Howard Marks on the AI Bubble, Irrational Exuberance, and Investing Under Radical Uncertainty

Howard Marks

You have to accept the likelihood that what you're doing is closer to speculating — and I don't say that word pejoratively — than analytical investing. There is a spectrum, which goes from analytical investing in prosaic, understandable companies to speculative investing in futuristic companies that can't be described at all. You should calibrate your activities based on where you are on that spectrum. That's the whole thing, and it's very hard to do. This is the hardest thing I think I've ever seen in the investment world, because of this enormous degree of uncertainty.

Howard Marks

There's no question about the fact — to use Alan Greenspan's phrase from about thirty years ago — that we have exuberance. That's the only thing we know for sure; he pioneered the phrase 'irrational exuberance'. The question is whether today's exuberance is irrational, and I don't think anybody can definitively say so. I've never heard anybody tell me exactly what AI will be able to do, or when, or for whom, or how much profit it'll produce and for whom. There's an arms race going on between what you'd call some of the greatest companies on the planet — I'd describe the hyperscalers as mostly the greatest companies I've ever seen — and they're engaged in an arms race. Can only one win? Is it winner-take-all? Can several win? Nobody can tell me these things.

So I don't think there's an analytical, value-based way to decide whether to participate in these IPOs, or at what price. AI, as far as virtually all of us are concerned, is a concept. We can't define its parameters. It's a great concept — likely to be the most powerful force any of us has ever seen — but that's all we know. Deciding whether to participate, and at what price, is what my South African friends call a thumb suck. You can't put numbers on a pad and figure out what these things are worth, which is what value investors like me have historically done.

Howard Marks

It does feel like that. I've seen several technological innovations, and read about many more over the last hundred and fifty years — this may be the greatest, the most powerful, and also, in many ways, the least specifiable. Take the railroads in the 1860s, radio in the 1920s, the automobile, computers in the '50s and '60s, the internet in 2000. It may be revisionist history, but I think we had a much better view of what all of those could do. They didn't have the unimaginable, unlimitable upside that AI has, or the degree of uncertainty. We knew the railroad would carry goods and people from coast to coast; we knew radio would carry messages. We may not have known exactly how they'd produce profits, or that radio would become television. But every one of them was accompanied by what we call a bubble: people got excited about unprecedented developments, they threw vast amounts of money at building the infrastructure, there was a winner-take-all race, exuberance, capital flowing in like water. In every case, too much capital flowed in, too much infrastructure was built, prices were paid that were too high, and a lot of the people who provided the capital lost their money.

I wrote in a memo recently that if this technological innovation, with its exuberance, doesn't produce a money-losing bubble, it'll be the first. It could happen — you can't rule it out. And maybe this is a good time to introduce the optimist's rejoinder: 'this time it's different.' That was true about the railroads, radio, computers, the internet — but this time it's really different, because we have a development of incalculable, unlimitable value, so there's no price too high. That's what they say. But the problem, Ed, is they always say that. This time it's different is never different — they said it in each of the bubbles I mentioned. Nobody, including me, should say definitively that this is a bubble, or that people investing in these early stages of AI will lose their money, or that the people investing in the companies you named will pay prices they'll never see again. But you must be alert to the possibility. The way people get into trouble is by not being alert to it.

Howard Marks

My favourite fortune cookie says that the cautious seldom write great poetry. Investing in these companies today could be a huge error — but it could be great poetry. The people who resist because it could be an error could miss out on the greatest thing in history. That's what makes these decisions so hard. People who invest today in traditional industries — transportation, distribution, retailing, real estate — don't have the risk of committing grievous error, but they also don't have access to the possibly best thing in history. When you sit here with something this young and unestimable, you just have to deal with it as a concept, or not deal with it at all.

Howard Marks

Most people deal with the future by coming up with a forecast. I argue you need two things, not one: a forecast, and a judgment about the probability that your forecast is right. You can make an optimistic forecast about the future of AI, but if you say, 'This is my judgment, and by the way, I'm highly confident I'm right,' I think you're probably making a big mistake. I've never met anybody who thinks they can tell me what the world will look like five or ten years from now — so why should a young person laying the foundation of his portfolio conclude that he's probably right when nobody else is?

We know AI could be great — I said in my last memo that, in terms of its basic capabilities, it's probably more likely to be underestimated today than overestimated. But the question is how much capital it should receive, and what a piece of a company engaged in this activity is worth. The old-fashioned value investor figures out what a company makes today and what earning power it's building, tries to work out its earnings in five or ten years, puts a reasonable valuation on those earnings, and compares that with today's price. I don't think I've ever seen an industry where that's less feasible. If somebody tells me what they think Anthropic's net earnings will be in 2036, I'll bet they're not within fifty per cent of the truth — of course, we have to wait ten years to find out. But if I'm right, then investing in Anthropic's IPO means accepting that what you're doing is closer to speculating — and I don't say that word pejoratively — than analytical investing. There's a spectrum from analytical investing in prosaic, understandable companies to speculative investing in futuristic companies that can't be described at all, and you calibrate your activity based on where you sit on it. That's the whole thing, and it's very hard to do — this is the hardest thing I think I've ever seen in the investment world, because of this enormous degree of uncertainty.

Howard Marks

That's what we do for a living, and historically we've made those judgments pretty well. But since the internet came along, roughly thirty years ago, we've had a new concept that's extremely important today: disruption. You take a prosaic company in a prosaic industry, and you think you can anticipate what it'll look like in five or ten years because it doesn't have technological things that will make or break it — but then you have to ask whether that's really right.

Thirty years ago, in the value-investing business, we talked about a 'moat' — the things that surround a company and protect it, that make it less attackable. The value investor has always preferred companies with moats. A great example was a newspaper: if you owned the newspaper in a given city, it would be hard for a competitor to start from scratch, because the used-car ads and the help-wanted ads and the movie times would be irrelevant coming from another city. It cost a quarter, so anybody could afford it, and if you bought one today you'd still have to buy one tomorrow, because yesterday's newspaper is already obsolete. That was a strong set of moats, and a lot of smart people made a lot of money investing in newspapers — the same was true of the movie industry and other communications businesses. Now the newspapers are largely out of that business, under profit pressure, because the internet and digital communication gave them competition nobody thought possible thirty years ago.

So what can't be disrupted now by AI? Who can't lose their job to AI? I used to say plumbers — but maybe a robot can come into your house with a camera, assess the situation, and make the repair. Then I said a masseur — why can't a robot give you a good massage? The world has become a much more uncertain place; the probabilities you can assign to the future are much broader than ever. When I was a kid, the world didn't change — a comic book was always a dime, new technologies didn't come along that often, and we were pretty confident the world would look the same ten years later. For the most part it did. Today you have to accept that much more change is possible, and the investor has to recognise that he or she is living and dealing in a much less predictable world.

Howard Marks

I still think there's a more predictable part of the economy. It'll probably be a while before the energy business is disrupted to the point where we use something in lieu of oil and gas — that's probably largely true of the food industry, the timber industry, home building, transportation. It's probably going to be a while before we walk into a station, become dematerialised, and show up in another city. Retail has been disrupted, but we're probably at a baseline level of in-person shopping that won't go much further. You can identify areas — metals and mining, paper, chemicals — where, for the most part, things with less intellectual content are less likely to be disrupted by AI, which is basically an intellectual problem-solver and productivity tool. We can make a list of things we think are less likely to be disrupted, but we shouldn't be too cocky about it. That's what we do for a living, and we're still investing according to the same philosophy in many of the same industries — but we have to constantly renew our thinking. The most laughable thing to do today would be to say, 'I found industries that'll never change.'

Howard Marks

We start with the traditional indicators of valuation — the Shiller, or Cape, ratio, or the traditional S&P PE ratio. I used the expression a year ago that the market is 'lofty but not nutty'. The non-Shiller PE ratio is about 23 today; the eighty-year average is 16, so we're roughly 50 per cent higher. But in 2000 it was 32. When I started in this business in 1969, in the research department at Citibank, the banks invested in what were called the Nifty Fifty — the best and fastest-growing companies in America: Xerox, IBM, Kodak, Polaroid, Merck, Lilly, Texas Instruments, Hewlett-Packard, Coca-Cola, Avon. Most of those stocks sold at PE ratios between 60 and 90. Look at the Mag Seven today, take out Tesla, and they're selling at PE ratios in the thirties — that sounds expensive to me. But PE alone is too simplistic: the companies are different, their capital intensiveness is lower, and their marginal profitability is higher, since the product is intellectual rather than a piece of metal — it doesn't cost much to make the next one, so their incremental profitability is much higher. And we've never seen companies growing at today's rates — companies growing 50 per cent a month, or 100 per cent a year.

Three years ago, most people thought software was a great industry to invest in, because everybody who used computers needed it, and if you had a software system that served your company, it would be expensive to change — a pretty good moat. More recently, people are wondering whether the whole software industry is going out of business, because nobody writes software any more — AI writes its own software; people just have to tell it what to write. Around the first of February we had what people called the 'SaaS-pocalypse', when the great AI companies announced coding models and everybody said that's it, the whole software industry is going out of business. That's probably an exaggeration, but it's very hard to figure these things out.

Coming back to how you invest given all this uncertainty: one of the greatest mistakes you can make is not being optimistic enough. Another is saying the future is unclear, so you can't invest — but the future is always unclear, maybe more than ever, and that's not a reason not to invest. You just have to invest carefully and knowingly, aware of the risks you're taking. There's a spectrum: at one end, ultra-high possible returns with great uncertainty; at the other, somewhat lower possible returns with less uncertainty. All of it is more uncertain than ever, but the spectrum still exists, and you can choose a point on it. You can invest in the hyperscalers — Amazon, Google, Meta, Microsoft — established businesses with moats and enormous operating cash flow, competing vigorously in this winner-take-all battle; you'd think investing in them is the low-risk way into AI, but if AI booms in the next three years, their other businesses hold back their growth rate, so they won't be the maximum winners. Then there are the established, one-product AI companies — Anthropic, OpenAI, Nvidia — harder to specify, since we don't fully know their profitability or finances, but with a high probability of still being successful in five or ten years, even if not number one; they may be riskier than the hyperscalers on price, but they're already up and running. And then there are the startups — no revenue, or revenue but no profits, maybe not even a defined product — where if you get in at ground level and one turns into a big winner, you can make an incalculable amount of money. I described this in a recent memo as a lottery ticket: most people who buy lottery tickets lose all their money, and a few become incredibly rich. You can pick where to play on the spectrum, mix positions across it, and decide what share of your portfolio each deserves.

Howard Marks

In the heat of the moment, in the exuberance, people say profits don't matter, what matters is the future. We used to value stocks on earnings; then, investing in companies with no earnings, we talked about a ratio of sales; then, companies with no sales, people in 1999 and 2000 asked how much per eyeball, how much per click, and put values on internet stocks based on how many people visited a site, even for free. But I believe it ultimately always comes down to value. At some point in the future, profitability will matter, and if a company is a great tech leader today but still won't be making money twenty years from now, my guess is that the price paid by an exuberant investor today will produce disappointment when exuberance is replaced by sobriety. Of course profits matter — we invest in companies we think will make money. Warren Buffett said, in connection with the internet in 2000, that there's no doubt the internet would add to efficiency, but that's not the same as adding to profitability. That's relevant today too. AI is going to change the world, I have no doubt about that — but who will it make money for? If all the hyperscalers, plus the Anthropics and OpenAIs, plus Tesla and the startups, engage in battle and compete against each other at enormous cost, how profitable will they actually be? And if AI is primarily a labour-saving device, who gets the benefit of the labour savings — maybe the customer, the shipping company, the retailer, the warehouse — rather than the purveyor of AI services? These things can't be specified now.

Howard Marks

How long do you have? When I attended the University of Chicago in 1968, a professor pointed out that the average mutual fund did worse than the S&P before fees, and then charged a high fee on top — so why not just buy one share of each stock in the S&P? There were no index funds then, but the idea came along, and today the majority of mutual-fund equity capital is managed by indexation or passive investment. That's one reason a lot of people have disappeared: the consumer was paying a high fee for a defective product, which is not a great business model.

On the other hand, over the last forty-odd years we lived through a period of declining interest rates, which made a lot of things very successful. A lot of people built very profitable businesses in what are called alternative investments — private equity, private credit and the like — in an environment that was perfect for them, especially since March of 2009, the low point of the financial crisis: things have been rosy for over seventeen years. Closest to home for Oaktree: in the last fifteen years a business called private credit developed — really just a broader term for direct lending, private loans for mid-size buyouts. It didn't exist in 2010; today it's 1.7 trillion dollars, with roughly 700 direct-lending managers. That availability of capital, alongside a favourable economy and generally declining rates, put a lot of people into business. But of those 700 managers, I'm told roughly 3 per cent were in business before the financial crisis — so we don't know how many have what it takes to deal with a harsh environment. Making money in a placid environment proves almost nothing; you can do it on judgment, hard work and skill, or on aggressiveness and luck, and it doesn't get sorted out in the good times. As Buffett says, it's only when the tide goes out that we find out who's been swimming naked. The period since 2009 has been salad days, halcyon days — the greatest period imaginable for the investment industry, especially alternatives — and nobody should look at those seventeen years and call them normalcy. One day the tide will go out, and some of this will be sorted.

Howard Marks

I think it's overblown. These are managers who collect money from clients and make loans for mid-size buyouts; some will be unsuccessful, but probably not a large percentage. This activity has been around under different guises since 1977 or '78 — I was lucky enough to start Citibank's high-yield bond activity in '78 — and we've been making loans to companies of moderate creditworthiness, and doing well, for 48 years. People who don't do it as well won't have great results, but most of the loans will pay, and I think the people throwing up their hands are exaggerating the difficulty.

Having said that, retail investors bought these products, and there's no market for them — you can't get out at the drop of a hat. If you went into a non-traded BDC, which is what these vehicles are, the terms said you could only redeem 5 per cent of investors per quarter. People said, 'What do you mean, I can't get my money out?' But that was always the term, in the prospectus, which admittedly very few people read. People do things in the good times, without adequate care, and regret them in the bad times — that's some of what's going on. I don't think there was misrepresentation; people shouldn't be surprised they can't get all their money out every quarter. One of the most powerful forces in the investment business is disillusionment: people went from unworried to now thinking the ship is sinking, and that's painful. The unworried feeling was mistaken, and the feeling that the ship is hopelessly sinking is probably mistaken too.

Howard Marks

I've enjoyed a great career, and I don't consider it over. Investing is a fascinating field — just think about this podcast, and the number of times I've said 'I don't know', or 'unpredictable', or 'inestimable', or 'incalculable'. What we do every day is peel an onion: we deal with uncertainty and make the best judgments we can in an uncertain world. Nassim Taleb, in Fooled by Randomness, compared investing with dentistry: if you go to dental school and learn to fill a cavity, you'll fill it the same way every time and be successful every time. That's not true of investing. If you're the kind of person who wants to be successful every time, don't become an investor — become a dentist, or an engineer, something with physical rules that are reliable. There are no physical rules in investing that will make you successful all the time. Warren Buffett, the most successful investor of all time, attributes his success to twelve investments over sixty or seventy years — he didn't have that many abject failures, but many of his investments were only moderately successful. He did twelve great ones. So ask yourself: do you like dealing with uncertainty and ambiguity? Can you live with a batting average far from a thousand? Investing has been an enormously profitable industry for those of us in it over the last fifty years — you shouldn't become an investor just because it's high-paying, but if you meet that description, I think it's a great thing to do. It's exciting, intellectually challenging, and you never reach a point where you say, 'I've got this figured out.' I find that a wonderful attribute.

Howard Marks

Thank you, fellas, for your great questions. It's been a pleasure.