Podcast: Play in new window
- NVIDIA Loses Over Half Trillion In One Day
- The World Needs Resources The Fed Can’t Print
- DeepSink Is AI’s Sputnik Moment
“So earnings growth in the AI space, admittedly meteoric if not miraculous through the early quarters of 2024. Earnings growth slowed considerably in late 2024 and as we come into this year. And if the supply chain for AI is put under the microscope and found to be creating overcapacity, you’ve got an eerie echo from 1999 and 2000.” —David McAlvany
Kevin: Welcome to the McAlvany Weekly Commentary. I’m Kevin Orrick, along with David McAlvany. Before we start, Dave, let’s go ahead and remind our listeners about the call with Doug Noland this week.
David: Yeah, join Doug Noland and me on Thursday afternoon, 4:00 p.m. Eastern, 2:00 Mountain, for the Tactical Short. This is a Q4 recap conference call, “Historic ’24 Excess Portends Precarious 2025.” I announced this call three weeks ago, and the word “precarious” seemed out of place. The comment might not have made sense.
Kevin: A little precarious now, though, huh?
David: Yeah.
Kevin: Yeah.
David: Do market conditions seemed more precarious? I think so. Perceptions shift, and with them market pricing does as well. I highly recommend that you join us. If you are inadequately hedged, you can still remedy that. Get informed, register for the call, submit your questions ahead of time, and we’ll do our best to address specific issues following the formal remarks. Short exposure has, as you might expect, been the inverse to the markets as they’ve been rising in recent years. You may not care about short exposure in a rising market without limits, but what about the limits? And what about a falling market? Are you adequately liquid? Do you have a form of financial insurance in place? We look forward to your presence on Thursday’s call.
Kevin: It’s important to talk about hedges. I was thinking this week, Dave, when we saw DeepSeek come out. We can talk about that through the show, but I love South Pole history. I love Ernest Shackleton, or Robert Falcon Scott, or Roald Amundsen, and it reminded me of the story. You probably remember this, but in January of 1912, a very well-stocked team of 65, it started with 65 people. The last five to push to the South Pole for the English, it was Robert Falcon Scott, and four other guys. 65 people, hardware heavy. They had dogs, they had ponies, they had motors that they had to move, but there was a shock when they got there on January 17th, 1912. A Norwegian team had gotten there first.
David: Almost a month earlier.
Kevin: A month earlier, and they had 19 guys—19 in, 19 out. No one was lost. The Norwegians beat them. It was very light on the hardware. Scott was heavy on the hardware, and unfortunately the five that went in and came back out, all five were lost for the Scott team. So I’m wondering, Dave, if NVIDIA isn’t a little bit like the Scott team right now, hardware heavy, and then you had DeepSeek this last week, come in and say, “Hey, you don’t need that much hardware.”
David: Well, it’s exactly right. You had a decline of $589 billion in a single day, and that marks the greatest concentrated single-company loss in financial history—
Kevin: Wow.
David: —in a 24-hour period.
Kevin: Almost $600 billion lost.
David: Forbes reports, “NVIDIA’s nearly $600 billion market cap loss Monday is larger than the individual market values of all but 13 American companies. More than the market cap of titans like health insurer United Health, oil giant ExxonMobil, and retailer Costco.
Kevin: Wow.
David: And at issue, if Chinese company DeepSeek can do what the large language models do at a fraction of the cost with a fraction of the hardware, then you’re looking at the AI supply chain in that Wile E. Coyote moment.
Kevin: Wow.
David: Gravity is in effect, and so the claim is that this more efficient open-source application was built for under $6 million, uses fewer than 10,000 chips, versus the hundreds of thousands, actually millions of chips required by its competitors at a cost stretching into the billions of dollars. So this would be the beginning of the AI revolution in terms of its applications and iterations globally. Can’t underscore enough the importance of that open-source nature, and it would be the beginning of the end for your AI tech firms that are pricing in 30 to 40 years of chip demand into the current market capitalization structure.
Marc Andreessen says, “This is the Sputnik moment. DeepSeek’s R1 is AI’s Sputnik moment.” And there was a gentleman who was in conversation with ChatGPT, and ChatGPT said that DeepSeek architecture replacing the dense large language models (LLMs), demand per unit of AI output is likely to decline by 80 to 90%, assuming similar productivity levels. So you’re talking about the graphic processing units that NVIDIA makes. If what DeepSeek says they can do they can actually do, you’re talking about a significant upset.
Kevin: It’s amazing that ChatGPT— Our listeners can go to ChatGPT and actually ask about the death of this version of AI being replaced by another. And it’s like, yeah, yeah, you’re right. Even NVIDIA came out and said, “Oh, well, this is a good development.”
David: Do you remember the conversation we had with Marc Faber and his classic quote, there’s a moment pause and he says, “We are doomed.”
Kevin: That’s right. That’s right.
David: Well, it’s interesting that DeepSeek announced its app the day of the inauguration and the week that Trump announces a $500 billion initiative to lead—and for the US to lead—the AI revolution in partnership with Larry Ellison of Oracle, Sam Altman, the team at ChatGPT. Don’t think for a minute that the politburo in China wasn’t briefed on the initiative, the announcement, and signed off on the timing. Mere coincidence? I think not. As the US asserts its global dominance via a more muscular Trump 2.0, overnight we’re not the leaders, we’re the laggards; not the leaders in this revolution—assuming that DeepSeek claims check out.
Kevin: Well, yeah, and Amundsen’s camp basically had a flag there with Norway right there on it. And in a way, I’m not going to cheer for a Chinese victory here, but we have to watch out, Dave. When something doesn’t feel like it’s quite right, usually something shows the vulnerability and shows that it wasn’t quite right.
David: Yeah, I think you’ll recall our comments in recent commentaries comparing the current period of AI enthusiasm to the dot-com era—
Kevin: Sure.
David: —in the internet revolution of the ’90s. And revolutionary it was. The underperformance in tech for the 15 years following the run-up, tied to overvaluation of those equities, requiring time to digest the blow-off levels of enthusiasm. That’s what happened next. Yes, it was a revolution. Yes, it’s changed life as we know it, how we function, the habits that we have, that we we social references like screenagers, which I have a few of them. That’s only possible because of the proliferation of the internet and ultimately the devices that complement it.
Ray Dalio concurs this week in a Financial Times interview. He says, “We’re in the cycle right now which is similar to where we were in 1998 and 1999. In other words, there’s a major new technology that certainly will change the world and be successful, but some people are confusing that with the investments being successful.”
He couldn’t have said it better. There’s the rub. The revolution is upon us, but what is likely to be disappointing is the investor expectations of continued success in those particular companies. The radical increase in efficiency implied by DeepSeek models imply a much faster and more widespread adoption of AI, but it’s far less constrained by energy consumption.
So we had a significant hit, not only to NVIDIA, but other companies in the AI space. And on the same day, NVIDIA shares drop like a stone. So did the energy companies with closest ties to the grid expansion needed to power data centers and needed to support the LLM proliferation. Constellation Energy, New Scale, NexGen, host of other nuclear names, were tagged for between 15 and 30% losses. And as a team, we’ve had for many years now an interest in uranium, but our interest in uranium, I’d say over the last 24 months, has been held in check by concerns that AI enthusiasm was mispricing those assets as well. And thus, we’ve tiptoed through that space. We are slated to expand exposure to the space, but have said repeatedly in the office that we need to see the AI bubble burst prior to major allocations being made. Maybe we get that green light in 2025.
Kevin: A good example of overpricing something too early and saying, “Well, you want to make sure that you’re part of the party.” Pets.com back in 1999 had more capitalization—market capitalization—than every airline in America, okay? It was a sock puppet. And to be honest, Dave, we buy our pet food online now. So the concept was absolutely right, but Pets.com, they’re no longer around. It was an overpricing in the dot-com bubble. They were wiped out. It didn’t mean that people weren’t going to be buying pet food later. Our dog and our cat get their food online, but not from the companies that were overvalued in 1999.
David: Yeah, I think the critical consideration is that markets are extra enthused by the nature of the revolution. Upside imaginations are unbound, and petty things like profits don’t matter.
Kevin: Yeah, profits. Yeah.
David: Who needs those when you’re running a company? So earnings growth in the AI space—admittedly meteoric if not miraculous through the early quarters of 2024. Earnings growth slowed considerably in late 2024 and as we come into this year. And if the supply chain for AI is put under the microscope and found to be creating overcapacity, you’ve got an eerie echo from 1999 and 2000.
Kevin: Do you remember when everybody wanted to have fiber optic cable? That was the thing. There was not enough at first. So that became the investment, and now look at it. As far as oversupply, can’t you just buy your own fiber optic cable if you want? There’s, what do they call it, dark?
David: Dark fiber.
Kevin: Yeah.
David: Underused, underutilized.
Kevin: Yeah.
David: Yeah, you basically have your own private network. Fiber was overbuilt. Corning benefited at the front end of that, and has never seen similar revenues, even 25 years later. If you believed last week, like Mark Zuckerberg did, that spending $65 billion in AI CapEx was necessary in 2025, the question is, do you still believe that today?
Kevin: After Monday, do you still believe it after Monday?
David: Can you get the same computational work done using different models at a fraction of the energy costs and a fraction of the hardware requirements? So if we use DeepSeek’s numbers, Zuckerberg won’t have to spend 65 billion to get the job done. He might reduce his costs to 10/1,000ths of a percent, right? So Zuck’s spend, to say it in reverse, is 10,833 times greater than it may need to be. That’s a big deal for the AI supply chain. You’ve got US cloud companies, which have estimated 250 billion in AI infrastructure spending this year, for 2025. If the capital necessary to scale their AI ambitions is significantly reduced, there will be a radical rethink over the next few weeks, and that comes right in the middle of tech earning season.
Kevin: Dave, I just wonder, what’s a trader to do? Because you remember back in 2007, Bear Stearns came out and they had some toxic debt. That was what we called it later: toxic debt. 400 million bucks, and it was like the cat was out of the bag, nobody bought it. And I’m wondering if DeepSeek isn’t a Bear Stearns moment for the markets, but the problem is, everything’s so overvalued at this point. Like you’re talking about. It’s not just NVIDIA, it’s everything tied to that. If there’s a far more efficient way to get to the South Pole and get out safely without all that hardware, what do you do as a trader right now? How do you play this game?
David: Yeah, the lean and mean operation has some appeal. 19 into the South Pole, 19 out, versus the larger operation, 65 who started the operation, supported the operation. Five ultimately launched to the South Pole, not a single one of those five survived.
Kevin: Well, that’s right. So what do you do? These companies cannot betray the fact that they’re scared.
David: Well, and I think traders can’t either. And this is exactly what happened in 2007 and 2008. As the banking sector was saying, at Bank of America at the time, “The music is playing, you’ve got to get up and dance.” It’s not an option not to play this. You have to play it out to the very end.
Kevin: Because fear is contagious.
David: Yeah. What is a trader to do? You panic quietly, right? You keep a stiff upper lip, you attempt to discount the DeepSeek’s claims while your concentrated positions in the bubble recover some value, play any bounce you can, and you exit those positions as discreetly as possible. So if the narrative has changed, you’d better not change your attitude or disposition publicly, because all you’re going to do is narrow the exit and increase the body flow through it.
Kevin: Well, and this highlights what you’ve been talking about now for months, and that is this concentrated trade that we have, Dave, the Mag-7. When you had 70% of all trading volume going into just seven stocks, and of course NVIDIA was one of the biggest.
David: We’ve harped on the danger implicit to concentrated flows into those narrow names, breadth within the market being the framework for analyzing how healthy or stable a bull trend is. And last week was no exception. We continued to see massive volumes into a concentrated few names. VanEck Semiconductor ETF added a billion dollars in new inflows. A double leveraged QQQ trading the NASDAQ-100 had inflows of 847 million, bringing it to a total of $8.2 billion, so almost a billion dollars last week. They’re buying the ultra bullish tech exposures and they’re doing it just prior to a sell off, and frankly, that’s not uncommon. Last week was no exception. The enthusiasm is there up until the last.
Kevin: Okay, so now what? What do you do if you have some control over the market?
David: Foment a rally. You have to foment a rally. You must, by any means necessary. You have to increase portfolio liquidity while conveying high conviction in your existing positions to make sure that the lemmings parade continues and the investor on the street provides you with liquidity to get out, and ignorantly they assume the bag holder role. So hedge funds, over-leveraged players, they will fight for a rally. They will reiterate the revolutionary nature, and that is the question, how revolutionary and who will lead from this point forward?
Kevin: When we first started, you said that there are only 10 companies on the exchange that are larger than the losses of what occurred on Monday with NVIDIA. Let’s put this in perspective, though, with gold. Okay, so we know that this is a large loss as far as Wall Street goes, but how does it compare to gold, like the ETFs?
David: Yeah, for comparison’s sake, if you took every ounce of gold currently held in all ETF products globally, including the US, Europe, Asia, take that number, 271 billion, double it, 542 billion. It’s still less than NVIDIA’s single day of losses on Monday.
Kevin: Wow. Wow.
David: So—
Kevin: All the gold ETFs.
David: Yeah. Again, going back to the Forbes report, single day losses in NVIDIA surpassing the total market caps of companies like United Health, ExxonMobil, Costco. NVIDIA gave up the highest market cap in the marketplace this week, and with the feverish enthusiasm for AI pressing the pause button, it’s fair now to ask whether the markets are overvalued.
Keep in mind, last week that would not have been the case. Infinite upside implies that they are not overvalued. But if you have finite upside, that might suggest something very different. The narrative has shifted. The narrative has shifted. Can Wall Street recover from the shift and somehow maintain upside momentum? I think that would require an incredibly healthy flow of capital continuing to rush in, as it was through the end of last week.
Kevin: Won’t you need continual new buyers to propel that forward?
David: Yep, prices rise till you run out of buyers, and prices fall till you run out of sellers. And Monday was not the first time buyers have gone on strike, just the first time in recent memory. Now to the exits, right? So if you’re an institutional allocator, you have to ask the question at this point, should I lighten my position? Should I hedge my position? Both actions exert downside pressure to the price, and you really do need to foment a rally in order to exit.
Kevin: As we’ve talked through these things though through 2024, there’s really nothing new, is there? The market has been overvalued for a while. We just wondered what would be the trigger where it actually had to revalue to the correct level?
David: Yep. Market experts on the way into the office this morning had nothing to say but, “Buy the dip. Buy the dip. Buy the dip.”
Kevin: Right.
David: Mohamed El-Erian wrote in Bloomberg that “the drivers of 2024″—the financial market landscape—”remain equally relevant in 2025, ultimately influencing greater market volatility.” And that was the main theme from last week’s program. Volatility is the theme for 2025. The three things he pointed out were AI focused technological optimism as the first. The second, a hope for leveling up, where the broader markets benefit from economic growth as a result of tech-driven productivity enhancement. And then the third, higher government bond yields. And El-Erian puts a focus on the current negative equity risk premium, which was last seen in 2002. In other words, you are not being compensated extra for your equity exposures.
Kevin: For the risk you’re taking, you’re not making enough money to be there.
David: Yeah, as the variable is calculated, you should be compensated an extra two, three, even four percentage points for owning equity versus debt. But right now, the math implies a negative 55 basis points instead. And that’s, again, as a result of forward earnings yields of 3.9% slipping below the yield on the 10-year Treasury very recently at 4.65.
So the January 23rd Financial Times article also discussed the same issue. The title of that article, “US stocks at most expensive relative to bonds since the dot-com era.” Guess what? It didn’t matter on the 23rd of January with the AI narrative intact. Guess what? It matters today.
Kevin: All of a sudden.
David: Now, instead of sanguine justifications for nosebleed-level valuations, analysts will have to consider that we have the mother of all bubbles and are dealing with end-of-cycle dynamics, not a new era where the old rules are tossed out. So last week we asked the question, is this time different? Well, is it?
Kevin: Watching football this weekend, there’s a commercial on where this sport team is in a halftime and they’re being beaten 78 to nothing, right? And the coach gives this resounding, cheerleading speech to the team that is zero, being beaten by 78 to zero. And one of the players goes, “Yes, let’s go get them.” Well, the end of the commercial is they still absolutely do not win. They’re just crushed.
What makes me think of that, Dave, is Wall Street is going to have its cheerleaders after this. These guys are going to go out and even though behind them they’ve got disaster, they’re 78 to zero, right? They’re going to come out and they’re going to say, “Hey,” like you said, “Buy the dip.” Maybe they meant “be the dip.” Be the dip or buy the dip. Yeah, I guess it’s just how you spell it.
David: Well, things in markets are cyclical. Earnings growth is cyclical. Margins are cyclical. Asset prices are cyclical. It’s a pretty easy bet that mean reversion kicks in at some point. All you need is a catalyst. You have Goldman Sachs analyst Ben Snyder, who claims that earnings growth is real and that should be the driver of equity prices to even higher levels. He’s not—
Kevin: You go get them. You go get them.
David: —concerned at all, right? So looking at P/E ratios, he claims the S&P is, and I quote, “in line with our modeled fair value.” Now, we’ve talked about the cyclically adjusted price earnings multiple. We’ve talked about the standard price earnings multiple as being in the 98th to 99th percentile, okay? That would be—categorically—we’re talking two and a half to three standard deviations above the mean, and yet somehow Goldman’s coming out with a “modeled fair value.”
Kevin: Right.
David: So being two and a half to three standard deviations above the mean is now in line with modeled fair value?
Kevin: Can’t you model anything you want, though? Can’t you get a model to say whatever you want?
David: That’s right. You can model whatever you want to justify anything. Maybe he meant to say that the equal-weighted S&P is at fair value.
Kevin: All the other stocks, the loser stocks that haven’t really been doing what the Mag-7 have.
David: And they are closer to, closer to—
Kevin: Right. Yeah.
David: —fair value. Or maybe his models were generated by a large language model and their learning inputs were all from Abby Cohen. Do you remember Abby?
Kevin: She’s always bullish.
David: Yeah. Goldman’s own perma-bull and tech hypnotist. In fairness she’s retired from Goldman, but her market hallucinations may very well feed what we have today as analyst hallucinations via ChatGPT. Models are only as good as the data they’re fed and the assumptions they’re built on.
Kevin: So let me ask, okay, with yields being where they are on US Treasurys, which has been a safe— I’m talking short Treasurys now, on 90-day to two-year Treasurys. Why would you go get less out of the stock market right now? Because El-Erian was basically saying, “You’re not getting paid to be there.”
David: Right, so we go back to El-Erian’s comments. Higher yields are a significant contributor to volatility, and I think there’s no greater example of that than the pressure you’re seeing emerge within the private equity space. We noted that corporate bankruptcies rose last year. Well, what we didn’t note, a few weeks back, is that that included 110 private equity-backed companies which failed to work through their liability management exercises, which is just a fancy way—LMEs are a fancy way—of saying, “We couldn’t make it, and we had to renegotiate the terms of our debt. We extended, we got forbearance from our lenders.”
There is a scramble for renegotiating terms as we speak, because floating-rate debt has become unmanageable in the private equity space. At the same time, recapitalizations are happening left and right. You think you know what recapitalizations are, well, when the original investors in a private equity deal are unhappy because they’re tied up in a company they don’t want to continue to own—
Kevin: They’ll find other ways to get paid out.
David: Yeah, they can’t get liquid. They can’t take the company and do an initial public offering. They’ve been unable to exit their portfolio companies at premium multiples. The partners vote for huge special dividends—
Kevin: And that’s more debt, right?
David: —funded from even more debt issuance. So you’re talking about entities leveraged to the hilt. That’s actually how the Financial Times describes private equity firms currently. That is, after all, the nature of an LBO. We called it private equity. Let’s not forget that in an earlier iteration they were called leveraged buyouts, and that’s exactly what they are. Leveraged to the hilt is what you start with. Overindebted companies then with these special dividends—the recaps—being saddled with even more debt just so that investors can get liquid without a liquidity event happening via an IPO or other form of exit. 2025’s going to be very interesting, to say the least.
Kevin: When you were talking about panic quietly, you’re not talking to our listeners because most of these listeners understand this. But instead of panicking quietly, for those who’ve seen this coming, again, US Treasurys, short-term Treasurys, gold, real things. Morgan, who was talking at the meeting this morning, I loved his quote. I’m not sure if it was his or somebody else’s, but he said, “The world needs resources that the Fed can’t print.” We’ve got an awful lot of inflated, fake things out there like overvaluations on artificial intelligence, tech stocks, that type of thing. Unfortunately, we’ve got $36 trillion worth of debt in the Fed. We’re just going to have to print our way out. You can’t earn your way out completely. So the world needs resources that the Fed can’t print. This is a really good time, I would imagine, Dave, to move from paper to real things, like McAlvany Wealth Management, gold, what have you.
David: Our hard asset thesis is playing out superbly in real time. And the last puzzle piece, if you will, is the overconfidence in the equity markets in financial assets, which has been extended. That enthusiasm has been extended by the AI craze in what is a bubble. So where we go next I think is absolutely critical. An emphasis, a reallocation, a diversification towards infrastructure, towards real estate, to a degree, towards natural resources, towards precious metals. These are things that do well in an elevated interest rate environment, do well in a sticky inflation environment, and do well in the context of broader market unwinds.
* * *
You’ve been listening to the McAlvany Weekly Commentary. Make sure that you join Doug Noland and Dave on Thursday afternoon at 4:00 p.m. Eastern Time, 2:00 p.m. Mountain Time. Go to mcalvany.com for that. Don’t forget to call us if you have any questions at 800-525-9556.
This has been the McAlvany Weekly Commentary. The views expressed should not be considered to be a solicitation or a recommendation for your investment portfolio. You should consult a professional financial advisor to assess your suitability for risk and investment. Join us again next week for a new edition of the McAlvany Weekly Commentary.
