Loading summary
A
Foreign.
B
Welcome to the Synopsis, a business and investing podcast. I am back again. I got another comment singing my praise and that's all it takes for Drew to get me back on for another week. We'll see if I make it back next week. But I'm excited because guess what? Drew's making me talk about AI again. Here we go.
A
Favorite topic. He's just, he's just always like, why can't we talk more about the same topic? And also software a lot. But good news, no software today.
B
There's no software. You know, I actually like this topic. So today we're talking about a recent YouTube video you made, which I think was a very honest take about the AI market. And so the kind of content of the video is. Is AI a bubble, which again, you hear about a lot. I know we talked about it last video briefly. I think I made some points around that. But I think what's compelling is your dual sided take. And you know, we, last time we brought up a little bit and I jog my memory here, we talked about Jeremy Grantham as, I mean, I don't mean to pick on him. I like reading his stuff. I think he makes some compelling points. But we talk about kind of the value investing or quality investing community, however you want to put it. Looking at tech companies that do well or these environments and kind of just like hoping for a bubble like sitting there seething that they're, you know, cyclical. Four times earning stocks aren't ripping like Micron is and there becomes kind of an emotional bitterness to it, which is, which is hard. They're like, this is just like the dot com bubble. Everything's going down 90%. Everyone's so dumb. I don't think that's a compelling way to look at things. Honestly. I think if you look at, and we can historically get into kind of each bubble, you brought up a few. But like all things in life, stocks aren't down for 70% for no reason. They're not up 70% for no reason. There aren't these manias or bubbles, however you want to call it, for no reason. I mean going back to the dot com bubble, you go to the back to the South Sea bubble, you can go back to the, I don't know, maybe the tulip bubble. That's when just tulips, you know, the flowers were very valuable. I don't know if I see that one.
A
No, but again this.
B
So there are.
A
Hold on, you're just going all over the place. So there are two kinds of bubbles. There are financial bubbles. And then there are technological bubbles. So there are bubbles that form without any kind of underlying technology. If we think about what happened in mortgage backed securities, the real estate bubble that happened in 2005-2007, that was very much a byproduct of people just getting caught up in valuations and thinking they could flip homes and keep going up and up indefinitely. So there are bubbles that exist without a technological backing. Having said that, anytime there is a new and innovative technology, it has always, always led eventually to a bubble. So that is kind of the way I would frame that. And if we're thinking about AI and the reason why I think a lot of people think AI is a bubble, and I honestly thought it was more of a bubble or we were further along in a bubble before I did this video. But because just simply the behavior we've seen of a lot of investors and the way a lot of stocks have moved, that felt kind of the most bubble. Like to me, when you're saying that cyclical companies that have been cyclical for decades are never going to be cyclical again. When you're seeing stocks move up 30% on, you know, a single earnings report, when you're seeing companies like the memory companies go up, you know, a thousand percent in the order of 18 months, something like that, that to me was the primary sort of thing I was looking at that made it feel kind of bubble like now. When I actually went through the video though, and was doing analysis for it, I realized there's a lot of things that are different about this current situation with AI and what we saw happen in the last tech bubble, for instance,
B
I think the distinction is meaningful here because again, there's usually reasons and there's usually reasons, and intelligent people can have strong opinions about outcomes that make sense and are grounded in reality. I think we go all the way back to. And I know you're saying I was a little off topic, but if we think about like some of the East Indian trading company and the South Sea bubble and things that, that were happening of that nature, part of that hype, which again, famously Isaac Newton was actually caught up in the South Sea bubble, for those who don't know. So, you know, very intelligent people are susceptible to these financial situations. But again, what was the excitement there?
A
It was, did that one make you feel better?
B
That one did make me feel better. Yeah. No, I like having some information and I was trying to get there because I wanted to drop some facts on some people. But again, if you look at those types of Bubbles which were the discovery of new places, new resources. Again, the entrepreneurial aspect is almost like venture betting of the new trade routes. And hey, is this trade route gonna work out? It wasn't like there was nothing there, right? It wasn't like there was nothing there. And I think again we're back into this AI world where it's an exciting technology, it's I would say again, new unchartered territory. Difficult to say that this is completely unfounded and everyone has, has lost all sense of reality. And then I think what I also enjoy is you going into the actual, you know, side by side of the dot com bubble to the current, you know, we also call them the AI stalwarts and how that these valuations are still very dissimilar from even the mania and the dot com bubble. I do think, you know, if I had one qualm with your video, which was you did leave out, I'd say the new rise of the AI companies. If you look at what's fueling all of these kind of, you know, what's fueling a lot of these companies, it's the clauds, it's the open AIs. I guess SpaceX to a degree is somehow an AI company now. So those companies I think on financial metrics are looking a little more dot com bubbly comparison valuation wise.
A
Yeah, and I think that's a good point. The reason why those weren't a big focus is because investors aren't invested in them unless you're, you know, a VC fund or something like that. And the kind of focus was very much on public financial markets and current investors kind of holdings. And so all of the existing stocks that could have exposure to that, all the semiconductor stocks, even you know, Google Meta, those kind of stocks. And so that was kind of really more, more the focus. Now we did talk about those companies as suppliers, capital suppliers and purchasers of a lot of different semiconductor chips for instances and power and their role in that. But we didn't talk about them from the valuation perspective, which you're right to bring up, but again we didn't do that because most investors are not exposed to those companies yet because they're private. So now if we do look at the actual valuations of them, and we don't know for sure what valuation OpenAI or Anthropic is going to come out at SpaceX that looks like is coming out around 1.75 trillion. That one's, you know, it is in part an AI company, but then it's also, you know, a space company as well. And you know, some of the estimates at the higher end are saying they're going to be doing about 30 billion in revenue in 2026. So you know, that valuation puts them a high 50s, you know, 60 times revenue. So a very high multiple. But it is also Elon Musk and it is kind of a little different than I think just talking about like the straight AI kind of bubble factors going on. And so if we're looking at, you know, what Anthropic raising a little bit below a trillion dollar valuation and yet we don't know for sure exactly what revenues are rumored, kind of around $30 billion ARR kind of at the higher end. So that's above OpenAI, which is probably closer to 20. And so if that's the case, that's like a 33 times multiple right there. And so that is going to be for anthropic and then OpenAI is probably a higher revenue multiple from there. And you know, are those valuations high? Yes. But at the other end of it is these companies are growing quicker than any company has ever grown before. You know, Anthropic has reached this milestone of ARR faster than any business has before. And that is kind of the thing that gives me a little bit of pause in saying outright. It's a ridiculous bubble valuation plus the fact that they are capacity constrained. So right now as it stands, it seems like there's more demand for their services than they can actually serve. Now, of course that's only at a given price and there's going to be, you know, a clearing price for their services and you know, it could be too expensive right now. And as a result of that, you know, demand is kind of falling, which we're seeing a little bit. Actually, we're seeing a lot of companies blowing through their AI budgets for the year and they're turning off. Claude. Microsoft very famously did that, you know, a month ago or a few weeks ago with their engineers. And so that's not necessarily saying though that it's ineffective, it's just saying that it's too expensive. Now you could kind of make the argument that they, if it was such a clear ROI on that that they would probably just keep, you know, Claude code on. I, I think it's kind of a mix where they're not exactly sure whether or not they're getting a high ROI on that investment. And then it's also the fact that, you know, maybe just getting people to do these things are still cheaper because if you're using AI and it's almost, you know, doubling whatever the cost of an engineer or whatever it is, you really need them to become a lot more productive. And maybe they're not seeing that. Maybe they're just saying the engineer does less and the engineer lets AI kind of do their work for them. And so there's going to be a lot of conditions. And right now we're kind of painting, you know, very broad strokes. But the crux of the claim here that you're trying to make is that the AI companies are in bubble like valuations. And I'm not as confident as you are in making those assertions. Certainly, you know, they're very expensive multiples, but I don't know whether or not that growth is going to offset that. And then plus the fact that they do seem to be capacity constrained.
B
Whoa, whoa, whoa, whoa. Listen, no one's making an assertion that they're bubble. Like I'm saying, if you were to draw comparisons of valuations, those companies seem to have more similar valuations to the darlings of the dot com bubble, which is as compared to Nvidia and Micron and some of the actual infrastructure companies that are more reasonably valued. And again, you could also make the argument depending on. And I think you did a great job that these companies. Okay, let's. Actually, I'm going to narrow it down. I'm narrowing myself down because I'm going to make a point that I think it's better that we just kind of cohesively go through this. Let me host here. I'm bringing us back on track. Are you ready for this, Drew? I'm done talking about the South.
A
I'm ready for this.
B
I brought up Isaac Newton. I read that in a book somewhere. And now people know that I read. So we can get focused back on that was really the whole point of this diatribe.
A
Do you hear that? That is our listeners being super impressed with everything you just said, especially that Isaac Newton point. They're like, how did he know that he must be AI?
B
He must be AI. And he must have Googled that before this started.
A
But I could see you getting a second positive comment by the end of this podcast.
B
Well, that would be good. That gets me another week on this thing. But okay, so let's dial it back in. You very articulately lay out two sides of the argument. I think the, you know, pros and cons of it. So why don't we talk about the. In the current market, ignoring the private companies, what would you attribute currently as the bubble? Like characteristics of this corner of the market.
A
So we touched a little bit about what we saw on the investor side. I mentioned that just, just the sheer kind of price movements of some of these stocks, like memory stocks going up 1000% in 18 months and being priced as if they're never going to be cyclical again. You know, AMD stocks similar to that, it's trading at some 20 times 20, 30 earnings, which is a high multiple for a company. If the company is going to be cyclical, if you're saying it's never going to be cyclical, then fine, maybe that multiple makes sense. But the whole idea that these deeply cyclical companies are done being cyclical and are done having earnings contractions, I don't
B
believe dive into that though, arguing a little bit more. This is a point you brought up a couple of times the cyclicality of these businesses warranting or multiples. Can you articulate why that's a given fact? And then secondly, couldn't the argument just be made? Well, the higher they're going to have higher highs and you know, higher lows essentially. Now with, with this new.
A
So the second part of your argument is actually going to suggest the multiple should be lower. Because if earnings are going to be higher, then you still want to put that multiple on like normalized earnings kind of an average between the low and the high. But since the highs are going to be higher, that multiple when you're running it is going to be even lower than it was in the past. And if you're looking at like what the lowest kind of multiple, like micron traded at, it was three times earnings. So a pretty low multiple. And the reason why it, I don't want to necessarily say it deserved that multiple, but it got such a low multiple was because earnings could fluctuate so much. So you could have one year where, you know, earnings were 90% less than what they were the year before, or they were loss generating and going from positive to loss generating. And so you put a multiple on kind of the average of earnings across the cycle. Because when you put a multiple on earnings, what you're really doing is a dcf. And so if you put a multiple on high earnings, kind of the implication there is that these earnings are going to continue to grow from there. But if they're cyclical, they're not growing, they're contracting in several periods through a cycle. And so that's why you want to put the multiple on mid year earnings because then that's like mathematically more equivalent to saying their earnings are going to grow kind of from here over the cycle. They won't grow in each individual year, but from like base to base they're growing, which was kind of the first point you're trying to allude to. And it's totally possible that AI does mean there's a lot more demand for these chips and so there's going to be higher highs and higher lows as well. But having said that, that shows up in the earnings, so it doesn't mean that you expand the multiple as well on top of that.
B
That's reasonable. Sorry. And one additional point here I'd like to make is that technically these companies were a lot more commoditized, would you say as well that's another reason they typically had low earnings multiples?
A
It depends exactly what company we're talking about. From memory. Yes.
B
And I don't know on the Micron memory. Right?
A
Yeah. And that hasn't really changed much. It's just that they were very supply constrained. So that's kind of the difference there. But then if you're talking about amd, like you could argue they've kind of become even more differentiated or more competitive, especially against Intel.
B
Right. So again, to assess your claim about the fact that cyclical companies typically have lower multiples. Multiples are a dcf, they're shorthand dcf. If over the lifetime of the company, these are the average earnings, that's how you should kind of assess the valuation of the company. Now the market is assuming that these companies are growing for a long period of time and then again at the end of your five year period, you're still now a very solid, competitively advantaged margin, high margin earning company at the end of that cycle. Right?
A
Yes. With the caveat that I would treat, you know, AMD differently than Micron because Micron's memory margins right now are, you know, the highest they've ever been and they're probably going to continue to go up for the interim, but I don't see that as being a long term sort of thing. Whereas AMD could, you know, structurally have a better business at the end of this, but still be cyclical.
B
And again, a little bit on the cyclicality, I think one point you made, which I want to highlight here, there's not a really good section highlighted, but people who've been following this have noticed the capacity constraint of the chip manufacturing, specifically tsmc, who has elected to not really blow their capex budget up, although it's been increasing. A lot of people said in 2022 when ChatGPT came out, the smart thing would have been to have laid out significantly more money in capex. And again their CEO and their company's perspective is that they've been through these cycles before where they have leaned into the over usage of capex. And because again they are a high utilization based business margin is going to really contingent on how utilized they are. They have elected to kind of have a more even hand in this, in this market which again people criticize them for. But I don't know, that says something. I mean someone who's been around the block for the last 40 years has kind of said, hey, we've seen this before, maybe this time's different, but TSMC doesn't think it is.
A
Yeah. And if you ask me what is different about this time, it is. That is one of the factors. You have both TSMC especially saying we're not going to expand capacity that much, but then also a lot of other players like the memory players are not leaning that much into expanding capacity. They are a little bit and they're retooling certain lines for high bandwidth memory. But they're not all saying, you know, we're going to build out supply to meet current demand. Instead they're saying, you know, we'll increase it a little bit but we want to be cautious. That's kind of the general take right now. Now when you have multiple companies though that you know, could take the potential demand, kind of the pressure to take on and earn those revenues from those custom eventually kind of weighs on them and that's what kind of pushes one of them to increase their supply and then the other one doesn't want to lose market share and so the other one does the same. That's kind of the dynamic of why that tends to happen in industry where you have multiple players. But it's different with TSMC because that's not really the case. Now we could say that maybe intel does get kind of their crap together and does become a more formidable competitor and maybe TSMC eventually becomes threatened by that. That's not going to be the case now, but that is something that could happen. So it's different when there's one player that's the bottleneck of a value chain versus if there's multiple players that could theoretically step in and fulfill that supply. Because then it's, you know, regular capitalism considerations that tend to kind of win out and they will eventually want to try to earn those profits which then results in them competing those profits away. And so the hope is that in memory they all kind of maintain this oligopolistic sort of supply structure right now.
B
Yeah, and again, this kind of gets into your. And right now we're all pro bubble, so we're just leaning into the very.
A
Well, that was an anti bubble argument or at least a slowing of the bubble argument because slowing it. If you think about what kind of the classic capital cycle is, is that you have an industry with high returns or is thought to have very high returns like AI and what ends up happening is there's a lot of demand, then a lot of money floods into that industry to try to build out the supply to meet that demand. Eventually capital overshoots. And so that means supply overshoots demand. And when supply overshoots demand, you'll have falling prices. Right, because now you have excess capacity. And so this would in analogy, it could be a couple different things, but it could be, you know, we overbuild way too many data centers. It could be that, you know, memory companies overbuild capacity, could be that TSMC overbuilds capacity. It could really be kind of any of these in the value chain that they all just kind of overbuild. And as a result demand is less than what the supply is there for. And that's when you tend to get price competition and you're getting price competition, you're not getting the returns you thought you would. So it could happen in many different areas of the value chain, but that's kind of worth emphasizing. And your original question though was about behavior, right? What makes us think this is the bubble? And so we just talked a little bit about, you know, originally the investor side, the stock price is moving up a lot on the business side, these companies are generally though being a little bit more capital disciplined with the capacity. If we're talking about the actual semiconductor players because they've been burned by past cycles. Now who is not being conservative though with their capex and capital outlay is mega tech. It's not Google, it's not meta, it's not Amazon, it's not Microsoft, it's not anthropic, it's not OpenAI. They are all being extremely aggressive with how much money they're spending. And so from the bubble behavior aspect, I would say it is all the capital they are putting into the industry, they're flooding it. You know, they all have the highest capex budgets they've ever had for like three years, four years in a row and they keep doubling them every year. It's just insane. And then Google just did an $80 billion primary stock offering because all of the cash flow they generate is still wasn't enough. And now Metta is thought to contemplate a similar thing. And this happened after my video. But one of my main kind of points in the video was that there is so much capital still flooding into this industry that I don't think we're anywhere near kind of the top of a bubble. I think we're kind of in like the middle of a cycle. And then there's other reasons I give for that, but I want to get too far ahead.
B
Well, first part I want to touch on is the capital market return cycle in that book. But it's such a basic concept. And again, this is economics and microeconomics 101, which is firms are earning outsized profits, money moves in, competition moves profits down. And now again, I think what the book expands on is that that naturally increases supply too much. And now there's almost a recession in that industry, and that's the bottom of the cycle. Unprofitable players move out and now it kind of stabilizes. Right? That's such a very basic thing to say. And then when you're in the moment, you just kind of look at that little curve and you go, where? Where are we on that thing? Like, I know you said, oh, we're maybe in the middle, maybe some people say we're at the top. But again, it's not to say that, you know, we're not trying to call bubbles or call cycles, but cycles. Everyone should be aware that cycles exist. And it's very difficult obviously to say where we are on them. But again, that is by nature how economics work. And eventually supply will meet demand and irrational players will leave. I mean, that's part of the market.
A
Of course, having said that, though, we know what the CapEx budgets are for 2026, and a lot of the hyperscalers said that they're increasing them next year and they're already kind of raising money, kind of anticipation of that. And we know that Anthropic is having an IPO and OpenAI is having an IPO and SpaceX is having an IPO, and they're all going to raise a bunch of money and they're all going to spend that money. So as long as there is all of this money still sloshing into the system, and we can't see that far ahead, maybe this goes on for many years, maybe it only goes on for one or two more years, but we can see there's still more money entering the system. There's More money being raised to enter the system. And there's a lot of commitments to spend money. That's why, you know, I had this thought that it's probably middle of the cycle now. Maybe it's earlier than I thought, though. Maybe this goes on for even more years. That's possible. It's also possible that, you know, AI could potentially hit some sort of a bottleneck, and we could talk about that later on what these bottlenecks could look like. And then people cut their spending because there's no point in spending on it either because the AI ROI isn't there or because there's just bottlenecks in the process, whether that's getting the data centers built, tsmc, whatever it is, that you can't spend as much money as you were hoping to spend. And then all of a sudden money is not actually able to flood into the system the way they expected it to. Not efficiently anyway. And so it's all just slower. And that could also, in a way, maybe, if it's not a very dramatic popping of a bubble, it is definitely slowing down. And if you own some semiconductor stocks that you thought would never be cyclical again, at multiples that suggested that, and then all of a sudden you had a year of revenue contraction and then decreased earnings, I promise you the stock's going to get hit pretty hard as a result of that. And so that is kind of painting the scenario a little bit.
B
Yeah, no, and you're right. I mean, the spend is not decreasing at any point. And I also would argue your one conjecture of like, oh, you know, maybe they can't spend all the money. I mean, to me it just seems like the way all these tech companies are talking about is very zero sum game, which is kind of, oh, well, if I don't buy it, Google's going to buy it, and then they're going to have the foundational model. And if I, you know, so it just seems like they're at each other's throats and nobody is going to say, hey, I don't think buying a Nvidia chip at this price makes sense. I just don't know if someone's going to pony up and say that.
A
Yeah, yeah, that's why that's probably not my base case. Now maybe you get to the point where it's like, hey, we can't put these data centers anywhere else because we just can't get enough energy to power them. And there's all of this pushback in all of these different, you know, local communities that don't want the data centers and for us to build it out, it's going to be extremely expensive and very hard to get the energy and the resources to do it. And, you know, it's five times more expensive than we are budgeting, whatever the number is, Right. And so, you know, either they come up with new ways to build out the data centers or that potentially slows. And that's why people are now talking about building them in space. Right. So there's obviously like a lot of different ways this could all develop. And kind of in, you know, David Deutsch's book, he talks about how whenever you face a problem, you have to find a solution and the solution creates a new problem too, usually. And so this is kind of par for the course that as they figure some things out, there's going to be more problems. And, you know, that's okay. That's okay. If all of a sudden we realize we need all these data centers and then it creates new problems that we have to solve, I think eventually we will. But just from a sheer timing perspective, I think that if we see the capital flowing into the industry doesn't continually lead to steady, to increasing revenues for a lot of these semiconductor businesses, these businesses kind of at the crux of the AI buildout, I think what you're going to see is a big valuation derating just because again, a lot of these companies are being priced as if they're going to continue to expand earnings. And I think that if you see one year, even a few quarters of contracting earnings, I don't think investors are going to say this is just a little blip and it's going to continue to increase thereafter. I think instead they're going to realize that, oh crap, these are cyclical companies.
B
Right. So if I were to. And I'm going to move on to our third point we've already summarized. First point, companies expensive, expensive companies embed a lot of assumptions. And assumptions are hard to make very far into the future with these companies that historically haven't been great. And again, that ties into, well, how are they great? Well, we have this capital market cycle, obviously demand higher than supply. Is that shielding and making these companies that historically are not that great look a lot better than they are, and is that a concern for valuation? And I think this ties into how does this all tie together? Which is, if there's a dramatic rerating of everything, how is that going to impact the broader ecosystem? And this is the third pillar, which I would say is the interconnected nature of all of these Companies and all of the dependencies they have on one another to where if the end user, and I know you alluded to kind of Microsoft, you know, the expense of kind of some of these, you know, utilizing these tokens has not generated the ROI that they needed to justify the continued spend. If the music kind of stops, how does this impact all the companies? And so maybe you can touch a little bit about the interconnected nature of everything.
A
Yeah, look at what happened in 2022 when a lot of these tech companies just their stock price moved down a lot. All of a sudden they were getting this sign from the market to stop spending so much money on growth and to become more profitable. And there's many companies that are great examples of this. You know, what comes to mind off the top of my head is, you know, Spotify, that's one of them. They became profitable, pushed more in profitability. Sea Limited, a really good example, they were losing a lot of money on a lot of growth initiatives. And when their stock price dropped 90%, they took that as a signal that they need to become profitable. And so a lot of examples of these businesses at this point in time where they're kind of taking direction from what valuation investors are placing on their stock price to then dictate their actual spending. And so this is a sort of feedback loop, or reflexivity, whatever you want to call it, between the stock price and management's actual investing decisions. And it does make some sense because, you know, in Sea Limited's case, a lot of their investment was stock funded. They were raising equity in order to go to new markets and grow. Right now that's the same case with, you know, OpenAI and Anthropic, they're going to go raise equity in order to continue to grow. Even Google, a company that, you know, IPO'd in what, 2004, they're still raising equity today in order to go after this investment opportunity, despite being very, very cash flow positive. And so when you see stock prices decrease, all of a sudden, equity raises become much, much less attractive. And you do kind of get this nod, if you will, from the market that your existing sort of strategy is not working. Now, you know, some managers can certainly try to just steamroll through that and say, this is still the right strategy. We're going to still continue to spend a lot. But if your business in and of itself is not profitable and you don't have cash flows to do that, you just won't be able to. So you won't be able to pull like a meta who Said, okay, we're going to still spend a lot, but what we're going to do is we're going to become a lot more efficient. So we're going to continue to fund reality labs to the tune of 20 some billion. And we're also going to put out these. At the time it was like a $32 billion capex plan and everyone was like, wow, that's so much money. In order to improve the signal and the face of att. Now that's, you know, a drop in the bucket relative to what they're doing. But the point there is that they had a real business. I was able to generate a lot of cash flow and sustain the investment. That's not the case with a lot of these AI companies. That's not the case with OpenAI and Anthropic, who are going to need to get the money somewhere. And if they can't do equity raises because valuations are down so much, where's the money going to come from? You know, they could issue some debt. That's usually not a great idea. And if your company's stock is down a lot and you're unprofitable, you can't really take on debt. And so then you're kind of forced to actually generate profits in order to fund future investment. And if you're forced to generate profits, then you also can't be that ambitious with how much you're growing. You can't have that much growth, expenses, that much growth capex, you have to cut all that back down in order to be cash flow positive. And then maybe once you're cash flow positive, the market believes you have a real business and then gives you a higher valuation, which then allows you to kind of get back into investment mode. But that is kind of the feedback loop that would happen if valuations all of a sudden fell back down. And so you might think, oh, well, you know, it's just the stock price moving because I think like Warren Buffett and the stock price is different than the business in reality. When you're actually talking about a business growing in their capex plans, the stock price dictates a lot of the movement and capital decisions of the business, of managers, decisions to allocate capital.
B
Yeah, I love that point. And by the way, I think Ben Thompson had a great article about SpaceX, kind of similarly to that, which is. And I think he's made this point, his point's been made before, which is Elon Musk uses his larger than life personality to drive high stock prices, high stock Prices allow him to issue equity at very high valuations. That equity cash infusions allows him to execute on some of the crazy ideas that were much less likely to be feasible if that whole feedback loop didn't happen. Right. And so, you know, similar here. Right now the market is letting people do a lot of things on belief. If that belief starts to get shaken. You know, Microsoft, a great example, a lot of great data about businesses, actual uptake of AI. I personally, when I get an AI agent on the phone, it makes me angrier than a recording because the AI agent is trying to do things that it can't and it makes me want to throw my phone against the wall. Is that just me or have you had better luck with AI agents answering the phone?
A
So what you're saying is this is bullish for Apple because people are going to have to buy new phones.
B
No, I'm talking about like AI agents. Like, you know, you try to call American Express and they're like, this is your AI agent for me.
A
I know, I made a joke there and you missed it.
B
Yeah, yeah, yeah, that joke went right over my head. Yeah, good point. You know, speaking of new phones, my 1 year old iPhone decided to start, like have its screen flickering for no reason. I haven't dropped it or anything, you know, so now I gotta go.
A
I mean, it's not for no reason, it's because your upgrade cycle has been getting ridiculous. And you know, they have a little thing in the phone that they have to flip. It malfunctions anytime you. They're like, we have a candidate for an upgrade here. He's got the money, he'll do it.
B
I literally told my significant other. I was like, hey, I'm gonna switch to a Google Pixel because I'm sick of the iPhone. And she's like, you're not allowed to do that because then we have shared albums. You can't. You're gonna be weird in group text. It's gonna ruin everything. You can't do it it. And you know, that's a lot.
A
So you guys broke up? Is that the story?
B
No, I kept. I'm getting a new iPhone even though I'm sick of them. But anyway, again, that's a whole nother discussion. But again, that was just my little anecdote about how AI customer service reps not, not floating my boat, but maybe five years or so they'll be good. I don't know, just not right now.
A
Well, I'm, I'm annoyed too on the chat boxes because they're, you know, they take a while to load, and then once they load, they usually don't have that many permissions on what they can actually do. And almost anytime I'm going to, like, talk to a chatbot, it's like, because I want a refund on something like that. Or like something's broken and someone needs to fix it and they can't do that. So it's just like an extra step to go through it before you get to the human and you got to do the whole thing over again.
B
And it's harder to get to a human because they keep trying to answer the question or they fired them all
A
because they have the AI bots.
B
They got to pay for the AI bots. So, yeah, now there's no customer service.
A
There's a lot of token cost involved in that anyway.
B
So, yeah, you're right. So that's the risk. Right, which is this doesn't materialize, or there's a shakiness in the timeline of that, which gives a feedback to the end users of these companies, which then flows through into all these other systems. And again, that would be kind of the bubble, like, everything pops scenario. Right? Which is everything that's interconnected kind of goes down in a way. And again, I don't say down bankrupt, I mean significantly rerated.
A
Yeah, And I think Klarna talked about that, where they had rolled out a lot of AI for support, and then they had to kind of curtail that a little bit because support satisfaction ratings went down a lot, which is kind of like the experiences we were talking about a little bit. And so that's not to say, like, because people always miss this, like, whenever you say, like, it's not good right now or something like that. It doesn't mean that I am negative on its ability to eventually get good, but I'm trying to make a point that the timing on when it gets good matters a lot for how much capital is flowing into this industry. And if it takes a while for these models to actually get a lot better, and I know you can make the argument with a lot more Capex data centers, the model improvement goes up a lot too. But you need to see a more immediate ROI from a business standpoint, because the way businesses operate is they do sort of exploratory budgets, they do experimental projects, and they want to see whether or not the ROI is there. And if it's not there currently, then they'll revisit it at another time. And that doesn't mean that it could be next quarter or year. It may be a couple years out. If they went through this big effort to replace all their staff with AI and it didn't go well, then they're going to kind of be negative on that for some period of time. Maybe wait until another company or two does it and sees whether or not it's actually working for them before they go all in on that. I think there kind of has been this mandate a little bit from the market for a lot of these companies to do everything with AI. AI, AI, AI. Do as much as you can with AI and a lot of companies went at it. And there's studies showing, there's an MIT study that was showing that it was like 90% plus of all of these projects had a negative ROI or an unclear roi. And that makes sense because like an average business, even if AI is powerful, it's like, how do you actually utilize that in your workflow in a way that is going to make you more money? It's one thing, you know, if we're looking at coding, because that does seem to be a uniquely good sort of value prop there, if you will, especially if it's alongside an engineer that can actually understand the code. And it just makes them a lot more effective. And there's, you know, something to using chatbots for, you know, support. It's not all bad, but there's a lot of other functions where they've tried to inject AI into the process and it just makes it more complicated because there's still errors. It just isn't working quite as well as it should. And as a result of that, it's just, you know, they're paying for tokens without really getting much cost savings or revenue uplift on the other end of that. So this is also kind of a bull case for a lot of software companies because their whole argument is that an average company doesn't really know how to implement AI. They don't know how to go to the Claude API or whatever and then build a product around it that's going to actually streamline a business process such that it results in an outcome that can improve efficiency. And so it's the software companies that are going to be utilizing AI to help help the companies actually use AI. And so if you were, you know, a prior user of ServiceNow, now all of a sudden, ServiceNow incorporates AI into the service ticketing sort of chain for IT management. And they're the ones that are able to use AI a lot better because they have, you know, all sorts of engineers and data scientists that can actually incorporate AI. Into the software product and that's an actual product you can sell versus the idea of just, you know, selling quote unquote AI. And so maybe it works too for anthropic when you can get a forward deployed engineer, you know, into Goldman Sachs and have them custom build an application for that. But you know, there's still no solution for like a middle market sort of company. So that's where you know, a lot of software companies could step in and kind of deliver that. They could be the ones to do that.
B
And you know, one, one last argument I'll make which is we're talking about customer service jobs and call center jobs and again these have really been offshore and the cost of the jobs as a percentage of U.S. jobs or some of these things as a nature already quite low. So you have to kind of overcome what's there and then also overcome the cost advantage as well. So again, I'm not shocked by the studies. I mean some of the data was interesting but I don't know. I've yet to see a super impressive consumer facing AI experience at this point other than Claude. I mean again I'm saying as a customer of a business, I personally use Claude and chatgpt my daily basis and I do think it makes me very efficient for monetarily for the 20 bucks a month I pay for it, it's a no brainer. Again the enterprise level suite and like firing a whole headcount to replace it with somebody. Yeah, I haven't really seen that yet. But hey, we're early on, we're early on. So let's get into the pros of the AI. I think that ends our cons. Do you have anything? I don't want to say cons. We're going to say this is pro bubble right now. These are pro bubble.
A
What do pro bubble, what does pro bubble mean?
B
Pro bubble means these are arguments that there is a bubble right now or pro the AI bubble right now. Is there any other closing arguments you have for the bubble exists? I think we talked about valuations, we talked about cyclicality.
A
Yeah. So it's valuations, it's the investor behavior we talked about, it's the business behavior, it's all of the capital flooding into the market. It's the fact that you know, you have have Google CEO Sundar Pichai talking about how he explicitly thinks there's too much of a risk of undershooting spend so he'd rather overshoot it. That that is clear, you know, bubble sentimentality and that eventually results in you actually Overshooting it. Right, because that's what they're kind of set on doing. What we're seeing between all of these different tech companies increasing their capex budget. So that's all suggesting that this could at least turn into a bubble. You know, you have weird investor behavior like Albert's deciding, you know, it's an AI company and the stock shoots up, you know, 500%. You know, we talked about some of the other businesses. You do kind of have a lot of circular financing arrangements, arrangements where Nvidia will take an investment stake in a business and they'll take that money and go around and indirectly buy more Nvidia chips. Now, that's not, you know, most of the economic activity, but that's certainly something that happens a lot or happened a lot in the last tech bubble. So the other one is, you know, debt financing is starting to increase a little bit too. You know, you have Oracle has over 150 billion in debt. They just did another 20 to 30 billion. Even Google is increasing their debt load now. They have 50 billion. They just did an equity offering though Meta added, you know, 30 billion. And so those are all kind of, you know, bubble signs when you start increasing the debt load. And it's kind of interesting they're also increasing equity at this point, which could still have, you know, pretty mixed ROI. And then adoption is kind of the big one. And so we talked about some of the things, whether or not adoption is actually there and getting there that quickly. You know, there's a few different studies. It's hard to know exactly just from the studies, but, you know, I listen to a lot of, you know, transcripts of what companies are saying and all of that, and more importantly, how it's showing up in their P. Ls. And, you know, we mentioned the Microsoft thing on them canceling the Claude code licenses. There's another MIT professor, Darren Acemoglo, that found that that only 20% of work can theoretically be replaced by AI but only 5% of overall work can be replaced by AI profitably. And so it's just too expensive for right now, especially if you are offshoring some of these tasks like customer support. And it's relatively cheap. So that's kind of the argument for why we are potentially, you know, in a bubble. A lot of this spend is just not currently warranted by the technology that exists today.
B
I think that was a good summary of the pro bubble argument.
A
Thank you.
B
I mean, I gave it.
A
So of course, I would say I'm
B
a little concerned about the bubble now. So. Okay, but let's convince me now that there is no bubble.
A
And I think you're easy to convince. I'll convince you of anything. I convinced you to host, I convinced you to host a podcast. I'm good at this.
B
That's true for like three years now. But let's get into the anti bubble argument or the fact that hey, this thing has legs, it's reasonable.
A
It's not pro bubble and anti bubble. That's like an opinion on whether or not we like, like bubbles.
B
I mean that's how I'm spinning it. So let's get.
A
How would you an argument for if we're in a bubble and like a counter argument. We're in a bubble. That's one way to say it.
B
Counter argument that we're in the bubble. Okay, good.
A
So pro, anti whatever.
B
Pro anti whatever we're gonna do. That's how my mind's thinking about it. So let's get into the meat of what's happening here. And I do wanna start with the straight up comparisons between here and the dot com bubble. And we did touch a little bit about, you know, valuations and things of that nature. But, but from just a high level, looking at the S&P 500 forward earnings, looking at the NASDAQ forward earnings, looking at the Nvidia AI stalwart valuation, these are actually, again, I know we were discussing the cyclicality and how they're high. They are high, there's no doubt about it. But relative to the original mania that this gets compared to, we are in a very, we'll say, reasonable period compared to that.
A
Yeah, I mean, I think the biggest things to just look at to say this isn't a bubble is valuations. If you're looking at like the Ford NASDAQ multiple, it's like 24 times versus like 60 something times at peak during the tech bubble. Even S&P 500 traded at peak tech bubble like 28 times forward, something like that. It's right now 21 times. That's high versus historical of 17, but not near 28. And then, you know, you could look at usage, usage in the Internet in the early 2000s. You know, it was a few hundred million users. Now we have a couple billion users that are using AI products. And just maybe anecdotally, I don't have actual data for this, but it seems like people's, you know, time spent on AI products is a lot relative to maybe their time spent on very early Internet when there wasn't a lot to do. There a lot of sites there and it was pretty slow. You know, I have everyone in my life that I know uses AI in some capacity and quite regularly, even, you know, going up to, you know, older, older ages as well. And so it's just the usage is a lot higher than what the Internet was. And the revenue is real too. There's real people, you know, paying revenues to access AI, certainly on the consumer side and also a lot on the business side as well. Even though we talk about, you know, whether or not some companies think that these tokens are currently worth their AI usage and all that. And then again, you know, Anthropic has some, you know, $30 billion or by some marks closer to $50 billion in ARR and is again one of the fastest growing companies of all time. And that's, that's real revenue. Same with OpenAI also very fast growth, although Anthropic's been even faster. You have the supply bottlenecks we were hitting on. So that's kind of constraining the bubble because people's ability to spend is going to be slowed. TSMC is one of those kind of constraints in the supply chain, but there's a lot of other ones. There's energy is a big sort of bottleneck. The memory companies are also so bottlenecks to an extent. And then there's, you know, the fact that this is still mostly cash funded, which could change. But generally speaking, when we're, you know, getting into a real bubble, the issue is a lot of it becomes debt funded. It's almost hard to find enough funding for all of these projects and so they move more into debt. And then the last one that I'll note is that a lot of the spend is being done by, you know, pretty high quality customers with real businesses that are cash flow positive. And so there's less of a risk that they go bankrupt and you know, renege on all of their contracts, which is something that happened during the tech bubble. You had a lot of companies that were promising to pay money they could just never theoretically actually do. So maybe OpenAI, maybe Anthropic a little bit is an exception to that. But you know, the capex budgets Google's laying out, it's real. I believe they're going to spend every dollar they say they're going to spend. They have an actual ability to do that. Same with Amazon, Apple, Microsoft, Meta. And so that's real money that's going to be spent and pushed into that. Which is kind of the biggest argument that, you know, this is not going to end because all of this money is still flooding in to the ecosystem. And so, Alex, did you flip yet?
B
I know I'm a little wishy washy here, but I'm now going to argue the other side. I just want everyone to know that I'm versatile in my arguments, but I
A
think overall, impressionable is the word, impressionable.
B
But I think that's what makes these things complicated, is that there's compelling arguments on both sides here. But, but overall what we see is, yes, there's a struggle and uptick in, in the business community, yet nobody's giving up on this. Right? Everyone sees the exponential increases that are made each quarter, every six months, a new model, a new solution. And I think everyone has the faith that this will get to where when I call American Express and they say, hey, this is an AI agent, I'm going to speak to you and you're going to be able to solve my problems in their totality. And that is a future we're heading to. And it's better to start integrating these systems now than later. And I think that that's overall the business community's perspective and the spend and use cases will continue to increase, continue to increase and justify the spend. If I were to have a kind of a base case situation of what's going to happen here, which is, I think the AI bulls are correct. This is a transformational technology. I think that there's going to be a lot of money to be made. Do markets get ahead of themselves sometimes? Will there be a rerating? I think that's a probable event as well. Do I think it's going to be a bubble that pops and crashes and we're sitting there like, oh, look at all these morons. You didn't see the second dot com bubble? I don't see that being my base case scenario. So here we go. Let's see how well this ages. But I think it's going to be a nice middle ground. You know, they always say no one's left in the middle. I'm in the middle here. That's where we're at.
A
Well, so I, I called the, the title my YouTube video. We're not in an AI bubble yet.
B
Yet.
A
I think that just human behavior is such that we kind of feels like we're going to eventually get into one. And to describe what I mean by a bubble, I pulled up because I was, I was looking this up. I'm like, what is, you know, a bubble anyway? And I really like poly markets definition because it was the Most concrete. And So a major AI stock falls 50% from highs and you need to meet like three of these five conditions. So, so it's like a major AI stock falls like 50% from highs and it lists a bunch of these major AI stocks, including Nvidia or OpenAI or Anthropic Go bankrupt. That was one of the conditions. I don't think that's going to happen. But if one of them did, It'd probably be OpenAI or Nvidia's H100. Rental rates fall from seven to a dollar. So basically the cost of renting out their chips falls a lot. Kind of suggests there's a Supply Gillette and, and then there's a few others that kind of all are around like just stock valuations dropping like 30%, 40%, stuff like that. And so, you know, if we're looking at just the current valuation levels, I said 24 times forward NASDAQ earnings, okay, maybe that goes up a couple multiple turns once we get OpenAI and Anthropic and SpaceX into the index. But it's still, again, it doesn't quite have room to drop, you know, 50%. It's not going to trade at a 12 times earnings multiple, something like that, that or you know, 18 with those in the index, whatever those numbers come out to be. So I think that for now what's going to happen is all of this, you know, actual spending into AI is going to continue. All of these companies that are beneficiaries are probably going to continue to see record earnings and revenues. I think eventually valuations are going to get ahead of themselves. And you could argue in some areas they already are. But in terms of how much something can get ahead of itself, it is so much more. It is so much more. And so, I don't know, it's a little weird to like predict actual price movements on a stock. So I don't want to do that. But if you're looking at what exists today, there are certain behaviors that feel bubble, like, but you know, this AI capital cycle as a whole, it doesn't feel like it's anywhere near an end. I said it somewhere in the middle, but I don't really know. And as we kind of get nearer to, you know, the end of it, that's usually where you see most of the behavior get out of hand.
B
Right. And you know what I see, which I liked your action items at the end of the video as, okay, if you're afraid of, you know, this potentially being a bubble, if you're afraid that or you're missing out. What are your action items here? I think that what is compelling to me here and Also in the 2021, I don't want to call that maybe the software bubble. I don't know that was a bubble necessarily, but yeah, you had a lot of frothy behavior. Yeah. Or I don't know. What's a nascent bubble? I'm not sure what that is. A baby bubble, we'll call it.
A
But that's not nascent. But anyway, go on.
B
Nascent is early. I think a baby is early. Is that not the right use?
A
Why is it an early bubble? What does that mean?
B
You're right. I don't know. Just like whatever. Not a big one. But there's. We'll call it. Like I said, we're using.
A
I see nascent and like growing. I see what you're saying.
B
It was a froth. It was some froth going on, but
A
it was a budding market.
B
Most parts of the market were very, just reasonably, I'd say, traditionally valued. And I feel like same thing here, which is, yes, you have these companies that are caught up in the AI hype and then maybe even some of the software companies that have been almost negatively associated with AI to a large extent. And perhaps there's opportunities there, maybe there's not. That's for people to decide. But I think there's a lot of parts of the markets that are just reasonable. I mean, it's not, you know, just kind of historically where we are there and the whole thing isn't caught up like we're in, you know, some late 20s period or nifty 50 period. That's not where we are. So I think there's a lot of pockets to operate in outside of this arena.
A
Yeah, no, I agree with that general take. I want to emphasize, though, that doesn't mean that, you know, AI stocks in particular are going to be great returning stocks from here, you know, forgetting whatever the price movements do in the short term, you could still have, you know, entry prices that don't make sense for where the businesses are in 10 years from now. And that won't really be apparent for many years later, even if the stocks do go up a lot more in the interim. But, you know, to your point, yeah, there is a lot of other areas in the market that are not, you know, as impacted by this. Now, the problem with that, though, is that when there is such a big kind of theme in the market, it does tend to soak up a lot of the capital of investors and so that could go on for a while too. And so you might want to invest in other sectors of the a market where they're not as attached to this AI narrative. But then you're going to have to be patient because a lot of money is not flooding into that area. I, I can promise you a lot of these investors at big funds that are watching how much the semiconductor stocks are moving, they feel a need to invest in some of these stocks and anything that has no AI story, they're just not going to be as interested in them. And that could last for some period of time too, you know, however long Warren Buffett's, you know, period of underperformance was. It was, you know, several years, literally several years during the tech bubble where people would say, you know, he lost his way. Why isn't he buying these tech stocks, yada yada. And you know, maybe now in hindsight we could, you know, say, yeah, you should have maybe figured out some of those tech stocks, but that's not the point. The point was that he was right to avoid a lot of that activity, but it just seemed very foolish because of how much the price of those stocks have, have moved up, up in the short run. And so I kind of expect something like that to happen again. I mean, to some extent that's already happening. When you talk about how the market's having its greatest dispersion by sector, some sectors that are down a lot, and then you have semiconductors that are up a ton and maybe energy too, and some consumer discretionary that's kind of making up for everything else in the market that's down. And so that is what's happening is that these AI stocks are attracting a disproportionate, you know, amount of capital and all of that. And we'll see kind of how that all goes. Now if you're concerned about all that, you can just avoid those areas. You can invest in an equal weight index. Instead of the S&P 500, you could do RSP. There's a lot of different things you could do to kind of limit your exposure to that. But there's going to be a lot of people that want to maximize their exposure to it.
B
Yeah, I know. I mean, it's a good point. And also, yeah, tough to be a hedge fund manager, tough to be a mutual fund manager, tough to be anyone tied to the S and P or nasdaq. And you know, you don't own AI, you don't own memory trips chips. This year. It's been a tough year. I mean, it's that simple. You might not be in a job by the end of the year. Right?
A
So, yeah, and we're already seeing people like talk about like throwing out old strategies and stuff like that. You know, quality companies are dead and blah, blah, blah. And you know, all these old strategies aren't going to work. You need to, you know, get some sort of AI strategy, something like that. And so, you know, this, this is something markets do. They go through cycles and certain stocks become more popular. And as stocks become more popular, some investors, they'll do style drift to better match what kind of is working in the market. And part of that is careerism. Right? Protecting their career is a big aspect of this. And I'm not going to say it's all stupid because if this goes on for three more years, you know you're going to lose your job if for three years you're saying like, I don't want to participate in something. And so that's why people do. Now, having said that, because, like, you know, you could listen to someone like Gavin Baker who's, who's talked a bit and, and he seems pretty smart on AI and all of that. And he'll talk about how he thinks a lot of this is real and he thinks that this time is actually different. He thinks that even he owns a lot of the memory companies. He thinks that this is not going to be the typical sort of cycle. He thinks that it continues maybe to make sense to own them, at least as of the last podcast I heard him on. And so there are investors making those kind of bets. To me, that seems to not be very conservative. And maybe this is just the way I invest is, is I prefer to be conservative than not. I think there's enough ways to make money in the market being more conservative than having to assume that, you know, past cycles don't repeat themselves or something is different about this one. But that doesn't mean, you know, he is wrong or something like that to believe that, you know, he has his own reasons to, to make his own argument and, you know, it could be well reasoned and he could very well be right. There's just, you know, lots of different ways you can make money and you don't have to run your own personal portfolio like a hedge fund does, you know, earn alpha and beat the market. I think for the vast majority of investors, conservatism makes more sense because your goal is different than his goal. If you run a hedge fund, your goal very often is to beat the market. If you run Your own personal portfolio. Your goal is not to beat the market as much as you might think it is and as much as you might want to play that game. Your goal is to actually achieve your financial objectives. It's to not be out on the street. It's to, you know, buy that home. When you want to buy a home. It's to, to meet whatever kind of the objectives are in your life. And so don't get caught up in thinking that you need to kind of hit your wagon on every hot stock trade. It may not actually matter for your life and there may be more risk involved in it than is necessary to take.
B
Well said, Drew. The divergence of goals can be something quite hidden or some people. Yeah, you're right. You don't even know what your goal is. You know, you want to beat the market every day. That might not be what you need. So I think this was a very compelling two sided discussion and I think time will tell who's right. Are we at the top of the capital market cycle? Are we where Drew thinks, which is kind of on the way up, or are we just, I think where I am, I think I'm a little closer to the top than you are. If I were to mark my little X on, on the capital market returns. But again, time will tell.
A
So Alex is calling a top. That's what we should tell people. People.
B
I'm saying, I'm saying we're, you know, I wish I could demonstrate in a diagram where we are, but what a
A
curved line going up like this and then you're like, we are here.
B
Yeah, exactly. Yeah, exactly. That's where I think we are. I know one thing for sure is I'm definitely wrong because this is not, not an easy thing to predict. And I, that I'd bet on.
A
But yeah, and you, you shouldn't have to predict it in order to invest well, in my opinion. I think if you're in the game of trying to make these predictions, and again, you know, some HED funds are going to try to do that and that's fine, and some people might want to play that game too, but you shouldn't have to have that much of an opinion on an individual stock in order to do well in it. I think you should invest in businesses that are going to do well in a lot of different circumstances. You shouldn't need to assume, you know, okay, I need, you know, this company to not be cyclical and earnings to grow positively faster than they ever have before for the next 15 years and margins to expand more than they ever have before. And that's what I need to happen in order to get a good return on the stock. I just think that the idea of margin of safety and conservatism should still be employed. And that doesn't mean you never invest in any new technology or anything like that. It's just that there's certain pockets where you can get into these technologies and stocks where you don't have to make such lofty assumptions. And that was only less than a year ago. A year ago, ASML was trading at a fraction of the multiple it is today. Same with tsmc. And, you know, there were a lot of value investors that owned a lot of the memory companies. You know, Lilou and Mohnish Prag both owned Micron, I guess now infamously, because they sold out. But there were times to. To own a lot of these stocks where the assumptions you would have needed to make were much, much more conservative. And so that is kind of my word of caution is just looking at the assumptions you need to make and whether or not you really feel comfortable with those. And, you know, it's one thing when the stock price is up and they seem to be going up and up every day, but is this a company you'd actually be comfortable holding for the next 10 years? I think that that's something worth kind of considering.
B
No, I think it's a great point, Drew. And overall, this has been quite a compelling discussion. Some people say that I'm a little wishy washy, easily convinced by Drew, and I'm going to deny that narrative. And we'll see if I'm back next week, depending on how I'm feeling.
A
Would you say you're wishy washy if I made an argument for you to be wishy washy?
B
I think I would like you to make an argument against my wishy washiness. And I will agree with you there. So let's see what we'll do next week. That's the full episode. You're going to argue about how steadfast I am in my opinions.
A
You need someone else to make that argument.
B
Yeah, I'd like you to make that argument. So I could. So I could agree with you. But I think we'll leave it there. And I don't know what we're talking about next week. We'll figure it out. But that's the reveal for next time.
A
Until next week, Sam,
Date: June 10, 2026
In this intellectually rich episode, Drew Cohen and his co-host vigorously dissect whether the current artificial intelligence (AI) market is experiencing a bubble, or whether comparisons to past bubbles—most notably the dot-com era—are misplaced. Drawing on hundreds of hours of research and a recent in-depth YouTube analysis by Drew, the hosts break down the nuanced dynamics of AI-related valuations, capital spending, behavioral signs of market froth, and the complex feedback loops between company performance, market expectations, and real-world business results.
Their conversation is a tour de force for sophisticated investors: rather than merely declaring AI a bubble or denying it, the hosts engage in a balanced, point-by-point debate that leaves listeners with a spectrum of arguments and plenty of actionable insight.
“Anytime there is a new and innovative technology, it has always, always led eventually to a bubble.” – Drew, [02:08]
Investor Sentiment: The hosts point to surging valuations in cyclical companies (like those in memory/semiconductors) being treated as if they’re now secular growth stories.
[11:09] Drew:
“The whole idea that these deeply cyclical companies are done being cyclical and are done having earnings contractions, I don’t believe.”
Price Action: Memory companies up 1,000% in 18 months, stocks moving 30% on single earnings, and assumption that cyclicality has ended.
Semiconductor Companies: Unlike previous cycles, companies like TSMC and memory suppliers are cautious about overexpanding capacity, recalling previous cycle boom-bust pitfalls.
[16:01] Drew:
“If you ask me what is different about this time … [TSMC is] saying: we’re not going to expand capacity that much... they want to be cautious... that’s kind of the general take right now.”
Mega Tech: In contrast, hyperscalers (Google, Meta, Amazon, Microsoft, Anthropic, OpenAI) are flooding the space with record capex, sometimes even issuing new equity or debt to fund capital outlays.
“There’s so much capital still flooding into this industry that I don’t think we’re anywhere near kind of the top of a bubble. I think we’re kind of in like the middle of a cycle.”
[26:20] Drew:
“...the stock price dictates a lot of the movement and capital decisions of the business, of managers’ decisions to allocate capital.”
Drop in valuations directly influences how companies spend: e.g., in 2022, tech companies shifted abruptly from pursuing growth to pursuing profitability as stock prices collapsed.
“I’ve yet to see a super impressive consumer-facing AI experience … I do think it makes me very efficient … Again, enterprise level suite, firing a headcount to replace it with [AI]? Yeah, I haven’t really seen that yet.”
Stocks in “AI narrative” areas carry high valuations, often with underlying assumptions about endless secular growth.
Surging capital expenditure (“blitzscaling”), especially among hyperscale tech companies.
Notable circular financing—e.g., Nvidia investing in startups that spend heavily on Nvidia chips—a feedback loop reminiscent of late-stage dot-com dynamics.
Early signs of increased debt financing.
Questionable adoption and unclear ROI in many business contexts, suggesting capital is out ahead of practical value.
[38:12] Drew:
“Google CEO Sundar Pichai talking about how he explicitly thinks there’s too much of a risk of undershooting spend so he’d rather overshoot it. That is clear, you know, bubble sentimentality…”
Compared to dot-com valuations, current Nasdaq and S&P500 multiples are not extreme: Nasdaq around 24x forward earnings vs. 60x+ in 2000.
Technology is being widely used—billions engaged with AI, unlike the fledgling internet ca. 2000.
Actual revenues: AI companies like OpenAI and Anthropic are real, rapidly growing businesses, not mere concepts.
Much funding comes from cash flow–rich, stable tech giants (Google, Microsoft, Meta) rather than unprofitable startups.
Supply bottlenecks in critical infrastructure act as a natural brake on runaway exuberance.
Most spending is currently cash-funded, not (yet) predominantly debt-funded—a key difference from historic bubbles.
[42:06] Drew:
“If you’re looking at the Ford NASDAQ multiple, it’s like 24 times versus like 60 something times at peak during the tech bubble…usage is a lot higher…”
“When you’re actually talking about a business growing in their capex plans, the stock price dictates a lot of the movement and capital decisions of the business.” – Drew, [29:53]
“I’ve yet to see a super impressive consumer-facing AI experience at this point... Firing a whole headcount to replace it? Yeah, I haven’t really seen that yet. But hey, we’re early on.” – Alex
“If you’re in the game of trying to make these predictions…some HED funds are going to try to do that and that’s fine…but you shouldn’t have to have that much of an opinion on an individual stock in order to do well in it. You should invest in businesses that are going to do well in a lot of different circumstances.”
— Drew, [56:48]
Next week’s topic: Undecided. Will Alex still be on? Tune in to find out!