Latest / Deep Values: Value Investing Deep Dives / Seth Klarman Warns of an AI Bubble
Transcript
- 0:01Imagine being in the market for like 44 years through
- 0:05blackmondaythe.com. Crash 2008, a literal global
- 0:09pandemic. Yeah, that has a whole lot of
- 0:11volatility to survive, right? And in all that time managing
- 0:14billions of dollars, you've only ever had five down years.
- 0:16It's honestly incredible. And your absolute worst drop
- 0:20ever? Just 10%.
- 0:22But here's the hook for you listening today.
- 0:25This exact investor is looking at the artificial intelligence
- 0:28boom right now and he is seeing a massive, dangerous bubble.
- 0:33Yeah, the man who basically never loses money is terrified
- 0:36of the biggest tech revolution of our lifetime.
- 0:38Exactly so. Today's deep dive is all about
- 0:41pulling apart this paradox. We're looking at a highly
- 0:43revealing interview with Seth Klarman.
- 0:45The legendary founder of Bowpost Group.
- 0:47Right, The interview is on the Iconnections YouTube channel.
- 0:49And what he laid out isn't just, you know, a basic critique of AI
- 0:52hype. No, it's, I mean, it's a master
- 0:54class in how to invert the way you think about risk.
- 0:57Because in an ecosystem where everyone is obsessed with like,
- 1:00catching the next rocket ship and maximizing upside, his
- 1:02entire superpower is an absolute obsession with the downside.
- 1:06O before we get into is brutal warnings about AI, I feel like
- 1:09we have to understand the mechanics here.
- 1:11Like how does someone actually survive 4 decades of crashes
- 1:16with a near flawless record? Well, you do it by making
- 1:20structural safety your absolute baseline.
- 1:22Like most investors look at a stock and ask, you know, how
- 1:25much money can I make if everything goes right?
- 1:27Right, the upside. Exactly.
- 1:29But Carmen looks at at an asset and asks how much money will I
- 1:32lose if everything goes horribly wrong?
- 1:34So he refuses to use leverage. Oh interesting, so no borrowing
- 1:37money to amplify returns? None, because borrowing money is
- 1:41the fastest way to get wiped out in a downturn.
- 1:43Beyond that, he demands a huge margin of safety on every single
- 1:46bottom up analysis and he is perfectly willing to just sit on
- 1:50cash for long periods if he doesn't see a deep discount I.
- 1:53Actually want to pause on the mechanics of that structural
- 1:55safety for a second, because I know he also relies heavily on
- 1:58something called senior securities.
- 2:00Yeah, he does. For anyone listening who isn't,
- 2:02you know, deep into corporate finance, what does that actually
- 2:04mean in practice? So it comes down to the literal
- 2:08legal order of who gets paid if a company goes bankrupt.
- 2:11OK. When a business issues both
- 2:14stock, which is equity, and bonds, which is debt, the debt
- 2:18is considered legally senior to the equity.
- 2:21Right, they're at the front of the line.
- 2:22Exactly. If the company goes under and
- 2:24has to be liquidated, the courts ensure those debt holders get
- 2:28paid back first from whatever cash is left.
- 2:30The stockholders are at the back of the line.
- 2:32And they usually get nothing. Right.
- 2:34Basically 0. So Klarman frequently buys the
- 2:37senior debt of a company he's legally insulating himself.
- 2:41Even in a worst case scenario where the business fails, his
- 2:44position in the capital structure means he still gets
- 2:46his money back. And you know, combine that with
- 2:49cross asset macro hedges like buying put options that payout
- 2:52if the broader stock market crashes, and he essentially caps
- 2:56his maximum possible loss. It's, well, it's like playing a
- 2:59sport where your entire strategy is just refusing to give up
- 3:02points, right? Knowing the other team will
- 3:05eventually get tired and make mistakes.
- 3:07That's a great analogy. But I have to ask, why is this
- 3:10strategy so rare if it's clearly so successful?
- 3:14Well, as Clement points out, talking about downside just
- 3:17isn't sexy. I mean, it doesn't play well on
- 3:19financial TV during a roaring bull market.
- 3:21Yeah, nobody wants to hear about crash protection when everyone
- 3:23else is getting rich. Exactly.
- 3:25But protecting your capital when times are euphoric means that
- 3:29when the market inevitably corrects, you actually have dry
- 3:32powder. You can aggressively deploy
- 3:34money into the bloodiest environments.
- 3:36Because all your competitors are too wounded to act.
- 3:39Right. Or just too terrified, which is
- 3:41when you buy incredible assets at fire sale prices.
- 3:43Which brings us perfectly to the biggest, shiniest object in
- 3:47today's market, right Artificial intelligence.
- 3:50Klarman explicitly states this current environment has the
- 3:53characteristics of a bubble. He really does.
- 3:56He points to this sort of optimistic new era thinking that
- 3:59always accompanies financial medias, and his primary piece of
- 4:03evidence for this is honestly kind of hilarious.
- 4:06Oh, the all birds indicator. Yes, the shoe company.
- 4:09They literally added the letters AI to their corporate name and
- 4:13their stock price jumped. It is the exact same
- 4:16psychological phenomenon we saw in like 1999.
- 4:20When companies wouldjustadd.com to their names and watch their
- 4:23valuations explode. Yep, completely irrational.
- 4:25But let me push back on Klarman's premise here on behalf
- 4:29of everyone listening duringthe.com boom.
- 4:31The Internet actually did change the world.
- 4:34It did. And isn't AI actually game
- 4:36changing? I mean, Carmen himself admits in
- 4:38this interview that generative AI could cure diseases or read
- 4:41X-rays better than human doctors, right?
- 4:44So if a technology is genuinely world altering, how can
- 4:47investing in it be a bubble? Well, the answer lies in this
- 4:49concept called multiples versus uncertainty.
- 4:52And this is perhaps Klarman's most profound insight for anyone
- 4:56evaluating the market today. We can break that down for us.
- 4:58Think about how a financial multiple works.
- 5:01A multiple is simply the premium you're willing to pay for a
- 5:03dollar of a company's earnings. So if a business makes $1.00 a
- 5:07year and you pay $20.00 for a share, you are paying a 20X
- 5:12multiple. Got it.
- 5:14In traditional finance, if you face massive uncertainty about a
- 5:17business's future, you demand a lower price, right?
- 5:21A lower multiple to compensate you for taking on that unknown
- 5:24risk. Because if I don't know what the
- 5:25future holds, I'm sure as hell not going to overpaid today.
- 5:28Precisely. Now apply that logic to AI.
- 5:31We are facing like ultimate uncertainty right now.
- 5:34We don't know if these models will hit a performance wall or
- 5:37achieve actual AGI. We don't know if the long term
- 5:40winner will be Google, a startup, or some open source.
- 5:44We don't even know how it'll impact global employment or
- 5:46copyright law or inflation. Exactly.
- 5:49Logically, that incredibly high level of uncertainty should
- 5:51force the market to demand lower multiples, but instead the exact
- 5:56opposite is happening. People are paying 4050 or even
- 5:59infinite multiples. Right for companies with 0
- 6:01profit, based entirely on this distant, unpredictable future.
- 6:06So what's fascinating here is that the bubble isn't
- 6:09necessarily the technology itself.
- 6:12Is the price tag. Yes, the bubble is the absolute
- 6:16certainty that investors have priced into the valuations
- 6:18despite the reality of massive uncertainty.
- 6:21So for you listening, if you're looking at an AI stock right
- 6:24now, Kwarman's framework really forces you to ask are you paying
- 6:29for the current reality, or are you paying an astronomical
- 6:32premium for a complete mystery? Exactly.
- 6:35But here's where it gets really interesting.
- 6:37Karman isn't avoiding the space entirely, right?
- 6:39He actually owns Amazon and Google.
- 6:41I think they make up about 10% of his entire book.
- 6:43He does, yeah, but his reasoning is completely stripped of the
- 6:46hype. I mean, he didn't buy Amazon and
- 6:48Google because of their large language models.
- 6:50He bought them because they are, in his words, enormous cash flow
- 6:54machines with highly versatile cloud data centers.
- 6:57And he didn't buy them at peak hype either.
- 6:59No, he waited for brief windows of market panic like the tech
- 7:02sell off a couple of years ago and he bought them at below
- 7:05market multiples. So the downside was protected by
- 7:08their existing massive retail and search businesses.
- 7:11Exactly. He didn't have to make any
- 7:13heroic assumptions about the future of AI to justify the
- 7:16price he paid. Contrast that with the companies
- 7:19he's aggressively avoiding, the massive foundation model
- 7:22companies like Open AI and Anthropic.
- 7:25Yeah, he stays far away from those.
- 7:27Because their business models currently require like,
- 7:29incinerating billions of dollars just to train their models and
- 7:33stay relevant. Yeah, he makes great point that
- 7:35a business forced to constantly set fire to its own cash just to
- 7:38survive is not Warren Buffett's definition of a great business.
- 7:42The capital expenditure required is just staggering and Klarman
- 7:46asks this critical question, is this actually a winner take all
- 7:49market? Right.
- 7:50What if it's not? If open source models just
- 7:52commoditize the technology, these huge companies might never
- 7:55recoup those billions in training costs.
- 7:58The uncertainty is simply too high for the valuation.
- 8:01But the most Seth Klarman play in this whole AI landscape isn't
- 8:05software. It isn't chips.
- 8:07It's raw land. This is.
- 8:09So smart. He's buying dirt that just
- 8:11happens to be located next to massive power substations.
- 8:13I mean, here's the analogy. He's not even selling pickaxes
- 8:16during the gold rush. He's buying the dirt where the
- 8:19pickaxe store eventually needs to be built simply because he
- 8:22confirmed it has electricity. Just a brilliant exploitation of
- 8:27physical reality over digital hype.
- 8:29You know, we think of AI as this invisible code in the cloud, but
- 8:32the cloud lives in physical data centers.
- 8:34Right, very physical, very power hungry building.
- 8:37Yeah, a single AI data center can require as much electricity
- 8:40as a small city. You cannot just plug that into
- 8:43any neighborhood grid. Building new power plants takes
- 8:46a decade. Navigating local environmental
- 8:48permitting for massive transmission lines takes years.
- 8:52The power grid is the ultimate physical bottleneck.
- 8:55By buying that specific land, he is basically holding the
- 8:57bottleneck. Exactly.
- 8:59He is creating valuable optionality.
- 9:01He pays very little for the raw land.
- 9:03If the insatiable demand for data centers continues, tech
- 9:06companies will be desperate for land with existing power
- 9:09hookups. And he can lease or sell it to
- 9:11them at a massive premium. Right, but if the AI boom
- 9:14fizzles out, he still owns a piece of land with valuable
- 9:18infrastructure. And because he didn't overpay,
- 9:21his downside is virtually 0. He also applied this
- 9:24structurally safe approach by buying into a private non China
- 9:29Asian data center venture right at a 40% discount to what public
- 9:32markets are paying. Yeah, he always finds the back
- 9:34door into a trend. Because, you know, the AI vacuum
- 9:37is sucking up all the capital and all the media attention
- 9:39right now. It leaves massive parts of the
- 9:41market completely ignored. Oh, absolutely.
- 9:44With his forty person investment team, Klarman is out there
- 9:47hunting in the anti hype. They're looking at AI agnostic
- 9:51companies, businesses like roofing suppliers or housing
- 9:54materials that won't be disrupted by AI, but whose stock
- 9:57prices are drifting lower just because investors are bored.
- 10:00They're also targeting perceived AI losers.
- 10:02Oh interesting. Yeah, the market is making these
- 10:04sweeping assumptions that certain software or service
- 10:07related businesses will just be destroyed by automation
- 10:11overnight. So the equity in these companies
- 10:13gets clobbered. Exactly, down to incredibly low
- 10:16cash flow multiples. So Klarman's team steps in,
- 10:20analyzes the actual fundamentals, and often finds
- 10:23that the market has wildly overreacted, which creates a
- 10:26deep value opportunity. But out of all the complex
- 10:29investments they research, his absolute favorite idea right now
- 10:33blew my mind. Yeah, it wasn't crypto.
- 10:35It wasn't some biotech startup. It was assisted living
- 10:39facilities. It perfectly illustrates his
- 10:41contrarian method though. Think about the timeline of
- 10:43assisted living. These facilities were totally
- 10:46battered after COVID. Families were terrified to place
- 10:49their relatives in them, occupancy plummeted and a
- 10:52massive wave of bankruptcies swept through the sector.
- 10:54Right. But the underlying demographic
- 10:56reality, like an aging baby boomer population, that never
- 10:59changed. The need didn't disappear, it
- 11:01was just delayed. And because of the bankruptcies,
- 11:03new construction completely halted.
- 11:06So now the bad debt has shaken through the system, the
- 11:10oversupply is gone, and the fundamentals are turning sharply
- 11:13positive. And most importantly, Kwarmin is
- 11:16buying these existing assets at significant discounts to what it
- 11:20would cost to build a new one today.
- 11:21Right, what we call the replacement cost.
- 11:23If you buy below replacement cost, no rational competitor is
- 11:26going to build a new facility across the street.
- 11:28Because they could never charge enough rent to make a profit
- 11:31compared to your much lower cost basis.
- 11:33Now I get buying physical real estate when it's cheap and the
- 11:36demographics are on your side. But I also saw in the source
- 11:40notes that he buys a lot of distressed debt, like he's
- 11:44involved in a Brazilian corporate restructuring and
- 11:46various private equity debt exchanges.
- 11:48Yes, yeah. I have to play devil's advocate
- 11:50here. How is buying the debt of a
- 11:53failing company considered structural safety?
- 11:55That sounds like pure gambling. It definitely sounds
- 11:58counterintuitive until you understand the institutional
- 12:00mechanics of how debt is traded. OK, tell me more.
- 12:03Often, when a company faces a crisis and its bonds are
- 12:06downgraded to junk status, the original bondholders, like
- 12:09pension funds or conservative mutual funds, are legally
- 12:12required by their own internal mandates to sell those bonds
- 12:16immediately. Oh wow, so they aren't selling
- 12:18because the bond is necessarily worthless?
- 12:20Right, they are selling because the rules say they literally
- 12:23cannot hold junk rated debt. So they become forced sellers,
- 12:27dumping assets at any price just to get them off the Bush.
- 12:29Exactly that forced selling creates an artificial price
- 12:32crash. Then you have early distressed
- 12:35debt traders who swoop in, catch the falling knife, ride the
- 12:39initial dead cat, bounce, and then they also sell.
- 12:42Because they don't have the patience to wait three years for
- 12:44the actual business turn around. Precisely.
- 12:46And Klarman love stepping into that specific secondary vacuum.
- 12:50He operates right below the radar of the massive trillion
- 12:53dollar gorilla firms. In that like 50 to $100 million
- 12:57range. Yeah, where the forced selling
- 12:59creates irrational mispricing, he buys the senior distressed
- 13:03debt for pennies on the dollar, knowing the underlying assets of
- 13:06the company are worth far more than what he paid.
- 13:08It's. Brilliant building this fortress
- 13:10from the bottom up, finding cheap real estate and mispriced
- 13:14debt. But no matter how cheap a single
- 13:17asset is, it doesn't protect you if the entire global economy
- 13:20seizes up. No it doesn't.
- 13:21And. Klarman is staring at some
- 13:23massive macroeconomic storm clouds that he feels the broader
- 13:26market is willfully ignoring. He focuses heavily on the
- 13:29geopolitical map, particularly the Middle East.
- 13:31You know, the market treats regional conflicts as background
- 13:34noise, but Klarman looks at the physical choke points.
- 13:37He argues that if the Strait of Hormuz were to close due to an
- 13:40escalating conflict, the global oil supply would be checked off.
- 13:44Oil wouldn't just blip up, it could spike to $150.00 a barrel
- 13:48overnight. Which acts as a massive
- 13:50immediate tax on every consumer and business on earth.
- 13:53It would be totally devastating to the global economic
- 13:56landscape. And you combine that
- 13:58geopolitical risk with the physical reality of the AI build
- 14:02out we discussed earlier. Look, Rate the materials.
- 14:04The sheer demand for copper, steel, specialized cooling
- 14:08equipment, and like high voltage electricians to build out these
- 14:12data centers is massively inflationary.
- 14:14In the short term, it drives up the cost of raw materials and
- 14:17labor across the entire industrial economy.
- 14:19And hovering above all of this inflation and geopolitical risk
- 14:22is the ultimate elephant in the room.
- 14:25United States debt. Yeah, the US just crossed 100%
- 14:29debt to GDP and we are adding over $2 trillion a year to that
- 14:33pile. Now before we cover his
- 14:35observations on this, I want to make it absolutely clear to you
- 14:37listening to this deep dive, we do not take sides or endorse any
- 14:41political viewpoints, right? We are strictly and impartially
- 14:44reporting Seth Karman's perspective as stated in the
- 14:46source material. Exactly.
- 14:48And looking at it strictly through the lens of a global
- 14:50investor, Karman's perspective is highly critical of how the US
- 14:53is presenting itself to the world.
- 14:55In what way? Well, he feels that policy
- 14:57spanning multiple recent administrations are making the
- 14:59United States look, quote, unserious to the foreign
- 15:03creditors who actually purchase our debt.
- 15:05Unserious. That's a strong word for the
- 15:07bond market. It is.
- 15:08He specifically criticizes the erratic nature of trade policy,
- 15:12where tariffs are abruptly announced and then reversed,
- 15:14creating chaos for supply chains.
- 15:16OK, he critiques the strategy in the Middle East, arguing that
- 15:20executing bombing campaigns without securing lasting peace
- 15:23or demonstrating A strategic end game shows a dangerous lack of
- 15:26looking around corners. He also mentions diplomacy
- 15:29right, noting that the US frequently alienates top tier
- 15:32allies like Canada, whom honestly any other nation on
- 15:36Earth would draft as their number one choice for a
- 15:38neighbor. Yes, and regarding domestic
- 15:40fiscal policy, he critiques the political spectacle of proposing
- 15:44government cuts using what he calls a chainsaw.
- 15:47Like floating the idea of defunding FEMA or pandemic
- 15:50preparedness offices without making a clear reason to
- 15:53mathematical case for it. Right.
- 15:55To the foreign creditors buying trillions of dollars of U.S.
- 15:59Treasury bonds, he argues. This doesn't look like serious
- 16:02fiscal discipline. It looks like political theater.
- 16:05And it sends a glaring signal that the country cannot
- 16:08genuinely rein in its structural deficits.
- 16:12So if the national debt is spiraling, inflation is being
- 16:15propped up by the physical demands of AI, and the
- 16:18geopolitical map is a powder keg, what happens to the
- 16:22traditionally safe investments? Like where do you hide?
- 16:24This is perhaps Clemens most chilling observation.
- 16:27He says the risk free asset is riskier every day.
- 16:30Wow. For decades the entire financial
- 16:32system has been trained to ignore geopolitics and just
- 16:35watch the Federal Reserve. As long as the Fed was managing
- 16:37rates, U.S. Treasuries were at the bedrock,
- 16:40the risk free asset. But Klarman is warning that the
- 16:42sheer mathematical reality of these structural deficits will
- 16:45eventually force A reckoning. You cannot have the foundational
- 16:48asset of the global economy becoming inherently riskier
- 16:50without it causing some massive tremors across the entire stock
- 16:53and bond market. It is a sobering thought.
- 16:56If the foundation is shaking, you need those watertight
- 16:58compartments more than ever. But to humanize this legendary
- 17:04investor, before we wrap up, I do want to share the one that
- 17:07got away. Oh, the Palantir story.
- 17:09Yes, because even someone with a 44 year track record and an
- 17:13absolute obsession with risk experiences bad luck.
- 17:17In the interview, he's asked about his biggest mess.
- 17:20And 15 or 20 years ago, Klarman's team did all the
- 17:23fundamental work on a 40 to $50 million early venture investment
- 17:26in Palantir. They saw the underlying value
- 17:29long before the public markets did.
- 17:30I mean, Palantir was doing complex data integration for
- 17:33intelligence agencies, right? They saw the insanely sticky
- 17:36government contracts and the massive Moat the company was
- 17:39building. Valpo's had the capital and they
- 17:42were ready to make the bid. They were at the literal finish
- 17:44line, but the seller simply changed their mind at the last
- 17:47second and walked away. Oh, that hurts.
- 17:49By the time Carmen's team could act again, the window was
- 17:52permanently closed. Carmen admits he missed out on
- 17:55billions, if not 10s of billions of dollars on that single stake.
- 17:58It just proves that you can do all the math right, protect your
- 18:02downside, identify the ultimate asset, and still miss the boat
- 18:05because of sheer chance. But he survived the mess, and
- 18:09that is the real lesson here. His entire portfolio didn't
- 18:13depend on hitting that one Palantir home run.
- 18:15He missed out on billions, but his structural safety ensured he
- 18:18never lost the capital he already had.
- 18:21So it's a good point. And that leads to a really
- 18:23important, final, provocative thought for you to Mull over at
- 18:27the end of the interview. Klarman, who's in his 60s,
- 18:29manages 10s of billions of dollars and has literally
- 18:33nothing left to prove to anyone, says he plans to stick around
- 18:36and lead his firm for another 10 to 12 years.
- 18:38Yeah, if an investor whose entire framework is built on
- 18:41downside protection, structural safety and sniffing out disaster
- 18:45feels the absolute need to stay at the helm for the next decade,
- 18:48what does that say about the turbulent economic transition we
- 18:50are about to enter? That is the $1,000,000 question
- 18:53right there. The man who refuses to lose sees
- 18:56a storm coming and he is battening down the hatches.
- 18:59It really forces you to look at your own portfolio and ask if
- 19:02you are truly prepared for uncertainty or if you are just
- 19:05paying infinite multiples for a future no one can predict.
- 19:08Seriously. To access more advanced
- 19:10researches, applications and our latest published personal
- 19:13intelligent Multi Agent Deep Value Intelligence.
- 19:16Welcome to explore and join Deep Values dot AI community to track
- 19:21total market temperature. Welcome to search and download
- 19:23the iOS App Buffett Indicator from the Apple App Store.
- 19:26Disclaimer All analysis is based on publicly available sources
- 19:30and may be subject to revisions or differing interpretations.
- 19:33All contents from this channel is for information and research
- 19:35purposes only and does not constitute investment advice.
- 19:39Conduct your own due diligence before making investment
- 19:41decisions. If you'd like to access wire
- 19:43topics including life and investing philosophy, please
- 19:46subscribe to our sister channel Deep Values.
- 19:48Don't forget to like this video. Much appreciated.
- 19:51So as you navigate this AI driven market, ask yourself, are
- 19:54you just chasing the upside like everyone else or have you built
- 19:56your fortress? Keep your downside protected.
- 19:58We'll catch you on the next deep dive.