Transcript: We Asked A 30 Billion Manager Where Ai Profits Will Actually Go

TITLE: We Asked a $30 Billion Manager Where AI Profits Will Actually Go CHANNEL: Excess Returns URL: https://youtu.be/Jm2aArWxvRw?si=Un3Cf6byn8vMkoc1

I think what investors tend to have trouble with is conflating the idea of here’s a big secular trend that’s going to grow and it’s going to change our lives with exactly which companies that are riding that wave have the differentiated business models. The revenues are only as durable as the spend from the person above them who is buying their products. And as you do get further down the this the layers, you do lose visibility in what’s going on above you. Nvidia’s revenues are OpenAI’s capex and Open AI has the capex to spend because they’re getting money from Microsoft and also from just because they’re spending in advance of that revenue. Bad times can happen to anyone. Things happen in the world and a lot of being quality is just being able to keep going through those tough patches. Tom, welcome to Excess Returns.

Hi, thanks for having me. You’re head of focused equity and a quality portfolio manager at GMO. You spent most of your career there. Um, and your focus is on building concentrated portfolios that consist of highquality businesses. So, a lot of the conversation today is going to be around this topic of quality investing, what it means, how you define it at GMO. But I thought where we’d start with you is on a topic that is kind of critical to today’s market. It’s represented in a lot of investors portfolios and even in market cap weighted indices and that’s how to think about AI as a investment opportunity. You recently published a paper titled hype versus high conviction. You could people can download this on GMO’s site uh arguing that you know AI may be one of the most important decisions facing equity investors today. So we want to kind of work through that paper with you which was excellent and then I think at the end we’ll you know really try to tie back to how you define quality what makes for a durable business in your view and how you construct portfolios that consist of those types of companies. Um so great so to start let’s talk about the paper. So you know AI is on top of a lot of investors minds. We see it all around us. many of us are using it, trying to figure out how we’re going to benefit from it, and really trying to figure out the investment implications of AI today and over the long term. So, what do you think in your opinion? What what are investors maybe getting wrong about the narrative of AI? Like, what’s your general sense here? Well, I think part part of it they’re getting right. I mean, clearly it’s a great opportunity and the fact that investors think that is displayed in the fact that AI related companies are such large market cap and such a big part of the invest market indices. And so investors clearly appreciate the significance of it, have not just a strong position now, but a strong position far into the future. And that’s that’s hard, right? in a growth area things change. So today’s leadership isn’t always tomorrow or next year’s leadership. So I think that’s sometimes what investors can struggle with is sort of playing things forward is as best one can realistically and investing for the future not just for the current state of play. Right. So it’s not you can’t paint all these companies with the same broad brush and think you’re going to get the same great investment outcomes. So to your point, there’s there’s opportunities, but there’s, you know, possible risks. Okay. So with that, and I think this this is what we’re going to put this image in the um in the presentation in the interview here, but you kind of broke the AI ecosystem down into really four layers. And you have applications, LLMs, hyperscalers, and suppliers. And what you were really doing is kind of looking at the value chain in these four different uh core buckets of the AI ecosystem and then looking at what both the risks and opportunities or I guess the risks and advantages of each of those areas of the you know the overall market. So just kind of walk us through maybe even each one of these, you know, at as much detail as you want because I think this is this is a very critical part of the article, right? Um yeah, so as you say, we divide the AI ecosystem into these four layers of kinds of businesses and some companies do span more than one of these layers, but we think these four key layers are important way to think of the opportunity. So at the top is applications. This is how you the whole world I guess interacts with AI. The an obvious example of an application is chatgpt or copilot or you could say you know autonomous driving car is an application or a software generating system like cursor or con. Those are all applications. uh that is if AI is a big commercial success, it’ll be monetized by people paying for applications that do things. If you think back to the the previous sort of maybe wave of innovation around smartphones, applications do things like Uber etc. Um the the next layer that we think of beneath that is uh the compute that runs those applications. So that is uh the hyperscalers largely. So that is you your your co-pilot sits within Azure and runs on Azure. Microsoft is providing cloud services in that case actually for themselves but also for um all the people who would build applications on top of AI. Not to foreshadow where we’re going. Obviously the what what’s interesting about AI is there aren’t so many applications now but for it to succeed in the long term and deliver on its investment promise there are going to be a lot more of those applications in in the future and they’ll probably be running on the the hyperscalers we see today um beneath that compute what the compute level is um pro sitting on top of is the LLMs themselves so chat GPT is an application GPT um is DLLM or Gemini is the LLM or Anthropic etc. There are relatively few of these. They’re often at least currently housed in private companies or Gemini obviously within Google. Um those are um those are kind of where the crux of the innovation is. the fact that these things actually work as well as they do is the amazing development of the last uh three or four years. Um but that is um the third level of this and then the the fourth level is sort of the infrastructure enabling it and that’s um where Nvidia sits. Nvidia is the kind of poster child maybe for AI over the last few years but it’s actually fairly far down the AI stack. The co-pilot sits in Azure runs GPT which is trained or it does compute and inference using um Nvidia’s GPUs and then that actually we’re we’re bumping bundling this all together as infrastructure but of course Nvidia depends on TSMC to make the chips and TSMC depends on various companies to you build the tools to make the chips. There’s in addition to the semiconductor ecosystem, there is this sort of parallel ecosystem around data structure power. It’s not as big in market cap but clearly an important aspect of it too. So that infrastructure layer sort of the ultimate picks and shovels I guess of the AI revolution is kind of the bottom level of that four level stack we’re uh describing. Yeah, I think this is a great way when you’re thinking about these AI companies to first ask what layer are they in and then if you just want to touch on these like you know what the what the risk or I guess the advantage is within each layer like I’m just thinking like in the first one applications you know you have chat GPT going after more individuals where anthropic is going after more B2B uh but you know so those are kind of different markets obviously but then you know and there’s a a lot of variation in there. And then so you got to pick the market, pick the winners, but then you also have the legacy players to the point I’m just kind of sorry to step on you here, but like they have, you know, legacy players have like the data and they have like the customer like lock in maybe more. So just talk to talk to these uh risks and advantages within the value chain. Yeah. And I think more fundamentally at the application level, if we’re honest with ourselves, I think we have to say we just don’t know what all the applications are going to be. Uh, I used the example of Uber before because I think that’s an illustrative one of it was easy to see that smartphones were awesome before it was easy to see that Uber would be a killer use case for smartphones. It really was in couldn’t exist without them, but became incredibly valuable with them. And I think that’s for us the caution of investing at the application level is it’s sort of hard to get those right until they’re proven. And I think it’s a little bit premature to be too specific in what you bet on. I think there there are some applications to be clear that are are proven now. I think probably from a commercial point of view, code generation and chat bots um are kind of the two things that you are clearly monetizing today, right? I’m paying as a subscription to get Gemini. We’re paying at at work, we’re paying subscriptions to Anthropic to to get access to Claude. Th those are clearly working. But there’s a lot that isn’t u determined yet. It’s not even clear yet which kind of companies are going to win. Is it going to be startups like an Uber? Is it going to be existing companies, you know, like um Anthropics trying to move up the chain, right? GB uh OpenAI is moving up the chain by providing their own applications. Will it be legacy software players like a soft um a Salesforce with agent force? And we to your point believe those companies like have a lot of advantage because they do have the data. They do have the embedding with their customers. They can layer AI AI onto what’s already there. But one has to realize that sometimes in these big technology inflections, it is new companies doing things we haven’t thought of yet that often do um get very big market position. One of the points of the paper was you know following the cash. So that’s trying to look through you know how these AI investment dollars are actually flowing from layer to layer that will give investors you know a better way to assess maybe the risks of any one individual company. So can you just talk us through that idea? Yeah, I mean that’s one of the things that led us to think about this layer approach is that basically at from the top of the layer a company gets revenues and then its investments or a capex or expenses are the revenue of the layer beneath it. So I’m an application. I get revenue from an enduser. I pay a fee to get my compute. Um and then the compute um has to invest in you know Microsoft they has to invest in capex that will go well they have to pay um the LLM they also have to invest in either directly or through the LLM like OpenAI is spending a lot of capex that’s going to Nvidia. So Nvidia’s revenues are OpenAI’s capex and OpenAI has the capex to spend because they’re getting money from Microsoft and also from just because they’re spending in advance of that revenue. They’re getting a lot of money from outside investors of course but in equilibrium this is going to be funded all flowing down from the applications not from just external investors who are bootstrapping the system today. And then we say Nvidia’s spending money that goes to TSMC that goes to applied materials etc all the way down. Um and I think that’s an important way to think about it because for any company however much they’re earning now their revenues are only as durable as the spend from the person above them who is buying their products. And as you do get further down the this the layers you do lose visibility in what’s going on above you. So it’s a it’s a harder way to manage a business perhaps sometimes. You touched on something that’s really unique here of this boom relative to say the dot boom which is this idea that a lot of this is being funded from cash. I mean do you think that makes this inherently like more stable um because the big players are spending their own free cash flow which is something like the do that was much more I think driven by debt. Yeah. Yeah, I think it absolutely does. Although to a limit, it doesn’t um insulate you from all the risks. But yeah, if you think of companies like Microsoft and Alphabet investing, they’re not going to stop investing because the Fed hikes interest rates by 50 basis points if that were to happen, right? They’re playing a long game that’s based on the fundamentals and they’re not borrowing money to invest it. So there are a lot of kind of macro shocks like they’re not going to stop investing at least immediately because the street of Hormuse is closed for example. So that that is um a real difference that you have long-term committed strategic investors. Uh a couple of caveats to that. One is while those have they’re very deep pockets, they’re not infinite pockets and we’re seeing some of these companies get down to a point of sort of break even on cash flow. And so the rate of growth that we’ve seen over the last few years, we are getting to the point where they will either have to take on debt or slow down their growth. I think frankly both are possibilities. Um there also some other things to think about. I just mentioned the straight of four moves like and right now we’re getting outside capital. So Middle Eastern investors are are a big uh supporter of things like Stargate and Open AI and stuff. So there there is still funding risk out there. Um, but I’m saying that just to sort of I don’t know modulate my answer a little bit. To your basic question, yeah, I absolutely think it’s a lot of a safer safer situation than the 992000 tech bubble. In some ways, you know, we talk about that as a bubble of course in hindsight, but we think about a lot of the business models that sort of were thought of then actually did succeed ultimately. They just had a very long period in the wilderness first. Yeah. I was thinking about this idea because you mentioned the idea that debt is starting to maybe first like get used a little bit more now. Like I was thinking about the the whole existential nature of the way the companies participating in this think about this whole thing which is like I’m wondering how much debt they’ll be willing to take on because they view this as like an existential battle between each other and that they’ll be like the winner that gets to AGI. So I think that’s just something interesting to watch going forward. Obviously they’re much more stable companies but just thinking about how much debt they’re going to be willing to take on like in pursuit of that goal. Yeah. And I think at some point we may see some of the companies step back and say, “Hey, you know, we don’t really need our own LLM.” I mean, Apple kind of didn’t get into the fray to begin with. And I think at the time when that first happened, they were kind of widely panned for falling behind and not spending enough. And I think as time passes, it’s kind of looking like they took the safe path there. They can license technology from Alphabet or others as they did with search. um and probably do fine. Of course, they have a lot of other advantages that most companies don’t have controlling the platform. And I think in terms of companies are maybe close to making that decision now or you could sort of imagine it. I think Meta for all um you all their reputation of being reckless spenders, they did have a year of efficiency not that long. They’re maybe a little closer to the edge than Microsoft or Alphabet is in terms of running out of cash flows. You could ask, do they really need Llama strategically to be their LLM? It’s not so clear. And the other to take the other side of M, the one thing they do have that’s really nice is they have their own use case in terms of directing content a ads and so forth. So they’re not providing cloud computing to other people who’s they can’t really know what’s going on with. They’re actually doing it for themselves. So they have probably the the best customer visibility of anyone uh in the conversation. Yeah, on Apple. Like I I don’t know anything about Apple, but it does seem to me it’s interesting like they they’ve kind of taken a step back and let everybody else do the spending. And I have a feeling like at some point they’re going to come from behind because they have everyone’s personal data through the phone and they’re they’re going to have some like breakthrough product probably built on top of other people’s technology that’s going to come through. But I could be totally wrong about that. Yeah, it’s interesting. Um I don’t think it’s clear now what the economics would be of a deal where um Alphabet partners with Gemini like who’s paying whom kind of when you think about with search right Gemini is I’m sorry Alphabet was actually paying Apple to get Google search on the iPhone because the iPhone is such a valuable property. I’m curious it seems to me just and I know you invested through this period. It seems to me like the tech companies of today are and I know you’re a quality investor are like significantly higher quality than the big tech companies of like the late ’90s. Do you think that’s accurate based on the types of metrics you look at? Um, yeah, absolutely. Um, now Microsoft was caught up in the tech bubble in

I think they were a high quality company then as they are now. And actually, one difference was that in 1999 Microsoft was at 50 something times earnings and they haven’t gotten really much above 30 in this last cycle or considerably less than that. So just this is a different point, but the valuations we don’t think are quite as extreme. Um, but to your point, yeah, there are a lot of companies that weren’t really very high quality there. We talk about sometimes the tech bubble. We also sometimes call it as the TMT bubble, right? It was telecom companies that were caught up in this just because they were laying fiber, you know, this very capital intensive, relatively undifferentiated business. Whereas now, u tech is probably the highest quality sector out there. Not to say every company’s high quality, but they do such differentiated hard things uh that um they really are able to maintain profitability is just kind of our core idea of quality. Um I used to sort of joke that tech investing was IBM and a bunch of crappy companies and now with all due respect, it’s a bunch of great companies. IBM, I mean actually even IBM, right, has come back. So they’re not really imply they’re not a a good company anymore these days. Can you talk about growth versus maintenance capex? We’ll put this chart up you had in the paper. Um because this is an interesting thing. I mean obviously initially in this a lot of this is growth capex building this stuff out but it’s going to require maintenance capex as time goes by here. So could you just talk about this chart and what you talked about with uh maintenance versus growth capex in the paper?

Yeah. So um imagine you’re Microsoft, you’ve been spending all this money in building up uh this Azure cloud computing capability and you have GPUs and CPUs and networking chips and all that. And if you were just to keep your present capability of compute going this stuff depreciates and you can argue about what the schedule is the from accounting point of view or in the real life lifespan of these companies but say every five years you’re replacing it. If you just said, “Oh, we have enough compute now. We can actually do everything we need to.” Your capex would be just kind of a fifth of what you your value of the stuff now just to replace it. But these companies are right now they’re growing capex like 60%. And virtually all that grow or more and all that growth is adding on new capabilities. So if you look at what the capex of these companies are today, we’re splitting between a steady but being diluted level of just keeping things running and this investing for future capability. That’s the growth capex. Um now if you’re Nvidia receiving that capex actually most of what you’re receiving is the growth capex. And so if Microsoft were to say inflect a little bit down and say not we’re not going to grow realistically but we’re going to grow a little bit less. We don’t need to grow as quickly as we were. We want to keep balanced with our cash flows or whatever. um then that would have a much bigger impact on um companies down below where in the in this in these four layers where they are getting more of the growth capex and the maintenance capex from the people above them which speaks to kind of the volatility as people talk about like the whip of the economic cycle right who’s at the end of the whip and who’s sort of close to the handle as you go down these four four layers you’re getting close to the end of the whip and then for better or worse you see a lot more volatility in revenue. Yeah, this growth versus maintenance capback is so interesting to me because I think it’s one of the big questions right now. I mean, we don’t expect these companies to keep spending at this level probably for a really long period of time, but like how much the maintenance capback is going to be um once this thing settles down seems like a very interesting question and a question that’s going to impact these companies a lot. Yeah. And I think what it’s actually you there’s actually been a fair amount of controversy over the last year about what the right depreciation schedules are and whether companies are not depreciating assets. But the fact is the youth use useful life of these chips is fairly long. They just they do wear out but it it takes a while. So that’s been one of the things in the industry is generally you can have fully depreciated assets that are still useful have still useful lives. Like people are still using ampear chips, right? There’s just so much demand that even if they’re not as efficient as the new ones, you still get a lot of utility out of them. And so um that that’s a good thing from the people who are buying this stuff. they get extra extra usage out of it. It’s a little bit of a I’d say risk for people lower down in the stack in that maybe the rate at which the the level maintenance capex the rate at which you just need to renew your current capacity is a little bit less than what you would think if you just looked at depreciation schedules and financial statements. How do you think about the changing nature of that we can call the mag seven or the hyperscaler companies or whatever in terms of they seem to be you know what made them great was they were these free cash flow generating machines they were capital light and now they’ve changed and you know I always ask this question about like have they changed forever have they changed for the worse and it was interesting we just had Michael Bosen on we posted it today and one of the points he made is obviously these companies are investing way more than they have in the past but they’ve always invested they’ve just invested in R&D and they’ve invested in things that got expensed you know through the income statement so maybe my question. The premise of my question is a little bit wrong in terms of maybe they’re not as change in their nature as as I think they are. Maybe they’re spending the money in different ways. So, how do you think through that? Yeah. And Amazon is sort of the famous one for investing through the income statement and sort of in an accounting sense really never at least not consistently earning very much money at all. um but at the same time being success fantastically successful in grow of course Amazon’s also fairly capital intensive companies from the warehouses and stuff even before you got into AWS and and cloud computing and we have with cloud computing that is a case where we’ve sort of seen this movie before like that was the critique of AWS and of Azure when they first came up is oh we thought these were software companies and they have super high gross margins and now you don’t because you have all this like dirty grimy hardware that collects dust and you have to replace and I they’ve proven that that can be a good business like capital light is great but capital heavy if you’re getting a high return on that capital that’s that’s fine too so I think the question is not so much are they capital intensive and is that therefore a bad business is okay yeah they’re getting these things that are capital intensive and is that a good business a TSMC right very capital intensive have been a great business for a long time. Um now admittedly these are the exceptions rather than the rules like most capital intensive businesses aren’t as great. Um but there is um something to the mo that great scale provides in these capital intensive industries that mean you can if the need to invest is really really big in dollar terms that means there aren’t that many people who can do it. And so if it there’s a real economic value to being it to it being done, you are able to get high returns out of doing it. So I’d say we’re not overly worried about just the fact that capital intensity is high. I think one thing you do see in capital intensive industries is the lead time between when you make that capital investment and when it pays off can be high. And that does present risk. But that’s also proved to your point about R&D spending whether it’s in tech or pharma or any of these areas that are R&D intensive. You’re talking about payoffs decades out. And yeah, a lot of the things you try don’t pay off. When you’re thinking about the return on capital on that investing, you’re sort of kind of playing the odds that on average it’ll deliver high return. In some ways, the capex that they’re doing now, it’s a little more certain, I think, that it’ll get a return. Maybe it’s the time frame isn’t so certain, but the eventually uh I think is pretty certain it’ll be used. So again, we’re not we don’t see that as a huge negative in and of itself, capital intensity. Yeah. To your point, the question I should be asking is maybe not the changing nature of these businesses, but the question I should be asking is their their investments in R&D in the past paid off massively. And the question is are are there investments in this tangible assets? Are they going to pay off similarly? I mean, that’s really the question that determines what happens from here, right? Yeah. Um, I think to some extent they won’t pay off as much as the initial R&D. I we’re sort of seeing kind of survivorship bias here. Like lots of people tried to start internet companies, a lot of them went nowhere, a few of them achieved escape velocity and because they’re the few that achieve scape velocity to get to where they are today. you know, they were they were lucky, they were skilled, they were in the right place at the right time. They got these unreproducibly high rates of return. So, they’re not going to see quite that again, but I think they are in a position where they will can get well above average exceptional rates of return on this investment. So, not as great as the past, but still pretty darn good, I guess, would be our forecast. And that would be different in some ways from past tech booms, right? Because in past tech booms, the builders have not necessarily been the biggest beneficiaries. You know, a lot of those builders like in the fiber build out like ends up going bankrupt. So because of because because of this intelligence, I guess because these better companies, you would probably expect the builders to benefit more here relative to the past, right? Yeah. Yeah. And that gets to the leverage point you mentioned earlier, like we haven’t gotten to a point where, at least for most of the companies we’re talking about or all the companies we’ve talked about so far, the debt hasn’t gotten to a point where if things go bad, there’s a recession, whatever, they won’t make it to through to the other side. Like these companies will. So that’s that’s a much different uh position. Um, you know, we talked about historically tech wasn’t viewed as much as a high quality sector and wasn’t because of the in or part at least because of the um pace of technological innovation too. You’d have one wave of technology and a company would win and then there’d be a new wave and that new wave would have different companies winning in that and there a lot of companies would fall by the wayside as you went from one generation to the next. And what we’ve seen that’s changed not just with the AI boom but over the decade or more that’s preceded that is it’s reached this sort of scale where it is the same winners of the previous generation who are in the position to win in the next generation. I kind of think that’s going to be true in a lot of cases in AI. Like obviously there are new companies like open AI, but I think a lot of the benefits to AI will actually fall to the incumbent tech companies, not just to new startups. How are you thinking about the LLM letter layer? You seem to be less optimistic on that and that’s an interesting layer because a lot of the technological innovation is going on there. But by the same token, you could argue eventually we’re going to have commoditization there. Some people have even argued eventually like Google because of their strong, you know, power relative to everyone else might make their best models free to try to take out some of the other companies. So like how are you thinking about that that LLM layer? Yeah, I I’d say I’m a little bit nervous about the LLM layer for for reasons you allude to. I think there probably more people trying to produce LLMs now than there need to be LLMs out in the world. I think it’s an open question like the one thing I I am a little bit uncomfortable with that with with the AI evolution is how much people are willing to extrapolate the progression of models into the future for any period of time. I think there is a real risk that we kind of squeeze all the juice we can out of all the data we have and just throwing more compute at it or slightly cleverer algorithms isn’t necessarily going to get anymore. that would be the the plateauing and and that hasn’t happened. And to be fair, people worried about that with Moore’s law of semiconductors for decades and decades before it kind of finally started to happen. So maybe I’m too cautious about that. But I think that is a real worry. And I think with LLMs, what would really different let one company differentiate from the other not so much we came up with a fancier algorithm or we have more GPUs than you, but more we have differentiated data. Then for an alphabet like apart from having the best scientists they also have the best data for a lot of purposes of any of those. So I do think they are in somewhat of privileged position but for a company that’s just sort of trying to come up with an LLM sort of with meto data and a lot of capital that’s going to be hard to differentiate and succeed. How about the software companies? They’re interesting to me because I would, and you can disagree with me on this, but I would have considered them very, very highquality businesses coming into this because of the recurring nature of their revenue. And this kind of came out of left field now. Everybody’s they’ve been completely derated. Everyone thinks every software company is going to disrupted. Like, how are you thinking about that? How are you thinking about the software companies in light of this? Yeah. Yeah. It’s it’s kind of ironic because three or four years ago, people would say on podcasts like this, I probably would have said with a straight face like, “Oh, of course, software is the best business in the world, right? recurring revenues, asset light, like close to 100% gross margins once you don’t stop reinvesting in re sales and suddenly this is a case to the earlier conversation I guess where suddenly not having tangible assets and being very assetike kind of worked against you with at least theoretically with a AI being able to come in and reproduce what you do in um at least in people’s minds a very quick and efficient way. Um now I think um I think those fears are overdone. I think the selloff this year certainly in the breadth of how many different software and software related business models have uh fallen is clear respect. Um investors are just throwing the baby out with the bathwater and shooting first asking questions later to toss my metaphors around. It’s um there is no mercy for any company that one can even imagine being um replaced by AI. I think a lot of what existing software companies have though is things like proprietary data regulatory lock in um entrenched position in the workflow of their customers. There even sort of network effects for software that is consistently used across practitioners in industry and becomes an industry standard. There are a lot of things that keep software from being disrupted other than just the code is really good. So I think the AI fears are a little bit high there, but uh it’s clearly had a big effect on markets and was clearly turned upside down what people thought of as high quality two or three years ago. And it’s interesting because in businesses where trust is important or in businesses where failure is a massive issue, you’d think people are probably overstating the the ability to go vibe code it. You know, if like we’re in the investment management business, like we can’t fail.1% of the time, we’ve got to fail 0% of the time. And so if you think like the the software that meets those types of tests, it’s going to be very hard I think for AI to overtake that I would guess. I mean I think to push back on that slightly. Um I have to say with all humility in the investment management business, you do fail a lot of the time in the sense of not every stock I pick goes to go. I guess more in terms of like losing client data or something like that. Not not in terms of your client. So client data, you know, financial system generally sort of regulated regulatory data. You can get in a lot of trouble with one slip up. So that’s the sort of thing that probably isn’t going to be replaced very quickly by AI. You know, the logistics in your supply chain, not not something you want to mess with. Uh the cost of failure is just too high. And the other thing is software is yes, it’s a huge industry, whatever one a half trillion whatever it is spent on enterprise software, but it’s actually a pretty small industry relative to the cost of the people who are hired to use the software. So just getting someone a cheaper version of Salesforce doesn’t really save a company that much money. They what they would really need to do is replace the person who’s using Salesforce to to save money. U I do the kind of software they be less um sanguin about. I think it’s more at risk is maybe like data visualization like you don’t have proprietary data. It’s more um if it’s not right it’s sort of immediately obvious it’s not right. So it’s sort of easy to check the AI. I think that’s a that’s really kind of a key concept of how how bad is if it’s wrong and how easy is it to check if it’s right too. Um so I think there’s there is stuff that’s definitely at risk. I don’t want to minimize the disruption of a disruptive technology but I think there’s a lot of stuff that’s not. This is a good time to take a step back because we have been talking about quality and and quality is like is probably one of the hardest factors to define. I mean first of all you got your discretionary people who you know they do it in their own mind. I mean the types of stocks Warren Buffett might own, people probably consider quality, but the for the quants like this is probably the biggest variation in terms of the quant shops and their definition of any factor is probably quality. People do it in so many different ways. Like how do you think about the definition of quality? How do you define it? Yeah. And you probably don’t have too many guests on your show and who say don’t say they buy high quality stocks when you talk to equity investors. The um so TMO’s history as an investment firm that goes back to the 70s. It actually started from very deep value groups. Jeremy Grantham and Dick Mayor were at battery march in the 70s and kind of made a name for themselves buying small cap value stocks. So GMOs on train equality was sort of coming from hey we buy all these low multiple stocks and we’re always missing out on these great companies that consistently have high profitability. They seem to do really well over time despite being at higher multiples because they just stay at higher multiples and they grow very well. And oh by the way during recessions and times of economic downturn these stocks actually hold up better. So it was not so much a this is this silver bullet factor that just gives you better performance. It’s more almost more like better risk adjusted performance. You get you can get an equity return but without low with lower risk or you could say you can take lower risk than the broad market without giving up return. So that’s what kind of got GMO into the idea of quality in terms of specifically how we define it. Um coming from that background is what kinds of companies deserve to trade a premium. And yeah, it’s growth, but it’s really profitable growth. If you can redeploy capital into a business and get a higher return on that redeployed capital uh than your investor could if you just dividended out to them and they did whatever with it, then you should trade a higher multiple. And that’s kind of what we’re trying to identify future high consi consistently high return on capital of course in a capitalist system with competition that means must mean you have some mode around your business you do something that competitors can’t equally duplicate um we do have backward-looking factors so we could have a like if you were to did in the 80s come up with a quality factor we would look at a history of high profitability high return on equity stability across the economic cycle strong balance sheet those are the kind factors, but really what we’re trying to get at where our fundamental work comes in is will that be true in the future? How much of it is balance sheet versus consistency? Like it seems like a lot of quality investors we talked to have maybe gone more towards the consistency side over time and maybe a little bit away of the balance sheet and that that kind of gets into the the good the big tech companies now being very high quality because they’ve been very very consistent performers, right? And they also have traditionally had very strong balance sheets as well. And we think we think do think the balance so it’s both not or in our mind the the balance sheet is important because um bad times can happen to anyone things happen in the world and a lot of being quality is just being able to keep going through those tough patches and if you have a lot of debt you are kind of at the the mercy of strangers as it were when the um when you go through those rough patches. You also have to dry powder to invest when there are opportunities and it is kind of a you know if you have a great business you’re generating a lot of cash you have to ask why a company would need debt too. So um we’re pretty suspicious and skeptical of high levels of debt but there also cases where hey a company just did a big deal this is transformative they have the cash flows to pay it down they just took a lot of debt on now we can we can be accepting of of higher debt levels. Um but I guess if you were to make me choose between the two, the debt sort of a safety shity thing, but the really cina for us you have to have is that view that you can deploy capital or earn a higher rate of profitability. Profitability in the sense of return on investments than the average company. That’s the the core idea and the debt is or strong balance sheet is an enabler of that. I wanted to stop and ask for that definition because I want to ask you about Oracle because you you refer to your decision to sell Oracle in the paper and I’m sure that has a lot to do with quality. So can you talk about that because Oracle is interesting and they are using way way more debt. We talked about the other hyperscalers you know using cash. Oracle is using way way more debt here. So can you talk about the decision to sell Oracle? Yeah and Oracle is a stock we’d held for a long time and you back when we bought it was a very very strong balance sheet a relatively low growth company but also lower multiples. It was kind of a very sort of boring in a in a good loving way boring kind of business and they pivoted very successfully and saw growth opportunities invested into them but they took that beyond the level at which we are comfortable as quality investors. So as you say they took on a lot of debt and that debt um is only serviced through their customers being willing to and able to pay them revenue. So um and they do have a lot of customer concentration. So Oracle can have whatever ironclad agreement legal agreements with Open AI um that they might have, but that’s still those are only good if OpenAI is solvent. We don’t hit a a rough patch or they aren’t the LLM that drops out in what’s potentially competitive world and losses out to others. So that puts Oracle in a sort of riskier position than we feel comfortable with as quality investors. So it’s not when we sold it, it wasn’t so much a call like oh the stocks got up a lot and the valuation’s expensive. It’s more like this is the kind of balance sheet that we can’t be comfortable underwriting in the portfolio. How much of that is quantitative and how much of that is you as a portfolio managers looking at Oracle and say I’m not comfortable with where they are right now. Yeah. Um so we have a lot of qualit quantitative metrics. We do qualitative screening on quality for sort of idea generation and to prompt the question. Ultimately the decision is always a fundamental one. So I would say it’s very our our fundamental decisions are very data informed is the way I’d put it. If this spending slows down, one of the things you talked about is is if if spending slows down and looking at this whole value chain you talked about earlier like where the places that might get hurt the most are like how would you think that through? like if this AI spending does slow down, who would be most impacted by that? Yeah, I think a simple rule of thumb is whatever’s gone up the most is probably what’s going to go down the most. So that’s, you know, in recent months that would be the memory stocks where that’s emerged as a a pinch point and the stocks have done fabulously well and I think if things go badly, they would unwind that pretty quickly. um that that the more general version of that getting back to the four layers is probably the further down um that stack you are or that whip metaphor I used earlier the further down that stack you are the more volatility there’s going to be in your fundamentals and therefore probably your stock price there are mitigating factors like some companies are sort of 100% AI plays and some have diversified businesses and some have stronger balance sheets than others so of course the more pure play you are in AI uh the more you have debt or anything like that, you’re going to be more volatile. But generally down that stack will do worse if you’re the hyperscalers like they’re not going to do well if there’s an AI slowdown, but they’ll do relatively well. So you end the paper with three principles you would think about in terms of where to invest. Uh durable competitive advantages, the ability to prosper across a wide range of scenarios and via uh resilience via less volatile AI revenue and balance sheet strength. C can you talk about those? Yeah. Um so sustainable advantages and think about companies that you know they may do very well now but five years from now is their products really going to be that differentiate. I think that’s where the LLM companies might come into question for us. Like there’s a big first mover advantage for OpenAI but I can’t really be convinced that they’ll be the best and probably aren’t the best LLM today or they’ll be even a relevant LLM in five years from now. That’s relatively unclear. Conversely, I’m pretty sure that everyone’s going to be using ASML to tools to build the chips that go into AI 5, 10, 20 years from now. You might have a question about how big the market is, but ASML is going to have their piece of it with with pretty high confidence. Um even uh secondly, however big you think AI is going to be 10 years from now, I think it would be foolish to think it’s going to be a linear path from here to there. They’re going to be draw downs. I would expect along the way. They could be very savage ones. And so you want to make sure you have companies that can survive those. And the way that the two things that help you survive those really are um you don’t have a lot of debt and you have sources of revenue that are kind of low beta to AI funding investment drying up. So you know if you’re youidia they have they’ve been sort of a victim of their own success and they’re pretty much all their revenue is the AI data center business. Samsung this is now vestigial gaming businesses that that is their root whereas broadcom they’re growing their their AI business a lot you they do have 40% of their revenue from very annuity like software they do have Apple revenue they have other networking chip revenue I mean it’s a very diversified business and so they’re actually they do have some debt but they have so many other things that we think that’s relatively safe or some of the other companies are very cashri that element of of safety in the business model is I think critical for not just even if you’re an ultimate winner, you need to survive um the tough times that come along the way to that to that victory. Just last one for me before I hand it back to Justin. Just when you close out the paper, you had a really interesting chart here. You’re looking at your you’re looking at the different layers of the stack we talked about earlier, but you’re looking at the GML quality, what percentage it owns of them versus the S&P 500. And I thought it was really interesting like the conclusions here. So can you just talk really quickly about how your your quality portfolio looks different than the S&P 500 from this perspective? Yeah. So um so really where we are is in two of those four layers. Uh we are in the hyperscaler layer um and we are in the infrastructure layer. So the the hyperscalers they have the safety they have the cash flow they have the other businesses. We also think they have the visibility um to deploy capital most effectively of anywhere else. They’re closest to the end markets and the applications. Um so we can and they also by the way are not trading at very high multiples right now. So um that that that’s sort of the safe place to invest and the place where in some ways it’s safe because they are sort of the arms dealer type companies is even below Nvidia like the TSMC’s and the semiconductor equipment companies the stocks are very volatile but um I can’t see a world in which AI plays out as successful and it’s different companies are building the chips or building the tools to build the chips like I actually can’t quite the say say the same thing about Nvidia like there is competition in Nvidia there are custom chips there’s AMD even if Nvidia is still relevant it may not be able to earn the kind of margins it does today so it may not be the same scale market dominant it has it’ll still be there but I don’t think its position will grow from here if I look down below that into the the semiconductor equipment chain that I think will at least grow at least with the market if not more as it becomes more capital intensive so it’s kind of at the top and the bottom of where we’re in invested. Um those are like over the long run I think they’ll they’ll rise together over the short run they do tend to uh move at different trajectories like this year we’ve seen the semiconductors do much better than the hyperscalers I think that is likely to rotate over the over the next year or so but we think those are both very attractive places to invest. We’ve talked a lot about AI and the tech space, but I I was looking at the the ETF that that you run, the high quality, the GMO high quality ETF, and clearly tech is, you know, I think the largest sector alloc I don’t have this the sector allocations up. I have the holdings up. It looks like tech is probably the largest sector allocation. But, you know, it’s not it’s not all tech. I mean, you there’s, you know, there’s some healthcare in here. There’s some financials in here, but I’m curious and I don’t mean this to be like like like put you on the spot like I got you, but like no Birkshire Hathaway. Like how do you have you guys ever held Bergkshire? Do you ever look at Bergkshire Hathaway? Like how do you view that as a quality company just given the business and the cash and stuff like that? Yeah. Um we don’t hold it. I wouldn’t rule out holding it. It’s we and we do hold some of the companies are in Bergkshire’s public equity portfolio which I guess is one of the reasons holding us back from investing in it but it’s also true that a lot of Berkshire is private companies that we couldn’t otherwise access. Um so uh I think we generally think it’s a mix of businesses is in aggregate pretty high quality because of the capital allocation that because there are some high quality businesses in there. It’s not um I could give you exact number off the top of my head but not two/3s high quality, one-third not. So uh we haven’t to date at least thought that aggregate it quite meets our standard. Um but it’s we’ve not quibble with a peer who had a portfolio. How often do you like how often would a new name come into a quality portfolio? sort of talk about the I guess the new idea generation process, what you go through there and then do you you know what would you do? Would you replace a name? Would you what what what would be the sell criteria to free up some cash to get that name in the portfolio? Like how do you manage that? Yep. Um so I mentioned we do we do fair amount of screening on financial statement characteristics of the signature of high quality companies like long trajectories of high profitability. That’s actually kind of what got us into tech originally is like when you know Berkshire Hathaway Warren Buffett famously was slow to invest in tech I think a lot of high quality oriented managers were if you sort of look at the digital signature the profitability just increasing and the balance sheets being strong even though it’s looking backwards with a lag it did kind of force you to look at these companies and um so that that kind of disciplined screening can be really valuable and that’s the first step of our process in both directions too. It can also tell you something’s deteriorating and sometimes things happen so slowly as a human you of maybe don’t notice these slow incremental changes. We do however it is always a human decision in terms of the the new names. Sometimes it’s a truly new company or company whose performance over the last few years has just risen up to grab our attention. And so like Uber which we had within the last year if there was a period where it wasn’t a high quality company and slowly its cash flow started improving and wow this is actually a very profitable company now and hey like we should look at it. Um sometimes it’s a company that’s been around for a long time and we’ve thought of as high quantity company something we’ve held in the portfolio and then something happens in the market it falls out of favor. We’re asking ourselves hey you know the stock’s gone down a lot. Do we think this is still a great business? you know, five years from now, it’ll be fine. If so, maybe we’ll buy it. We’ve actually bought some of the sort of software related companies that would feel like I’ve been overpunished by this AI disruption narrative u within the last year. Sort of examples of that. It’s not so much our view on the quality change, although we had we had to reiterate with AI, but the valuation opportunity was there that hadn’t been in the past. It’s um somewhat similar on the sell side. Like the good problem to have is hey we have a great company. It’s done really well. You know it’s a quality business but the stock price is just so expensive that we don’t see how we can generate a great return from here. Um and we have liquidated a couple uh tech companies over the last couple years for that reason. Or sometimes it’s a case where this is a less happy situation. which is a company we’ve held for a while. Things are deteriorating. Um, we just don’t think it’s high quality company anymore. Actually, we mentioned talked about Oracle before. That was one that we essentially downgraded their quality. That was a nice one for us because the stock had done well. Often that is the case though where you’re selling underperformers. I think that’s that’s a hard thing for investors to do, at least a hard thing for us to do is sell underperformers. But sometimes those are your your best trades. um it’s not strictly one in one out in the portfolio. We do like to have around 40 to 50 names as sort of a sweet spot where you can get diversification but have a concentrate enough portfolio that you really feel like you’re on top of every position. So I I think I probably recognize, you know, 95% of the names in the portfolio. You know, they’re mostly large cap, you know, strong big brand names that I think most people would recognize. But is that how do you think of and there’s not really small caps represented in here so maybe it’s a large cap mandate that you’re really trying know that’s where you want to kind of play but is there anything to be said for the different the differentiation between looking for quality companies you know across different market caps and different sizes and do you ever find that you come could come downstream a little bit from like maybe the 250 or what’s in the S&P 500 into other areas of smaller companies. Yeah. And there absolutely are uh high quality companies that are in smaller cap and size isn’t part of our definition of quality. I would say if you’re a great company in a big industry, you’re likely to get pretty large. But there are some niche industries where even the strong player in the industry is never going to be that large. Uh in fact, GMO does have a small cap quality strategy. um which is u it is a more cyclical group. There’s not obviously mega cap tech. It’s a little bit of a different kind of universe, but there are definitely companies that meet that criteria. And there are a couple of smaller cap companies in the um what you might call our large cap or all cap strategy, but smaller I’m talking about that sort of 10 to 15 billion type market cap range, not real pure small cap companies. Do you think that um you know international stocks typically have traded at multiples lower than here in the US and you know some people argue that you know the US market has better quality more diversification um and therefore that’s why you get sort of the premium you know in the US market and particularly but do you think if you were to be looking at international stocks or building international portfolio would you look that defining quality any differently or would you kind of stay in the same lane? Oh, and we do do some global investing in um broader portfolios. We do not define it differently in different markets. I’d say the prevalence of quality is less outside the US. One of the reasons why international markets are cheaper uh is that they just have a different mix of companies. They have more of the kinds of businesses are naturally low multiple businesses. But that doesn’t mean that you I mentioned ASML earlier with lithography. That’s a great company, right? That’s a high quality company. I have no qualms about that. I also observe that ASML has traded it a premium to its closest peers in the US. So just because it’s non US doesn’t mean it’s cheaper. In fact, often I would say there’s maybe even a scarcity premium for the highest quality companies outside the US. So um they aren’t cheaper. there hasn’t been this great valuationdriven reason to invest globally. There are certain industries where the best companies are outside the US. I think investors are benefit from investing globally. Sometimes we do actually manage a international quality ETF as well as the US quality ETF and it is I’d say similar kinds of companies or these companies that meet a sim the same threshold for quality. One of the things that I know Jack and I have always valued in talking to people from GMO is you guys have a good ability to kind of get down to just, you know, like first principles type of things important in investing in terms of how you’re building portfolios, how you’re looking at companies. And so, you know, I think this discussion has been a valuable one today be largely because of that. Um, and so we we really appreciate it. Uh we have two standard closing questions we like to ask all of our guests and you can go anywhere you want with these but um the first is what is the one thing you believe about investing that most of your peers would disagree with you with. Um, well, this is kind of a small one, but it came up uh recently for me is um share repurchases. Like you go to an average investor meeting and the stocks down and you say, should you do a repurchase now? Or um the in management teams will talk about, oh, we’re opportunistic in our repurchase program. To me, like repurchases is sort of a blanket way to return cash is fine and tax efficient. If you’re a company that’s trying to cherrypick your stock price, you’re basically trading on insider information because you know more than everyone else does about your company. And while you’re benefiting the investors who hold your stock if you’re right and buying low, you’re doing the opposite, you’re disadvantaging your investors who are selling to you or and so I I don’t really feel that that’s appropriate for companies to have share repurchase programs just based on the valuation of the stock. So that’s a little bit of a pet peeve of mine where I think I’m on a island of pretty close to one in the investment world. Yeah, I haven’t heard. So what would you do? Would you just have like if you would have like an automatic like if a company approved billion dollar buyback, you would just automatically set it and have it Yeah. Do you like if you’re the CEO of a company, right? You can’t just go out and buy stocks anytime you want. You have these programs where you buy on a certain cadence or or sell you more often on a certain cadence that’s sort of set in advance. So you know you are trying to game the stock price and sometimes you’ll get lucky sometimes not but you can’t just be opportunistically trading your own stock. And then the last question is based on your experience in the markets what’s the one lesson you would teach your average investor? Um well I’d say the average if you’re think by average you mean sort of professional investor not professional investors have a big advantage which is by definition they’re not paid by someone else which means they don’t have a boss. So don’t you don’t have to be conventional, right? Think think you can’t get fired. So take advantage of that that you can be a lot less constrained by benchmarks. I’m not saying you should concentrate overly and do necessarily risky things, but I feel like people too often just fall into what’s conventional or what a professional who’s paid and has career risk of being wrong tells you what to do. And I think people can can be a little bit more creative without necessarily taking more risk. Um, so that’s I guess my sort of big picture advice. Maybe my more micro advice is don’t pay as much attention to what happened today. Like I think a year from now is what’s happening today going to matter that much a year from now? If the answer is no, you shouldn’t be trading on it. Great. Good discussion, Tom. Thank you very much for joining us. We appreciate it. All right. Well, thanks for your time. I enjoyed this very much. Thank you for tuning into this episode. If you found this discussion interesting and valuable, please subscribe on your favorite audio platform or on YouTube. You can also follow all the podcasts in the Excess Returns network at excess returnspod.com. If you have any feedback or questions, you can contact us at excess returnspod@gmail.com. No information on this podcast should be construed as investment advice. Securities discussed in the podcast may be holdings of the firms of the hosts or their clients.