Speaker 1: How much money does AI actually need to make to justify the massive investment going into it? Aswath Damodaran breaks down what it would take for these companies to justify their valuations and why AI may need trillions in revenue to make the numbers work. To hear the full conversation, listen to episode 780.
Speaker 2: So think of this collective factory being built to make AI products and services. You have all these companies that are feeding into the factory. You got Nvidia selling the chips, you got the power companies supplying the power, you have the data centers being built, the real estate. So there’s this architecture companies that make their money in building the factory. But included in your mix are those companies that are spending the money on this architecture because they want to make products and services. I would include Meta and Alphabet and Amazon in that mix because they’re not interested in the architecture. They want to use the architecture to make money. The first thing to do is step back instead of adding up all of those market caps, because some of these companies make their money from building the architecture and some hope to make their money from the products and services that come out of the architecture, is to separate the companies and look at it differently. The building up of the architecture, and forget the market cap because that can get contaminated or affected by other things, is the actual amount being invested in the architecture, building of the data centers. By my estimate, it’s in excess of $2 to $2.5 trillion. This is the largest buildup in business history for any new business.
Speaker 3: Over what period is that investment going to be?
Speaker 2: I dated it back to November 30th, 2022. That’s the day ChatGPT went public, ’cause that’s the day AI came into our public consciousness. This is really over 4 years. So this is bigger than what the railroads spent in the 1800s in terms of, you know, dollar value. It puts the dot-com investment to shame because that was a tiny amount. This is $2+ trillion. So you built a factory that cost $2 trillion, and the companies that supplied the inputs to build the factory have clearly made money. Nvidia’s already made its money, right? It hopes to continue to make money. The electrical equipment companies, the power companies, the real estate—all of that, the components have already made money building the architecture. But we’ve also built the largest factory in history without a sense of what that factory will produce as products and services and whether people will pay for the products. It’s the largest leap into the unknown as well that we’ve made in history. We have no sense of what it will deliver, but we’re investing upfront. So the debate about AI is: what will come out of these factories? What are the products and services? What will people pay for them? And collectively, is there enough money to be made from those products and services to justify the $2 trillion plus in CapEx? So that reframes the discussion, allows you to separate companies that are architecture companies from the companies that hope to make money on products and services.
Speaker 3: Fantastic. Do you have answers to those questions, or speculation?
Speaker 2: I think we’re starting to see clues, right? I mean, the company that probably makes the most money from AI products and services right now is perhaps Anthropic with OpenAI, because they sell through their subscriptions and usage models the AI products and services collectively. Even the best-case estimates of the products and services that come out of AI right now—the revenues, not earnings, the revenues from those products and services—even with the most optimistic numbers, it was about $250 billion over the last 12 months. If you remember, the architecture of the factory you built was a $2 trillion factory, and you have $250 billion in revenues. Let’s play the worst-case scenario. If revenues level off at $250 billion, that $2 trillion is almost entirely going to be written off. Your revenues will have to climb from $250 billion to a much, much, much, much higher number. How much higher? Well, it depends on what kind of margins and profits you can make. So let’s say there’s a steady state 10 years from now, 15 years from now, where the AI business is well established. We’re looking at the collective revenues from that business, asking how much would those revenues need to be to justify the $2 trillion invested upfront? The revenues will have to be maybe $8 to $10 trillion, because if you think about the expenses to produce the revenues, and the profits, and the tax on the profits, you’re very quickly going to start to scale the ladder up to $10 trillion. And that becomes then the core question you’re asking: can the AI product and service market be large enough to generate $10 trillion in revenues with decent profit margins along the way? It can’t be mass market products. So that’s, I think, the big unknown we’re jumping into. I cannot rule out the possibility that it can be there, but I use what I call the 3P test. Is it possible? Is it plausible? Is it probable? Right now it’s possible—you can have $10 trillion revenue, and I’ll explain why that possibility exists. Is it plausible? You’re probably pushing the limits of what’s plausible with $10 trillion. And is it probable? Right now, I’d attach a low probability to it. The reason I’d be cautious about using the words “impossible” or “this is a bubble” is you’re making a judgment that I don’t think you can make with the data we have right now.
Speaker 3: One of the other components with these companies attached to this factory are a number of other very profitable, durable revenue streams, like Google’s or Alphabet’s, right? Like Microsoft’s.
Speaker 2: And those should not be part of the AI discussion. There are ways in which these companies are coming up with the cash flows to fund them. Bringing them in will just contaminate the questions you’re trying to answer. So in the case of Alphabet and Meta, they have incredibly successful advertising businesses. The only way they enter the discussion is if the factory turns out to be not profitable in hindsight and we have to write it down and there’s losses to be taken. The question is, where will the losses go? In the case of Meta and Alphabet, it’ll be their shareholders who will lose, because the money from the advertising business that could have been used for dividends and buybacks was instead funneled into the business. There’d be write-offs, those shareholders will lose money, and they will have regrets about what was done, but the damage will be, you know, limited there. With CoreWeave, it gets messier. The reason it gets messier is if the factory has to be written down, CoreWeave has enough debt that you worry about not being able to make debt payments. And when you’re unable to make debt payments, that pain gets spread to the rest of us. That is pain that damages not just the companies involved and the players involved, but everybody else. So I think that’s where I would draw the line. The fact that some of these companies have incredibly profitable businesses means that the damage from AI is going to be more contained in those companies. But the companies that don’t have the side businesses are going to be much more exposed if or when there’s a write-off of the AI investment.
Speaker 3: I think the question that millions of Americans are asking, to some degree, is: if I’m an S&P 500 investor, I’ve got something like 39 to 40% of my position in this complex, in companies attached to this factory. What I was attempting to get at there is, if the factory does not generate the returns necessary—if that possible but improbable outcome does not happen, right—then my investment in Alphabet is not going to go to zero, most likely. How can I correctly separate out the factory from what I’m doing with my money, in today’s market, and how I think about allocating to an index fund or any of these companies individually or in aggregate?
Speaker 2: My question is, what’s your alternative? I mean, you are exposed to AI risk, whether as an investor, as a human being, as a worker. This is an incredibly large disruption. There are 2 choices you can make. One is you can say, “I want to get out of this AI space because it’s likely overpriced and I don’t want to see the pain,” which means you pull your money out of companies that are AI-related. That’ll include most of the Mag 7, a lot of the AI companies, and you put your money in the rest of the index. The good news for you then is when the correction comes, you’re going to be hurt less. But the bad news is it could take 3 years for the correction to happen, and what you lose while you wait might vastly exceed what you benefit. It’s a market timing question. It’s a question as old as time. And I know in hindsight we crown the people who ride out these bubbles that burst as heroes. We did it with the dot-com, we did it with 2008, but we don’t follow through and look at what happened to them in the after years. I mean, take somebody who managed to get out of the market in 2008 because of the crisis that was coming. That’s good news. We view them, look back, and we made movies about them. We make them stars. If you track it, what most of them did in the decade after—you know what they ended up doing? They ended up staying out of the market for much of the last decade, because once you get your money out of the market, especially if it’s all of your money, it gets very difficult to decide to get back in because there’s too much at play. You get too cautious. So what you gain by not being in the storm, you lose by staying out of the market for too long. The facile advice you could give is “stay away from AI companies, invest in the rest of the market, you’ll be more protected.” But the long-term answer might be, hey, ride the wave through. You will lose money when a correction happens, but if you have a long enough time horizon, you’re going to be okay. And a long enough time horizon doesn’t have to be 50 years—it could be 6, 7, 8, 9 years. If you’re 2 years from retirement, for God’s sakes, get your money out of stocks then. Put it into something—and bonds right now deliver 5.5%. Don’t get greedy. But if you’re 35, I don’t think it should alter the trajectory of your investing in any significant way, even if you believe there’s an AI bubble.
Speaker 1: Today.
Speaker 3: So you’re going to be mad at me. I moved out of the market-cap-weighted index into the equal-cap-weighted index, and I also could remain in some factor tilts. I also own real estate. That was my answer to it. Sounds like you did not agree.
Speaker 2: That’s a mild core shift, right? It’s not a major one. The bigger concern I have is people pulling their money out of these stocks and essentially going into utilities or consumer products, thinking that’ll protect you when AI collapses, or worse still, taking your money out of stocks and leaving it in cash because you want to wait for a good time to get back in. An equally weighted versus a value-weighted index, I can live with as a choice you make because you feel too exposed. I mean, in investing, I have what I call the sleep test. And the sleep test is, if you lie awake at night wondering what your portfolio is doing, you fail the sleep test. And you have to do whatever you need to do to pass the sleep test. And if your concern is, “I’m overexposed in a value-weighted index in the S&P 500 because of these big-cap companies, I’m going to sleep better if I have an equally weighted index,” I think that’s worth an insane amount of, if not money, well-being, to be able to do that. So that, I think, is a reasonable course correction to make if that is your biggest concern.
Speaker 3: If I understand correctly, you have an opinion about each or many of the companies that we’ve just discussed in this complex, and you make personal investment decisions based on that analysis. Is that correct?
Speaker 2: Mm-hmm.
Speaker 3: What are you doing in response to today’s environment?
Speaker 2: I used to own all of the Mag 7, but I bought them at very different points in time. I bought Microsoft when Satya Nadella became CEO in 2013, ’14. I bought Facebook—half of it after the fiasco they had in 2017 with Cambridge Analytica, and the other half after the metaverse fiasco, where people decided that Mark Zuckerberg could not be trusted to run a company. I bought Tesla just ahead of the COVID correction. So I bought them at different points in time because at that point in time I wasn’t buying them for AI. I was buying them because they looked undervalued to me given what I thought about their future. And I make that point because people always look at the Mag 7 and say, “I could never have bought the Mag 7 because they’re such expensive companies.” In the last 20 years, each of the Mag 7, including Nvidia, has had at least 3 or 4 points in time where they were incredibly cheap because of something that happened to them. I bought Nvidia in 2018. I’d love to tell you I saw AI coming—wasn’t the case. I bought Nvidia because it had a pricing collapse that made it look cheap to me on a valuation basis. And coming into 2023, I owned all 7. I now own 5 of the 7. And there’s a story behind the 2 that I sold. I sold Tesla right after the election because I don’t like to own companies that become political as well as business plays. And whatever you think of the politics of it, politics is now part of the Tesla story, for better or worse, and I don’t feel comfortable with it. I sold Nvidia reluctantly because it’s been an insanely big winner for me. My split-adjusted price per share for Nvidia is $1.80. So when it hits $100, you know, you’re looking at an insanely big payoff. I sold a quarter of it in late 2023, a quarter in 2024, and the rest over the last 6 months. Have I left money on the table? Absolutely. Do I have any regrets? Absolutely not, because I think Nvidia is an awesome company, but it’s being priced as the greatest company ever, and to me that’s not a good investment. You can have amazing companies that are not good investments. If I look at the remaining 5, my guess is all 5 are either fairly valued or overvalued by a little. You’re saying, “Why aren’t you selling?” Because just as there’s a margin of safety when you buy, there’s a margin of safety when you sell. Taxes especially, I think, contaminate the investment process because they can affect when you sell. Because when you sell, you have to pay taxes, and that’s going to be larger on your biggest winners. And to me, that means that something has to get overvalued by about 25 or 30% before I sell. Because living in California, by the time I add state and local taxes on top of my federal taxes, I’m looking at a 26, 27% capital gains tax on winners. So I’ve sold all of my Nvidia, I’ve sold all of my Tesla. I still have the remaining 5, and I watch the pricing. If they get overvalued enough, I will start to shed them. The fact that I will leave money on the table when that happens doesn’t bother me. I mean, I also have simple rules on my portfolio where I will not let an investment get above 15% of my overall portfolio. It’s on autopilot. I sell it once it hits 15%, which means none of these stocks are in double-digit levels now on my portfolio. On any given day, I could have a really bad day with Meta or a really bad day with Alphabet—I don’t notice it at the portfolio level, which is what I want in my portfolio. So I think it requires kind of monitoring what you have. It’s one of the prices you pay as an active investor—you can’t buy something and forget about it, especially in your biggest winners, because then those companies can very quickly become 30, 40, 50% of your portfolio. I don’t care how great you are as an investor, that’s taking a risk you should not be—
Speaker 1: Well, I’m feeling a little bit seen here right now. My husband and I have been investing since the late ’90s, very tech-heavy. Scott, I don’t remember exactly what the money guy said. Is it 70% of our portfolio is in companies that are headed by Elon Musk?
Speaker 3: Yep.
Speaker 1: And then 85% are these, like, 7 tech stocks. At what point do you start to get rid of stocks that are such an outlying weight in your portfolio, like when you still believe in the viability of the company?
Speaker 2: Mine is 15%. It’s an absolute cap. So no individual company can be above 15%. If you say 70% of your portfolio is in Musk stocks, then you have a lot of Tesla in your portfolio, well above 15%. But you gotta do it with open eyes, which is you can’t have regrets. What I mean by that is there will be stocks you sell that will double after you sell them, and you’re saying, “I wish I hadn’t done that.” So here’s my suggestion: selling all of it is going to be too much. It’s going to be too much on a tax basis. It’s going to be too much emotionally. Do it in stages. So sell a quarter of your Tesla stock, you know, put it on autopilot. Every 6 months, I’m going to sell a quarter of my Tesla stock no matter what the price is. Because if you have to think through whether this is the right time to sell, you will find a reason not to sell stocks that have done well for you. It’s human nature—you hold on to your winners ’cause they’ve been so good for you. And we have the same kind of relationship with our losers—we hold onto our losers because we hope things will turn around. So sometimes you’ve got to take these decisions out of your hands and put a limit sell. So one of the things I use is limit buys and sells when I know I will not have the stomach to make the decision myself. There are stocks I love, but I don’t like the price they’re selling at. And 3 of the stocks that I track are BYD, Palantir, and MercadoLibre—3 companies that I like for very different reasons. But when I first looked at them, they were all overpriced. I valued them, which required that I kind of get comfortable with what they are as companies. And I put limit buys at prices well below today’s price, without an expiration date. You’re saying, “Why would you do that?” Because if they drop by 30, 40, 50%, they’re dropping for a reason. There’s going to be a lot of bad news around them. You will not feel ready to buy because you’re surrounded by bad news, you’re surrounded by selling. So sometimes you almost have to take decisions out of your hands, because you know emotionally you will not be ready to make those decisions. So do it in stages. Don’t try to do it all at one go, because, you know, trying to do it all in one go, it’s too much. You will not pull the trigger. It’s just too much to do.
Speaker 1: Yeah, we had this same advice with the Mad Fientist right at the beginning of COVID when the market was crashing and he was putting more money in. He’s like, I’ve got this plan, but it required me to execute it every single time. So when the market comes down 10%, I’m putting in more. When it comes down another 10%, I’m putting in more.
Speaker 2: March 23rd, 2020 was the absolute low. That was the day when if you looked at the news stories, your first reaction is, I’m going to sell everything and head for the caves. And that’s, I think, part of the problem — the times when you should be buying are often the times when it’s most difficult to make that buy decision. In the abstract, intellectually, you can say yes, but emotionally you won’t be ready. So times when you see a company that you really like, you’ve got to do your homework, understand the company, because most companies, when you see prices fall, they fall for good reasons, right? So you want to separate out the companies you want to be buying when the price falls from the companies where you’re catching a falling knife, where you’re buying them and they look cheap, but they get cheaper and cheaper and cheaper because there’s something fundamentally broken in the business. So I think doing your homework is critical as well when you put in these limit buys and limit sells.
Speaker 1: Do you invest in any index funds, or do you just do individual stocks?
Speaker 2: I do individual companies, because my portfolio now is 50 companies built up over a period of 40-something years. They’re all over the globe, different sectors. So one of the advantages of my job, which is teaching, is I have to go to other parts of the world. I have to value companies I’ve never heard of. And once in a while I look at a company and say, this would be an amazing company for my portfolio. It’s a company that’s on nobody’s radar because it’s a Turkish company making infrastructure investments. I’ve never heard of this company, but then you take a deeper look and say, this is a great company. So my portfolio is diversified enough that I don’t gain much by going to an index fund in terms of diversification. And to the extent that these companies are even slightly better than an average company, I’m going to be okay in the long term. So one of the first rules in investing is do no damage. It’s like the Hippocratic oath — don’t do something that can hurt you big time. Those are the kinds of things that hurt you as an investor. That’s why you should never have 80% of your money in one stock, no matter how well it’s doing. You should never get overinvested in a particular macro story, no matter how strongly you feel about it. Because it’s not that you won’t get a big payoff if you’re right, but if you’re wrong, it takes you down and it changes your lifestyle. Your kids might not be able to go to the college they were planning to go to because you screwed up. So avoiding those big screw-ups is as much a part of designing an investment portfolio as it is finding the big winners.
Speaker 3: I would love to go back for a minute here to the AI complex. And this is wonderful advice. And Mindy, I’m going to call you out here — your SpaceX and Tesla stock are in your retirement accounts, so you don’t even have to worry about the tax hit that he’s talking about here when you actually make those moves. I hope you guys act on it before, you know, that changes things for you.
Speaker 1: We’re in the middle of a big move. Once we’re finished moving, which should be the end of this week, then we have a lot of time to have conversations about it.
Speaker 2: And I think — especially getting more balance in, especially if it’s in your pension, it’s in your 401(k), or somewhere where it’s tax protected — then I would park it for the moment into an index fund. Even if you can’t find something specific to invest in, if you don’t like index funds, put it in — you know, I go directly into the Treasury. You can invest in Treasury bills directly, put it into a 6-month bill. You’re going to get close to 4%. And while you wait, start looking around, because once it goes into cash, it gets sticky. It’s very difficult to get it out of cash, because, you know, it just feels safer staying there. So you want to kind of do your homework while it sits around.
Speaker 4: I’d love to go back to the AI complex for a minute here, and I’d love to talk about Alphabet. Alphabet specifically, because I think it’s a wonderful illustration of what confuses me as an amateur relative to someone like yourself by a long shot. But when I look at Alphabet, I see revenue of $120 billion, price-to-earnings ratio of like 15. I think we had $90 billion somewhere in that range on their income statement, or $77 billion unrealized gain in the last quarter’s income statement on what was likely Anthropic and SpaceX. Yeah.
Speaker 2: Mark-to-market, which you shouldn’t even count as part of earnings, but no.
Speaker 4: Perfect. We have that. We also have an enormous — I think it’s $200 billion — in deferred revenue, revenue that will be realized in the coming years, coming from Anthropic, I believe. Right. And Microsoft has its parallel universe over there with OpenAI.
Speaker 2: And I think that’s what troubles me most about the AI story, which is the companies that claim to be companies that want to make their money on AI products and services. Let’s take xAI. You know where xAI’s primary revenues came from last year?
Speaker 4: xAI? No, I do not.
Speaker 2: It didn’t come from selling subscriptions to Grok or usage of Grok. It came from taking data centers they built and leasing it back out to Anthropic. And the problem with that story is, if you are telling me as a banker that xAI is worth a lot because the AI business is going to be big and Grok is going to be a key player in that business — the catch in that story is you’re now leasing out the factory you built to your biggest competitor. And if you truly believed your story, that would never happen. So either you don’t buy into your own story and you want to make your money by selling to other people who might buy into the story more, or there’s something I’m not seeing here. So you’re right, a lot of those deferred revenues come from renting out data centers they’ve already invested in to other people and collecting. There’s nothing wrong with doing that — you’re getting money while you do it, but it cuts against your bigger story, why you’re making the investment in the first place, which is you think there’s money to be made from selling AI products and services. And if you believe that’s so, how come you’re leasing your factory out to competitors who can do that? What is it that you hope to do? There’s a circularity here that comes from many of these companies reporting revenues, but they’re basically leasing the factory space to each other. That’s why I would keep my eye on the product and service part, because without that, all of that intra-company leasing is all going to fall apart, because none of the data centers will be valuable if nobody buys products and services that use them.
Speaker 3: You know that feeling you get when checking your finances means logging into 5 different apps?
Speaker 4: That’s why I use Monarch. Setup takes about 10 minutes. You can link your accounts, and everything you own and owe lands in one clean dashboard — banking, cards, investments, even keeping track of recurring charges. After that, you can stay on top of your money in just minutes, not hours per month. Monarch saves you time and actually helps you use it wisely. Use the code POCKETS at monarch.com to get your first year of Monarch half off, at just $50. That’s 50% off your first year at monarch.com with the code POCKETS.
Speaker 1: You had said earlier they’re building data centers for $2, $2.5 trillion, and right now Anthropic’s biggest revenue is $250 billion.
Speaker 2: No, that’s not just Anthropic. Anthropic is less than that — their annualized run rate is about $70 billion.
Speaker 1: Oh, that’s not just Anthropic. That’s everybody, $250 billion.
Speaker 2: Everybody. Palantir, Anthropic, OpenAI — the collective revenues from selling products and services. So no contamination from leasing. So that’s not fair, right? Because that’s just the factory being leased across people building. This is the actual stuff being sold from the factory. There’s only $250 billion collectively. That’s the bad news. The good news is it could be huge. So now let’s ask a question: how huge can it get? And that requires dealing with an existential question about AI — what is AI going to do? Let’s take a business. Let’s take McKinsey. Let’s assume Anthropic comes up with Claude consulting agents, which will help McKinsey consultants. It’s a tool that makes them more productive. It’s a cost to McKinsey, right? McKinsey will pay for these tools, but because it’s a cost on top of their existing consultant pay, it can’t be huge, right? It can’t be billions, because, you know, if you’re paying $800 million in consultants, you can’t pay another $800 million on tools, because then where are your revenues going to come from? So it’s — if it’s a tool, it’s a relatively small number. So already, if AI is just a tool, the potential revenues you’re talking about are much smaller. I’ll give you a sense of what that collective number, the maximum, can look like. Last year, across the globe, the collective amount paid to employees in salaries, wages, compensation, all the good stuff collectively was $26 trillion. Every employee in private and public companies across the world. So let’s say AI’s pipe dream is it can replace every employee at every company — I’ll tell you in a minute why this is going to be a nightmare for the rest of the world, right? — but potentially revenues could be up to $26 trillion. They’re probably going to be smaller, because why would you replace an employee with an agent if it costs the same? Potentially up to $26 trillion. But here’s why the story crashes and burns almost instantaneously: if your end story is AI will replace every employee, it’s true, that’s a lot of money, but then all these employees who are now displaced have lost their jobs, right? And their incomes and their consumption power. So what are all these companies that have replaced their employees with agents — who are they going to sell this stuff to? So that story kind of crashes and burns because the macroeconomics don’t work out. So when people talk about total addressable markets, $26 trillion is the absolute limit, but it’s got to be much lower than that, because you have to have a sustaining economy where people really have jobs. That’s why I think if you’re talking about total addressable markets of more than $10 trillion, you’re already reaching the limits of what’s plausible. Because you’re talking about AI then displacing a fairly large percentage of the workforce — it can’t be just a tool anymore. If it’s a tool, it’s a $2 trillion total addressable market. We can already start writing off big chunks of the factory. If it’s potentially replacing a lot of white-collar employees, it could be $8 trillion, $9 trillion, $10 trillion. You could get to some of the AI infrastructure being justified, but $2 trillion is probably still going to be too rich a CapEx. The problem is $2 trillion is right now, and these companies are not done, right? They’re continuing to build data centers. They’re continuing to add to the infrastructure. It could be $3 trillion very quickly. We’re adding about $700 to $800 billion every year to this CapEx. So unless you stop, this is very quickly going to become a number that is $3 trillion, $4 trillion, $4.5 trillion without the—
Speaker 1: As a reminder, you can listen to the full episode, which is number 780, or find it in the show notes below. We know you may have a lot of trouble finding a fee-only or advice-only, not AUM, planner who actually understands financial independence. Scott and I have started building out a network of these professionals and more over at biggerpocketsmoney.com/fipro. That’s F-I-P-R-O. That wraps up this episode of the BiggerPockets Money Podcast. Thanks for tuning in.
