >> Wow. >> Much of it in treasury bills, and we'll get into it I guess a little bit later.
36:23 But in the mining I'm starting to see that maybe we're seeing a bottoming process there already. But I took my positions way down. I sold into that. I started selling late last year, the precious metals, after the miners were up. Gold went up three times over a two and a quarter year period, from 1,800 to 5,600. >> Silver went from 30 to 120-something.
36:48 Yeah. >> And the miners went up by four and a half times over that period of time. It was terrific. But I grew concerned and I pulled back. I sold throughout the fourth quarter of last year and all into the first quarter to take my positions down, as I told the subscribers, as far as I could and still sleep at night. And that's what I always do.
37:09 It doesn't matter what market, whether it's tech or metals or whatever. If I see too much speculation, I'm going to pull back and wait for the opportunity. It always comes. >> Okay. Well, congratulations on walking your talk and seemingly really timing things very well.
37:27 And we'll talk near the end of this video, folks, on how you can follow Fred's analysis going forward. All right, Fred, just a couple of clarifying questions about the AI space because I find this so fascinating. You talked about how companies, particularly really large companies like Microsoft, are starting to say, hey look, these tokens are just getting way too expensive with the frontier models like Anthropic and OpenAI.
37:56 And so folks, that's like Claude and ChatGPT in normal-person parlance. And so they're looking for alternatives that are a lot cheaper to use and still get about 90% of the tasks done. And you said they're turning towards Chinese stuff. So a couple quick questions about that. One, national security.
38:14 Is that a real factor? Is that a real worry, a real factor here? I mean, if you're running DeepSeek, do you need to have a little bit of worry that this thing is basically looking into my company? >> Yes, and so certain companies wouldn't be able to use all of >> And does the US government place blocks on our largest corporations using a Chinese >> I don't know if they can do that. I mean, it is possible, I thought about this, it is possible that they could ban the
38:50 multiple, there's many of these Chinese LLM makers, they could outright ban them, ban their use. It is possible, if that happened. However, Microsoft is starting to use the same techniques to build their own low-price models. There's a company called Reflection that's US-based that is very similar, using the same techniques that the Chinese modelers are using.
39:24 So I don't think they could stop the use of Chinese models, but it wouldn't stop the pricing pressure. >> Okay. And if I can interject, because this pricing pressure was exactly where I was going next. So the current valuations for all the companies in this space is, we're going to build all this compute and then the compute's going to be sold for this average price. And you're basically saying, well that average price might be lower than what
39:57 the analysts are currently forecasting because companies are going to switch, right? Just like when steak becomes too expensive, you switch to chicken. Anthropic becomes too expensive, you switch to the cheaper Chinese models. And what's the differentiation between the two, the reason why the Chinese models are cheaper, is we have not allowed China to access our best chips. So they basically had to figure out how to do more with less. Whereas in the West here, we've taken largely a brute-
40:29 force approach to compute. We're going to try to create the most powerful chips as we can and then make as many of them as we can, and win on basically raw volume and horsepower. The Chinese have said, well, we've got to be smarter. We can't win on hardware.
40:45 The West has got an advantage there. So let's win on software. Let's write the software such that we get more from using less compute power. So, to your point, this is where my next question was going to go. You would think that the West would say, all right, well, rather than just brute force, why don't we do brute force and better software? And of course the market seems to be demanding that already at this point in time.
41:11 And if you do that, at the end of the day, what you need is less chips, or at least not as many chips as quickly as is currently being forecasted. And so that means less revenue for these companies than is currently projected out there. So do you feel like that wave is coming, that that will >> Yes.
41:30 And then even more dangerous than that is Vinod Khosla has a startup company that supposedly had a breakthrough this past week where they shrunk Qwen 3.6, which is another Chinese model from Alibaba, >> Okay. >> they shrunk it to run on an iPhone 17 Pro. So [laughter] >> wow. So you don't even need an expensive >> So you won't need the data center.
41:54 How about that one? [laughter] >> Wow. So that's going to put fright in Jensen Huang's dreams. >> Yes. As it should. As it should. So things are moving so rapidly. I've been involved, as we said, for nearly five decades. Things move rapidly all the time.
42:13 Obsolescence is a feature of the tech world. But I've never seen it move so fast than just in the last months, just this year. It's incredible. >> That's great. So another question just to make sure I'm understanding this right. What I just described is a threat to the top line, right? The revenues may be lower than currently forecasted, as efficiency starts driving everything. >> Then you were mentioning a squeeze from the bottom, which is that the
42:43 costs of running these data centers, we're not realizing the majority of those costs yet, so these companies look more profitable than they are, and so future profits are going to start shrinking because of these rising costs. So now we've got a profit risk from both directions, rising costs and shrinking revenues. You're nodding as I'm saying all this, but this is I guess maybe the foundation of your thesis. >> We do, and we can see it.
43:12 I mean, Meta came out just, again this is all happening in the last week, they came out and said that they're going to slash prices 75%. Well, we're seeing it, this isn't theoretical. This is happening and we can see the companies that are shifting. We can see the usage of the lower-priced models going up. We could see OpenAI, Anthropic slashing their prices and offering discounts and subsidies. It's all happening really fast.
43:29 But I don't know if anyone has really a model here that shows what the profits were going to be. They were just all building data centers to be number one, to be the winner of the race, >> which of course the government is incenting them to do too, right? Because it just wants to dominate China. From the government standpoint, it's an arms race.
44:03 >> Yeah. And of course the Chinese LLM makers are also subsidized as well, >> right? >> So yeah, and yet, as in your preface, everything relies here. The whole market is relying upon the success of these hyperscalers' buildups. >> Right. Right. The whole market, as I started the intro with, revolves around the axis of this pillar continuing to be there. And you're basically giving all this evidence why the pillar has a lot of cracks and it's built on sand >> and it's not just that, the rest of the economy is not good either.
44:46 So David Sacks, he's the Trump czar on AI, >> yeah >> he estimated that of the 2% GDP growth that we've been running at, and we had 2% this past quarter, 3.1% is the AI is driving 3.1% and everything else is 0.1%. So the rest of the economy, this AI buildout is driving all economic growth in the US right now.
45:14 >> So let me ask one other question around that. I'm right there with you, like the US economy and the US financial markets are all in on the AI trade continuing to boom from here. So of course if it doesn't, that's going to have really big negative repercussions. So this isn't my original thought.
45:37 I've heard it from folks like Jesse Felder and others, which is, the current forecast for capex buildout that the analysts have projected is basically, okay, if the capital's there, we can build out X many data centers. And I think they're probably going to be right. I think the capital will be there to build these data centers at whatever schedule we want to build it at.
46:04 I'm increasingly concerned that the limiter is just physics. Like there's only so much available land and permits and water and copper and capacitors and memory chips and public support, and those things are going to be the limiter on how fast this stuff grows, unless the economic models deteriorate even faster than that.
46:31 But that to me, just the limitations of the physical world, seem like a very underappreciated constraint here. Do you share that thought? >> Absolutely. We had $300 billion dollars worth of data centers that have been cancelled because of backlash from the authorities in states and cities and towns, >> right? Well, and the consumers themselves.
46:55 I live in Reno, which has been very data-center friendly. We just passed a moratorium on future data centers, >> right? And that's only growing. And the problem is that AI is seen very negatively. People do not see the benefits to them. There was a survey out that showed that 55% of US adults expect that AI will do more harm to them than good.
47:19 >> Right? >> 75% of adults believe that AI will cost them jobs. So we're seeing this backlash and you're seeing it politically. We're seeing social democrats winning, running on anti-AI campaigns. And so those are trip wires here too potentially. >> Yeah.
47:41 I've heard it said, I think very accurately, AI needs a much better PR firm, because right now the motto of, we're going to make your electricity more expensive and steal your job, that doesn't play. [laughter] >> It doesn't play. And so, but they just don't see it. When we didn't have this kind of backlash with the internet in 2000, people could see this is a big change.
48:02 Information at your fingertips, this is great, right? >> Well, enhanced search, which is basically what most people are getting. That's just not cutting it for them. When you're saying we're going to have water shortages, my electricity bills are going to soar, I'm worried about what it's going to do to my kids' minds.
48:20 All of those fears, jobs, those are those. So there's this huge backlash as well, and those are trip wires for the spending to continue too. >> Sure. >> All right. So for all of these reasons, and I've been saying this a lot in recent weeks, but you're just totally validating and affirming my point of view here. I think we're going to build out a ton of data centers over the next decade.
48:45 Maybe not as many as the analysts currently have forecast, but I think it's going to happen. I think AI is going to be transformative, and it sounds like you even feel long-term optimistic, you're going to buy these companies once the big correction happens here. But if the analysts are currently expecting a buildout at this rate, it's probably going to be more like this rate for all the reasons that we're talking about here.
49:09 So yes, that'll create a big corrective event in the AI space and we should watch out for that as investors, not be too exposed now, jump in to get values if and when they present themselves. But even at that lower rate of growth, they're going to need a lot of real things to make that happen.
49:30 And so if you are a believer that the AI buildout is going to happen just like the fiber optic buildout happened with the internet, or the railroads built out the roads, you're going to need a lot of things from the real world like copper and other key materials. So it's hard not to still be, even though we've got these concerns about AI, it's hard not to be a bull on commodities in this situation.
49:52 Do you feel the same? >> I'm not sure, yes and no. I'm concerned that when the internet was overbuilt and you ended up with a >> bunch of dark fiber >> you had, not just fiber, it was computer equipment, servers, storage, you had too much of everything there, and that led to malinvestment, it was excess, and there was a slowdown that occurred.
50:36 It wasn't that the internet wasn't useful. It obviously was a great revolutionary invention. And it's a similar thing with railroads where they just vastly overbuilt, and I'm not sure, I know they'll continue to build them now, I'm not sure they'll get anywhere close to where they thought they were going to get in terms of the estimates from like McKinsey is $6.
51:04 7 trillion dollars, almost $2 trillion dollars a year. I don't think they're going to get anywhere near this because the economics are just not going to work and there's going to be excesses here as a result of this. So I agree that there will be a continuing buildout, but nowhere near the pace, and it could be a period of time where there's excess. Certainly, we're seeing in DRAM, Micron's talking about $250 billion dollars of capex spending. Okay, the Chinese are coming public, the CXMT,
51:41 the CXMTs have an IPO offering coming out for $5 billion, and YMTC is another offering, and they're all building, Samsung, they're all building these semiconductor fabs for data centers essentially. I just think we're going to see oversupply of some things.
52:08 >> Okay, so I appreciate that, and I could definitely see the possibility of that happening. But I guess when I look at this, I look at our concerns about AI and then I look at the current valuations of these companies, the adjusted valuations of these companies to use your term, and it's like, wow, if that's the case, these things are just wicked overvalued right now.
52:26 >> Whereas I think you can look at a lot of commodity producers and say there's just not that froth in them right now. >> Correct. There's no doubt about that. So if you look at the positioning of investors, they're not in commodities. They're completely out of energy, for example.
52:48 The hedge funds have no positions whatsoever. And it's the lowest we've seen in a long time. And because they've been out of favor for a long time, and because it's difficult for a lot of reasons to build out capacity there, there isn't a lot of supply coming into those markets.
53:11 So I think they're going to be places that will do better for sure. They're undervalued. They're severely undervalued relative to the rest of the market. And as we saw in 2000, those value areas did very well. There was a shift. It was the best market for value stocks in that 2002 to 2004 period that anyone had seen in a long time.
53:40 And I think similarly you'll see a shift too into some of the most hated areas, and these companies are generating, they have low PEs usually, they're paying dividends, and they have good cash flows, so they'll be attractive to investors. So I'm going to ask you in a moment specifically where you see the most opportunity right now, but for what you just talked about there, it sounds like you expect to see a capital rotation at some point into the sector, probably because
54:21 people are starting to lose faith in this mega AI trend. Would you wait to move capital into this sector until there's a washout and investors get burned by the current trend and say, I'm going to look for something better? Or is it time to get in now? And I'm going to guess your answer is wait, because you just told me you're in as much cash as you've ever been in.
54:42 >> Right. So I have some energy positions. They're a lot lower than they were three or four years ago when everyone hated them, and then I built up significant positions. But I do like them. I'm just waiting. I'm concerned that everything's so warped here, that there's so much debt.
55:13 There's so much consumer debt, consumer savings have collapsed. People can't afford the cost. We have negative real wage growth. And this is all before you have some kind of market collapse potentially, if Grantham's right, but at least a severe correction, >> and so I'm concerned that demand might be weak for some of these things, and it took a little time, in 2000 it took a little time for the shift to occur.
55:49 For example, gold didn't get going until really 2002. And it was similar for other commodities out there. Gold isn't just a commodity, it's also a currency. But it was similar. It was a delayed reaction a bit. And so, if I didn't have any positions, I would have some. That's I'll say that. But I'm still looking for the moment.
56:17 I'd like to see the collapse start to happen and then I would start to buy. >> Okay. All right. So you said energy is a sector you like. >> I do, and I have positions in them, but I'm just holding them. >> Yeah, you're just holding them right now.
56:42 So one of the things that I've always found really interesting about you, Fred, is you follow the most cutting-edge industry, that's your primary day job, but for a very long time you've also been a big fan of one of the oldest assets out there in the world, which is gold. And you've been a very vocal, I think, champion for gold at cycle bottoms for gold,
57:10 and you've generally been proven very correct over your career. I think from comments you made earlier in this conversation that you think gold has brighter days ahead of it still. What is your current outlook for gold? >> So we had that correction. We've had a correction now for four months. >> And just to set table stakes, it got incredibly frothy.
57:33 Massive speculation. >> Yeah. And Chinese were involved too, Chinese speculators, and we've eliminated a lot of that froth, for example in gold stocks, which by the way Scotiabank came out and said are the cheapest they've seen them, they have 10 PEs on average for the gold mining stocks now, and it's the cheapest they've seen them in over four decades. But we eliminated a lot of the froth.
58:09 So there's something called the gold miners bullish percentage index, the BPGDM, that went to 100 in January, 100% bullish, and then it fell all the way to 2% here in this correction. 2%. And it's about the mid-20s right now, which is very low. I have another chart that shows, it's a really interesting chart.
58:36 It shows total world gold ETFs, change in total gold holdings by year, and this chart shows on the left side of the chart the inflows into ETFs, and this includes the GLD but also all of them around the world, and you saw steady inflows into gold ETFs all the way from 2005 all the way into the top in 2011-12, and then you saw the big decline that occurred.
59:11 We had a big selloff that happened and we saw outflows. But then look on the right side of the chart. It's all outflows except for last year. Yet we had this huge bull market just last year. And now we've seen outflows. We've seen 110 tons come out of the GLD, for example, just in the past six months here.
59:36 We never saw the retail participation that we saw leading up to that top in 2011 and 12. We've not seen that. And that's reflected in these outflows that occurred and have continued to occur. Now we've seen them again now, but now it's starting to see some stabilization in the last week or so, they've started to stop going down, and that's a sign that maybe we have reached the capitulation stage already.
1:00:04 And it's a similar thing with the open interest futures, the speculators, and I have another chart there that shows that, and the blue line there shows the number of future open contracts, and it went, at least in the last couple of months, at least to a 13-year low. So you can see all the way back to 2000 when gold was under $1,000.
1:00:40 We are lower today than we were then. We are at 378,000. It fell as low as 328,000. They are completely washed out, the speculators, out of gold. And so you could see there's a lot of potential buying that could occur. You could see what the contracts were at the top in 2020. It was almost 800,000 there, right? And you can see much higher, lower levels, but still very high in the 600,000 level ranges more recently on the right side of the chart. We're in the 300s right now.
1:01:14 So there's huge pent-up buying potentially from the speculators who are out of gold. So I think we've seen capitulation there. I think we've seen capitulation in the ETFs. And yet the things that were propelling gold were the central banks primarily, because it wasn't retail.
1:01:38 It wasn't retail that we saw in the past. It was primarily central banks, and the central banks are still buying. >> So we've had, China is accelerating their buying again. They're the biggest buyer. At the same time, I'll point out that they have cut their treasury holdings from $1.3 trillion down to 630, almost in half.
1:02:00 And they're continuing to sell treasuries and buy gold. And there's a chart here that shows, this is a World Gold Council chart, shows their intentions are to continue buying even more, the central banks continue to buy. I think it's, let me get it here, yeah, it's 45% expect to increase their holdings versus 43% last year and 29% in 24.
1:02:32 You can see it was only 8% back in 2019. They've been the primary propellant of gold here and they're continuing to buy. There's a whole host of these countries that have been buying and continue to buy. Poland's been an aggressive buyer, Pakistan, Uzbekistan, many of them are not US allies. Some are, like Czech Republic and Poland and Hungary.
1:02:56 But around the world we're seeing this continuing movement out of US treasuries into gold. And so now just recently we saw that central banks hold more money in gold than they do in US treasuries. And so we're going to continue to have that, even more so they're going to continue to buy and propel gold higher here and keep a floor to it too.
1:03:27 I think this is the reason why gold didn't get pushed below 4,000 for very long. It popped back up and it's been able to hold it even with today's decline with the activities going on in the Middle East. It's still holding, last I looked, still holding 4,000 or certainly isn't very far from. And so, >> just so you know, gold futures are at 4,000 on the dot right now.
1:03:48 We're speaking. >> Okay. On the dot. So anyways, they haven't been able to push them below 4,000. And it's been kind of just vacillating around this level for a period of time here. But we know the central banks are going to continue to buy. We know that de-dollarization is still a trend, and so central banks are moving out, countries are moving out of US dollars into gold, and I think that will accelerate once we have this decline in the stock market
1:04:27 because money will shift out, because, I'm not talking about the central banks, but people around the world have been pouring money into chasing the AI-related stocks. It's not just the US investors that have been doing that. I just saw, I think it was $900 billion, a recent period here, that they had poured in, foreign money going into US stocks.
1:04:52 And so I think when this thing busts, then you see a movement back into gold as well. >> Okay. So, totally follow the logic there. My thesis, which it sounds like to me you may agree with as well, is in the near term. So start of this year, ton of froth in the precious metals that had to get cleared out.
1:05:15 It largely got cleared out. But then at the end of February, kinetics started between the US and Iran and oil prices spiked hard, and you started to get some of these sovereign gold buyers, at least in the short term, start selling some of their gold reserves to pay for the higher oil price.
1:05:37 And so that was a depressive force on the price of gold. >> Correct. >> Okay. So you're with me there. I was personally starting to tweet a few weeks ago, or maybe even a week or so ago, maybe we're out of the woods here, folks, maybe the bottom might be in. But then we went back to kinetics with Iran and oil prices have come up a bit, and I think that's probably responsible for why gold is sitting at 4,000 right now when we're talking.
1:06:06 So okay, there's that dynamic going on. Would you expect to see a similar dynamic if and when this AI bust happens, where there's a big downward vortex in the market with the AI stocks, it pulls everything down with it, as we talked about, margin calls galore. So even though you expect gold to do well eventually coming out of that, do you expect there to be kind of a selloff as everything is getting sold off? >> That could happen, and it's one of the reasons why I've hesitated here, because I'm starting to see massive
1:06:37 inflows, massive buying in China for example, right? We saw it in May, a huge increase in imports. We're seeing those signs, but I am worried that we'll get a whoosh down. I don't think it will be a liquidity event to the extent that we saw in 2008, because the banking system isn't in as difficult a spot as it was then, but yes, we've seen it before, but usually it doesn't last very long. Even in 2008 gold rebounded quickly, maybe a couple of months of downturn, but I am, and that's one of the reasons
1:07:15 why I'm hesitant, that, and I've only started nibbling at certain things. It's because I am worried about that occurring. But I also am seeing these signs of maybe potentially bottoming and capitulation and the buying coming from China. It's enormous right now.
1:07:36 And it's much bigger than the declines that we're seeing from Turkish selling, some of the countries that have been under some pressure, their currency has been under pressure that they've had to do some selling of their gold reserves. So I think there's a potential that we start rallying anyways.
1:07:56 And if there is a decline, a big decline, I'll be ready to buy it because I have a lot. >> Okay. I'm glad to hear you're seeing those signs. And again, those are validative of some of what I was sensing. I've got to imagine some of your outlook though is tempered by saying, but it depends what happens with oil, right? >> Right.
1:08:13 >> If we go back to 100-plus dollar-a-barrel oil, I imagine you think that would be a bit of a constraint on the gold price. >> Yes and no. It would be, certainly the correlation has been that way. So the market correlation has been oil up, gold down, right? [clears throat] >> Down, and, you know, but remember we're running high inflation.
1:08:37 We've been running high inflation. We're at a 4% level PCE and CPI level. We've been running above the 2% level for now for five years. And I think that would lead to a lot higher inflation, not dissimilar to what we saw in the 1970s. So oil prices went up in the 70s, inflation went up in the 1970s and the Fed didn't react very quickly [laughter] and you had negative real rates, and we're already at near negative real rates, certainly in the low end when you're talking about 4 point
1:09:22 you know, three and a half to 3.75 fed funds rates, and so I think that inflation can be the bigger driver and that correlation can break, because that's what we've seen before. Gold can be an alternative asset to protect yourself from inflation. That's what was going on in the 1970s. [clears throat] >> Well then what's your thought then with a Kevin Warsh, where you've got a new sheriff who's basically saying I'll be the next Volcker if I have to.
1:09:54 You know, I'll >> I'm not afraid at all. Well, he came in and he had to talk tough to establish some credibility, but there's some gating problems he has to deal with. And I have the chart here of interest expense on the debt [laughter] and it's growing like topsy still. It was $1.35 trillion dollars. It's a record level, and I don't think [clears throat] he's going to, the market right now, and gold has priced this in too, the market assumes that instead of the 50% basis cuts they expected in the fed funds rate,
1:10:40 we ended up expectations shifting to 50% increases here. And that's, but here's the problem. $1.35 trillion, plus you have the $2 trillion annual debt that keeps growing and the rates are going up. They're only paying 3.3 to 3.4% on average rate, yet rates are going higher right now.
1:11:06 If he was to raise rates, this line goes even higher, making the deficit even bigger. And Mr. Trump, I don't think, would like it. And I don't think Mr. Warsh will do that. I think that he had some kind of a deal with Trump that he wouldn't be raising rates very much at all, and that he would more likely, if anything, cut back on the $6.
1:11:34 7 trillion balance sheet, but I don't think that he's going to, I don't think he's going to be tough and raise rates. I don't see it. >> Okay. All right. >> especially with the economy as weak as it is. >> Got it. Yeah, especially minus, if something starts compromising the AI capex spend, which you said is pretty much all of GDP growth, then it becomes even harder. Okay.
1:11:55 I hate to do this, I'm looking at the time. We've gone far over the hour. This, it's just always such a pleasure talking to you, Fred, because you're so knowledgeable. And you explain really complicated things very clearly. Thank you so much for coming on given your busy schedule and for being so specific about where you see opportunities.
1:12:16 So in a nutshell here as I wrap up, what I'm taking away is there's a lot of unpriced risk currently in the AI complex and a lot of reasons why you think that it's going to go through a pretty big correction at some point. Don't know when. My vibe from you is you think we're probably closer to the end of the story than the beginning, but you admittedly don't know for sure.
1:12:36 You're nodding as I'm saying this. So your advice to general investors is basically play defense. No need to be a hero right now. Don't get caught up in the FOMO. You personally are sitting on more cash than you ever have as a percentage of your net worth. Basically be patient and if indeed the correction happens, have dry capital to deploy into some of the surviving hyperscalers because you think they'll end up doing well in the long run.
1:13:06 But maybe just as importantly, you'll have dry capital to then put into some of these hard-asset sectors that you think capital's going to start shifting into. I got burned on the stove of AI. I want something that's a lot safer and interesting. And >> and cheaper. They're so much >> pardon me. >> And cheaper. They're so much cheaper.
1:13:24 >> And cheaper. Great point. So, maybe with some of those companies, maybe people want to start doing their research now to come up with some targets, maybe take some initial entry positions just in case there isn't that big of a dip, but you're kind of waiting for that big AI downdraft.
1:13:42 That's your signal to start pushing chips under the table meaningfully. >> Correct. >> Absolutely correct. Yes. The old saying, keep the bat under your shoulder and wait for the fat pitch. I'm waiting for the fat pitch. >> Waiting for the fat pitch. Love it. Ted Williams. The Science of Hitting. Okay.
1:13:59 So obviously there's probably a lot of people here saying, Fred's really smart, when he moves, I want to know when he moves and what he's putting his capital into. Folks can follow that by following your High-Tech Strategist newsletter, correct? So tell folks where they can go if they want to learn more about getting that.
1:14:19 >> Yeah, you can send us an email at thehightechstrategist.com. >> Okay. Send you an email and just say, hey Fred, I'm interested. And you'll >> all the details will be sent to you. >> Okay. And you also have an account on X that you post to, Fred, right? What's that? >> HTS at, I always forget that.
1:14:44 [laughter] >> I forgot it again. I meant to look it up. Yeah, HTS at, oh, you know what, you'll have to put it on the screen. >> You know what, Fred? I will put it up on the screen. Folks, I'll also put Fred's email and the link to his X account in the description below this video too, so you can get to those with just one click.
1:15:06 Well folks, please join me in thanking Fred for giving us so much of his time and expertise. Do that by hitting the like button and then clicking on the subscribe button below as well as that little bell icon right next to it. As a reminder, we're really trying to hit 200,000 subscribers for this channel by midsummer.
1:15:22 It's very helpful for us if we hit that milestone. Only takes you a couple seconds to click that subscribe button. And then lastly, if you would like to get some professional help and guidance for positioning for and taking advantage hopefully of future developments, especially if the future unfolds the way that Fred thinks it will.
1:15:44 If you don't already have a good financial adviser advising you, consider talking to one of the ones that Thoughtful Money endorses. These are the firms you see with me in this channel week in and week out. To do that, just fill out the very short form at thoughtfulmoney.com. Only takes you a couple seconds to fill out the form.
1:15:59 These consultations are totally free. There's no obligations involved. It's just a service these firms offer to be as helpful to as many investors like you as possible. Now, Fred, I can't thank you enough. It's always just such a pleasure, a joy, and a privilege to get to talk with you. Thanks so much for taking time in your busy schedule to do this.
1:16:16 And I look forward to seeing you hopefully next time on this channel at the Thoughtful Money fall online conference in mid-October. >> Yes. Yes. I'll look forward to it. >> All right. Well, can't thank you enough again, my friend. I'll see you then. Everybody else, thanks so much for watching. >> Bye.