Mike Green: AI is fueling another 1987 market CRASH!

Well, I want to be clear. I'm very bullish on the application of the technology. I actually think that this is a true genuine innovation. I actually was sitting and having breakfast earlier this morning listening to a young woman explain to somebody the stacks that she was building within her AI applications. And I was thinking to myself that you certainly didn't hear anything like this with crypto or really anything else. It reminds me very much of the rollout of the internet, which had a very profound impact, was clearly formative in terms of my perspective on this stuff. But I do think that there is a very material misunderstanding between the economics that are being presented and the likely economic reality that is going to emerge.

I think that there are a couple of things that are really critical to understand. One is that the adoption curves that we have seen to this point have largely been about people just seeking advice in their daily lives. Right? They're not building broad applications. They're not revolutionizing the world in any material way. They're simply saying, where should I go on vacation? Which hotel should I stay at? How should I plan my trip? Etc. I like to give the example: my mother-in-law asked me to install ChatGPT on her iPad. And the very first question she asks is not, you know, how do I build a multi-factor agent to simulate my retirement flows or whatever. Instead, she actually said, 'How old is Julianne Moore?' Right? It's a replacement for Google search. It is better than Google search. It does not have the same ads that have become, you know, completely endemic in the context of something like Google Search, although you're starting to see that flow over into things like Gemini as they seek for ways to pay for it.

But the reality is people have an endless demand for opinions or insight into their lives. It's the reason why the Dear Abby columns were so popular within newspapers for years and years and years. And this has simply taken that and elevated it in a variety of ways, right? It can be your therapist. It can be your girlfriend. It can be your boyfriend. It can be the person who explains why you're not the star in the conversation. Everybody else is. These are all things that people have desperately sought throughout time. And the fact that it's available in a personalized, non-recourse manner, meaning there isn't somebody thinking ill of you for asking the question, by and large, is completely validated by the adoption curves that we see. Right? People want that information and they love it when it's subsidized at very cheap costs.

Now, it's somewhat absurd that we're seeing the AI companies lose thousands of dollars per user in order to provide this type of advice. Newspapers subsidize that with subscription fees and with advertising. My hunch is we'll ultimately end up seeing something very similar, but it's going to be at a much lower cost because people don't value advice at a couple of thousand a month. They value it at $20 a month or ad-supported or whatever. That I think is going to create a real challenge for the AI space is effectively separating the two.

The much more impactful component is going to come from an overtime realization that many of the procedures and processes that we're now building agents to start to replicate, things like a call center or a customer service helpline, those can be totally rethought in a world of agents. We can answer questions that may not actually be asked or maybe haven't been asked yet. A really simple example: a call center for customer support currently exists for you as the user to call up and complain and say, 'It's not working. What do I do about it?' An AI-based agent at zero cost could actually follow up with a customer who made a purchase and say, 'Hey, I noticed you haven't brought it online and registered it yet. Was this a gift? If so, you know, here's a coupon to get another one for yourself. And if it is not a gift and you're experiencing troubles, how could we proactively begin to address it? What could we incorporate?' Right? That would be rethinking things.

And the closest parallel that I can draw to that is the early stages of the 19th century when manufacturing and factories moved from places where you pulled together the factors of production to the assembly line when you fundamentally rethought what you were actually doing. That I think is the transition that sits ahead of us with AI. And I think it's actually incredibly exciting and will likely deliver an unbelievable number of services and tools that individuals can use at prices that are far lower than we're currently able to access them.

Think about a fiduciary agent that replaces an investment adviser with your own personalized agent operating at extraordinarily high levels of intelligence with unbelievable access to all the information, but most importantly, information to your own life. Right? That changes the process materially. Likewise, the process of buying real estate. I currently get an agent who ostensibly is going to show me the available properties. Well, since MLS has existed, the vast majority of people know more than their real estate agent because they're using Zillow and Realtor to find homes that they've pre-selected. And so we're still paying somebody a sizable fraction to effectively shepherd us through a transaction that becomes immediately available for AI disintermediation, and the process could materially change in that manner. We simply don't know. There's unrevealed preferences that exist out there that we haven't yet begun to explore.

So I'm very bullish on the rollout of it and the opportunity that it has. I just think that the current path that we're following, in which unbelievable amounts of money are being expended in pursuit of, you know, a TAM that is described as all of space and time basically, are highly unlikely to generate the returns for the current investments that exist.

Well, unfortunately, it's become self-referential in its underlying components. So if you think about Nvidia's margins, effectively what's happening is Nvidia is selling out its entire production. It's creating scarcity. It has scarcity for its chips, but it's manufacturing that scarcity by financing its customers. It's the same underlying phenomenon as the customer financing that existed, the vendor financing that existed under Cisco's buildout in the late 1990s.

Ultimately, what you needed if you wanted to launch a website and compete for the best online shopping experience or the online customer experience in 1999, you needed servers, you needed routers, you needed LAN equipment, etc., because you had to build that all yourself. Cisco stood ready and willing to finance it. That meant that the customers that were being financed didn't have the opportunity to negotiate on price. Now, that is really what has driven Nvidia's margins to such extraordinary levels. And when you finance your customer, you don't write down the margin that you're actually realizing on the credit riskiness of the underlying business. You simply book the sale and accept the fact that the paper that you issued is held currently at par.

If it turns out that the opportunities don't generate the revenues and profits and cash flow that are necessary to service those obligations, there will be a write-down in the future that will reverse those profit margins. But that happens off into the future. And so when we look at what's actually occurring here, I'd suggest that there's two real things that people should be paying very close attention to.

One is in the late 1990s, the US competitors to Cisco and Juniper Networks and JDS Uniphase, etc., were largely Canadian and Finnish companies, companies like Nokia or like Nortel. I'm much less worried about competition coming from Finland or from Canada than I am from China. And China, in part because of the policies that we've pursued where we've tried to behave in an exclusionary manner, has had an entirely different approach to it, which has been to focus on efficiency. How do I use less memory? How do I use less processing power? How do I do this more cheaply and more effectively? And that's always facilitated by the second-mover advantage.

We've forgotten the components of the dot-com cycle and the insights that the second mover often has the advantage of not having to use the bleeding-edge technology and incur the costs associated with it. Ultimately, we're now starting to see an explosion of the usage of those Chinese models. It's a little bit like watching Cray Computer, which was the supercomputer manufacturer, compete against Sun Microsystems or Intel, where the capacity and capabilities underneath in the lower end of the market were appreciating rapidly and effectively cutting off larger and larger share of the available market.

The frontier labs are finding themselves in the same situation and ultimately are now taking actions that are designed basically to lock out the rest of the market. Whether that's agreements with the US government or whether it's embedding within Fable and Mythos, the inability to work on competing frontier models, these are all steps that are being taken by these companies because they have to protect their end market and the rationale for it simply because their valuations have become so absurd. That's very much a self-defeating process.

And so I think there's a very real risk that just like we saw with every other playbook that China has pursued, whether it's solar or nuclear or anything else, that the US runs a very real chance of squandering the lead that we've created because we're unwilling to actually experiment and drive competitively derived outcomes.

Well, I want to be perfectly clear. I think that part of what is actually happening is just that investors have quote-unquote figured out the game. So this is not dissimilar to the late 1990s when we had all sorts of theories around why is the stock market doing really well? Why were valuations rising? Jeremy Grantham, Warren Buffett, etc., wrote, you know, very profoundly in late 1999 about the exceptional margins that were being realized in US equities. They were dead wrong in terms of the expectation for forward earnings, but they were very right in terms of the valuations that we were ultimately paying for those things.

Now, we're actually facing a much more complex situation where we're embedding extraordinary margins. The gross margins that Nvidia is generating under any reasonable model should attract tremendous competition and replacement, which is exactly what we're seeing from the Chinese models. But there's also no reason to recognize that because the vast majority of investors are simply buying the US stock market right in line with what they're expected to do through their 401k. They're buying into a target-date fund. That target-date fund is allocating not on the basis of valuation or forward expected returns but simply by using the historical return profile to suggest this is what we expect in the future.

Every investment prospectus out there has the disclaimer: past performance is no guarantee of future results. And yet that's how we invest. And so, you know, you and I are very focused on this. We're very interested in this. 65% of American households that have 401ks should be very interested in the intricacies of this. But I'm going to guess it's about 5% that actually have any awareness of what is really going on.

Well, what's really interesting about it is if you think about passive, it basically reinforces whatever the last price was. It accepts it and provides an additional bid. And so there's a theory under what's called the Grossman-Stiglitz framework that ultimately the active discretionary manager should be competing away those opportunities. The problem is because passive vehicles like 401ks receive continuous inflows tied to the American retirement system, they effectively are able to exert continuous shocks.

It's a little bit like fantasizing that a phenomenal right-hook swinger can fight back against a guy who's an inside guy, but if that inside guy never lets up and never gives him a chance to draw back his punch, the game is over. The boxing match is going to be over before he gets a chance to take a swing. That's really the phenomenon that we're seeing in the financial markets today, which is an earnings report comes in, it's exceptional. That causes short covering from the skeptics. It causes enthusiasm and continued holding from the proponents of it. And into that changed price, passive has to step in to fulfill its fiduciary obligation to meet its allocation. It doesn't actually do any analysis. It's not thinking about it in the same way. It's simply reinforcing that price and effectively turning it into a higher floor.

It's what's called idiosyncratic volatility and it reinforces it. This is actually a paper written in 2023 by Jen at Michigan State that highlights this component of idiosyncratic volatility, and we absolutely see that in the research that we do, that once these sort of shocks occur, passive effectively becomes a propagation mechanism for it.

Yeah, I think the earnings are wrong, right? I think they are overstated certainly relative to what we are likely to see on a forward basis, but that's my opinion. That's not fact. What I would highlight on those earnings is there's no real mechanism for them to actually drive stock prices. Right? A company reports earnings, that doesn't change the stock price. What changes the stock price is people's reaction to those earnings. And so there's still a sizable fraction of the investment public that responds to those earnings and says, 'Oh my gosh, I got to get into this.' I would suggest that a sizable fraction of those responding in that manner are actually people who are covering shorts. We've seen them taken out over and over and over again as they have confidently tried to express views on valuations being completely absurd.

But, you know, eventually we enter into a regime, and I'd suggest that we're there somewhat now, where things like triple-levered ETFs begin to emerge that reinforce and magnify the impact of flows and ultimately create the conditions similar to what we saw in the late 1990s in which narrowly focused funds, with Cathie Wood and ARK, narrowly focused funds are attracting extraordinary pools of capital and driving those individual securities to valuations that were previously unimaginable.

I don't think anybody realistically expects Micron's or SK Hynix's margins to maintain at these levels for any period of time, or anyone who has studied the cyclicality around these types of businesses expects them to be persistent and maintained. The question is kind of to what magnitude do they ultimately go, because the corollary to Nvidia financing your purchase of GPUs is those GPUs are useless without the memory to go with it. So of course you're going to pay whatever is necessary to get this.

It also has become a somewhat convenient excuse for companies like Apple to announce price increases where they've more than doubled effectively the price increase associated with the memory because they're unwilling to cede the margins that they have begun to rely on increasingly. That in turn is creating conditions under which a growing fraction of people are looking at their iPhone and saying, 'I may not replace this next one.'

Well, I think that there's two connections to it. One is 1987, part of the boom that was being driven was actually being driven by the false security of portfolio insurance. If you believed your portfolio was protected to the downside, you're willing to allocate more to a risky asset. 1987 was effectively the saturation of liquidity in those markets by these types of strategies. That's what Paul Tudor Jones really identified. If you ever watch the documentary Trader, you'll see all sorts of nonsense about him, you know, tracking each individual movement and comparing it to 1929. And I know any number of people have produced overlay charts since that are probably even higher fidelity in the world of, you know, computer replication as compared to the hand-drawn charts that he was working off of that Peter Borson had created.

But, you know, the simple reality is that isn't actually what happened. What really happened was that the liquidity in the futures market was exhausted by the size of the portfolio insurance need to sell that overwhelmed the system. Specialists came under pressure in terms of their ability to finance the book and that caused the ultimate collapse in liquidity that we think of as 1987. It wasn't fixed by cutting interest rates. It was fixed by Gerald Corrigan, the president of the New York Fed, telling all the chartered banks in the United States effectively, 'You will lend money to these specialists or you will lose your charter.' Those were two choices, right? Obviously, the lines of credit reopened and liquidity was restored.

This is a very similar setup where we have a relatively new technology, index investing and ETFs, that have combined with a general awareness and perception that the market never goes down for any sustainable period of time, that has led people to increase their allocations to equities so that today they represent a far larger share of total household wealth than we've ever seen before. The mechanism for that and whether that continues, I would argue, is largely driven by the phenomenon of passive investing and how we've chosen to invest, and therefore it's subject to the flow components of the retirement of the boomers, the potential for an increase in unemployment, etc. There's a variety of components that could cause that flow to reverse.

The question is, does that play out in the same fashion? And ultimately, I would suggest we may be even worse off because the liquidity provision in the market today relies on what are called market makers, effectively individuals who no longer have the obligation to maintain liquid markets that specialists had. A specialist could get fired by the firm that they specialized in. I worked for Spear, Leeds and Kellogg, which was famously fired by Morgan Guaranty, a company now absorbed long ago. But Morgan Guaranty fired them for failing to maintain liquidity in 1987. They were incapable of putting up enough capital, but they still lost a business franchise because of it.

Today's specialists have the opportunity but not the obligation to maintain that liquidity. They can simply walk away with no damage to their franchise and their balance sheets held intact. That creates a risk that liquidity evaporates suddenly and quickly without the traditional warning signs.

Well, I think we have evidence of many things breaking. The question is whether the aggregate system breaks. So, we've now seen Korea close multiple times in the past few days because the growth of capital into things like SK Hynix or the Korean memory stocks has actually created volatility that is candidly uninvestable. And that is actually one of the great ironies is unfortunately in the pursuit of making the markets more investable and more accessible to the average individual, we've destroyed the underlying link to fundamentals that historically would have allowed us to properly price the cost of capital for publicly traded companies. Korea is just the most egregious example of that.

I was looking at the SOXL triple-levered ETF referencing US semiconductor stocks that was down 20% last time I checked, right? That's not an investable framework. It's been up and down 20% and because it is a levered vehicle, it actually creates its own endogenous flows, which is something that people generally struggle with to fully appreciate, but I'll give a really simple example.

So, let's imagine we have a 3x levered semiconductor ETF. We won't name it, but let's assume we've named it. You put $100 million worth of equity in there. It actually has $300 million worth of purchasing power because it's triple-levered. If the underlying index rises 10%, the total value of the portfolio has now risen to 330, but the debt outstanding was only the original 200. As a result, you have an additional 30 of equity. That means you actually need to go out and borrow an additional $60 million to buy the exposure that you promised your investors.

So, it can create its own endogenous flows. Those flows impact the securities, driving the prices higher and theoretically taking the demand even higher, the manufactured flows even higher. The next day it falls 10%. The reverse process occurs. You now actually have to sell an additional amount to maintain the 3x levered exposure relative to what you otherwise would have had to do. That type of volatility and that type of flow is actually exactly the sort of characteristic that you saw in events like 1987 in which selling begets more selling, buying begets more buying.

And so those components I think are very much in place in some areas of the market in which they're already, I would describe as fully broken. I would also extend that analysis to things like the VIX pricing itself because if you look at the implications of large companies like Nvidia or Micron that are both in the S&P 500 and in the SOXL, if they continue to get their pricing behavior largely determined by this triple-levered ETF, which in that example created $60 billion worth of flows. To put that in perspective, the daily flow associated with traditional passive is only two to three billion. So we're talking magnitudes directed at a much smaller set of companies that are not weighted on a market-cap-weighted basis.

It's modified market-cap weighting, equal-weighting type framework to give a diversified exposure. So companies like Broadcom or Micron have massive overweights in those components. As that becomes the determining factor for that subsegment of securities, it loses correlation with the S&P 500. That drop in correlation then lowers the price of implied volatility. And the question is ultimately do these triple-levered or levered-type vehicles ultimately end up the path that almost all levered vehicles do, which is that volatility drag that is magnified by the leverage ultimately drives them to zero even if the underlying is actually going up.

And I think this is the part that is quite, you know, most concerning about that is just that ultimately we know that these pools of capital are not sustainable, the volatility drag associated with that. And just to identify what volatility drag is, if you think about it from an arithmetic average standpoint, if I'm up 10% today and down 10% tomorrow, or down 10% today and up 10% tomorrow, I'm flat. If I run that in a continuous series, if I'm down 10% today and up 10% tomorrow, I'm still down 1%. Right? 1 * 0.9 * 1.1 is 0.99. I'm down 1%. That's what we refer to as volatility drag. You add leverage to that, that gets even larger.

And so these are largely self-liquidating pools of capital that are currently causing havoc in the market. As they become less impactful, that implied correlation will actually reassert itself because we're seeing less impact from these external flows of capital. That will cause a rise in implied volatility, potentially creating conditions under which people are forced to de-risk because of the higher realized and implied levels of volatility.