Cliff Sosin from CAS on Carvana and a bunch of other stuff $CVNA
Sosin believes Carvana now offers its most straightforward risk-adjusted setup yet: an enormous addressable market and margins well in excess of 2.5-3 times the industry’s on one side, a better-capitalized business able to trade price for volume on the other. His illustrative long-range math is deliberately huge—$3,000-$3,500 of net income per car across 20 million cars equals $60 billion in today’s dollars, perhaps 10-20 years out. More immediately, he argues that shocks might reduce profit or growth, but “they won’t lose money.”
Carvana’s nearly $2,000 of financing profit per vehicle is not simply an anomalous subprime gain: roughly two-thirds of its loans are prime, dealers ordinarily earn for originating loans, and vertical integration captures economics that otherwise accrue to finance companies. The genuinely debatable advantage is closer to a subset of roughly $800 per car. Sosin attributes that remainder to somewhat higher prime rates, lower vehicle prices and LTVs, higher subprime down payments, better cars, and possibly superior online underwriting—while stressing, “I’m not 100% sure that that’s the case.”
Alternative data helped Sosin understand Carvana’s business but did not make the stock, which Walker framed as traveling from roughly $300 to $3 and back toward $300, easy to trade. Daily sales contain weather, seasonality, and logistics noise, while knowing an earnings result does not reveal the market’s reaction. Its best use is slower validation—whether Carvana consistently offers the cheaper make-model-trim-year-mileage combination—not forecasting every weekly move: late 2022 instead became “water torture.”
Carvana’s 2022 collapse showed how a correct structural thesis can coexist with worsening operating facts that are difficult to interpret in real time. Omicron stretched delivery times from roughly three or three-and-a-half days to seven and cut displayed inventory from about 90% to half, initially explaining weak conversion; when logistics recovered, underlying demand still did not. By early 2023, falling headcount, stabilizing units, a visible delivery line, and the banking crisis’s support for a strong supposition that lending markets were no longer firming rapidly and would normalize finally suggested that the setup had turned.
Walker’s hardest downside cases—a 30% used-car contraction, a compelling $20,000-$25,000 EV, and autonomous fleets replacing ownership—do not strike Sosin as existential, though sustained vehicle-price deflation would shrink Carvana’s market. His cyclical stress case uses estimated price elasticity of negative 7 to negative 8: a 4% price reduction could offset a 30% demand decline, with perhaps six points of total price concession against an 11%-plus margin. For robotaxis, 30%-50% empty miles, wait times, storage, personalization, and heterogeneous consumer economics undermine the vision of “the same golf carts.”
Sosin treats concentration as a continuing sell decision, not a mechanical portfolio rule: trimming every winner above 30% guarantees that an investor never owns a truly transformative compounder at full weight. His warning is Stanford’s sale of Google around its IPO, but Walker’s pushback matters—Google, Amazon, Facebook, and Apple developed profit engines early buyers could not foresee. Walker also raises Walmart, Costco, and Home Depot; Sosin responds that robotics, self-driving trucks, and scale could keep strengthening Carvana’s core system.
His Herbalife exit illustrates how updating the reference period and monitoring a named tail risk can matter more than being right about the original controversy. Sosin still considers Bill Ackman’s pyramid-scheme claim “entirely incorrect,” but years of sideways performance displaced decades of teens growth as the relevant base rate; then effective GLP-1 weight-loss drugs triggered the exact risk he had identified. His process remains intentionally open-ended: let experts talk, “shut the fuck up,” and learn one brick at a time. Walker’s trite Cliff saying is to wait for the stocks to go up.
1. Public visibility made Carvana harder to own, but worth memorializing
Sosin describes investment publicity as “sort of a one-way ratchet”: followers can impede buying and selling, create an expectation that every move be explained, and form “a big crowd” ready to celebrate failure. He had therefore decided to stop appearing publicly.
Carvana’s near-death and recovery felt too unusual to let “disappear gradually into the mists of time.” After reaching out to Patrick, Sosin returned to Walker because of their five-year friendship and the chance to address finer points—“the details of retail margin”—for a more specialized audience.
2. Prime lending distributes capital; non-prime lending manufactures information
Sosin divides consumer lending into two different economic activities. Prime lenders identify borrowers with established repayment reputations, then compete largely through distribution; thinner margins and apparently safer loans encourage greater leverage, leaving the lender effectively “selling disaster insurance” against synchronized unemployment or other shocks.
Non-prime lenders search for “the good borrowers amongst the ones that were thrown out by the prime market.” They add value through underwriting information and behavior modification—engaging customers in ways that increase repayment—rather than merely performing distribution.
The higher rate is compensation for genuine work and risk, not necessarily predation. Many borrowers live with income roughly matching expenses, so an interruption or unexpected bill makes credit important; a lender that identifies the repayable customer can become “a valuable sort of partner” and improve that borrower’s life.
That combination of human behavior, finance, and banking created Sosin’s circle of competence. Complexity and historical blowups keep generalists away, while repeated study can reveal businesses whose economics are better than the market’s broad treatment of “subprime” implies.
3. Subprime blowups begin when repayment signals become self-reinforcing illusions
Sosin’s core warning is that lending is an information business, so “false or misleading signals can really screw you up.” In the late 1990s, non-prime auto and unsecured borrowers refinanced from one lender to another; each lender saw an apparently performing loan, while the industry collectively passed borrowers around “like a hot potato.”
Reported performance attracted more capital, which funded the refinancing that created the reported performance. The system “works great until it doesn’t”: once capital stopped expanding, lenders discovered that borrowers had been rolling debts rather than repaying them from sustainable cash flow.
Housing from roughly 2004-2007 repeated the mechanism through collateral. Rising prices appeared to validate borrowers, encouraged more lending, financed additional purchases, and drove prices still higher. Sosin’s general lesson is to identify the feedback loop that may be manufacturing the very evidence used to justify underwriting.
4. Carvana’s financing profit decomposes into ordinary and differentiated pieces
Carvana generates nearly $2,000 of financing profit per retail unit, but Sosin resists treating all of it as evidence of exotic subprime economics. Ordinary dealerships are paid to originate loans, while CarMax retains loans and earns the economics over time; selling those loans would produce a recognizable gain on sale.
Carvana does over-index to non-prime, yet roughly two-thirds of its originations are prime, depending on the date and definition. In prime lending it earns more than CarMax largely because it charges somewhat higher interest rates, although Sosin says lower vehicle prices can still make the consumer’s total transaction better.
The non-prime loans in CarMax’s business still get made—by a finance company such as Westlake, Sosin says. Carvana instead integrates that finance-company layer and captures economics that ordinarily sit outside the dealership.
ABS data provide a reality check: interest income less expected charge-offs, servicing expense, and funding costs leaves excess profit. With an average pool life of about two years, Sosin says Carvana’s roughly 9% non-prime gain on sale is supportable when those cash flows are capitalized on a discounted basis.
5. Better cars and richer digital signals might explain the remaining edge
Walker narrows the dispute: if an ordinary dealer might earn $1,200-$1,300 against Carvana’s nearly $2,000, the roughly $800 gap is material to a thesis built on superior scale and execution. Sosin narrows it further—the portion attributable specifically to better non-prime performance is only a subset of that difference.
Sosin’s operational explanation starts with vehicle quality. He says the primary reason a subprime borrower defaults is that the car breaks down; Carvana’s cars should fail less often, while lower prices, lower LTVs, and a higher down-payment mix reduce loss given default.
Online origination may also permit more verification with less friction. Every lender balances confidence against adverse selection: onerous requests drive good borrowers elsewhere, but Carvana can embed verification “in a click” inside an already streamlined purchase, financing, and delivery journey.
He offers illustrative—not specifically Carvana—digital signals: applicants with little phone battery have historically underperformed those with more, while Internet Explorer users once proved worse credits than people who downloaded Chrome. Such correlations might enrich underwriting, but Sosin preserves the hedge: “I’m not 100% sure” online data explain Carvana’s advantage.
6. Alternative data validates the flywheel better than it predicts the stock
Carvana lends itself to unusually detailed measurement, and Sosin hired a consulting firm to build more of it. Yet “this week’s sales are light” says almost nothing definitive about next week; weather, subtle seasonality, inventory availability, and delivery constraints introduce meaningful high-frequency variation.
His preferred analysis asks whether a customer seeking a precise make, model, trim, year, and mileage can find it at Carvana or CarMax, then compares total price in a randomly selected market after shipping fees and delivery time. Carvana wins “the vast majority of the time,” though the result moves as both companies change prices.
That evidence tests the structural proposition—selection, price, and convenience—not whether a quarterly estimate will move shares. Sosin recalls a blog about someone who hacked a law firm and stole earnings releases yet got only about 65% of the trades right: knowing the result is not the same as predicting the reaction.
Carvana demonstrated the distinction in 2024. Sosin felt prepared for several earnings beats, but similarly strong quarters produced opposite stock moves: Q3 rose and Q4 fell. Alternative data can make volatility easier to endure by keeping the investor connected to the business; it cannot make “life easy.”
7. Omicron obscured a demand collapse before operational data revealed recovery
Early in 2022, Omicron disrupted Carvana’s logistics. Average delivery lead times expanded from roughly three or three-and-a-half days to seven, and customers saw about half the inventory rather than the roughly 90% available in a healthy system; weak conversion therefore had an obvious temporary explanation.
The difficulty was that Omicron also masked weakening underlying demand. Logistics improved, but sales did not rebound; competitors charged interest rates Sosin believed were uneconomic, yet he could not know when that behavior would end. Selling because an unsustainable distortion persists and buying because it must eventually disappear can both sound rational.
By late 2022, sales weakened almost every week. The information advantage became “water torture”: instead of suffering one bad quarterly release, Sosin received another daily drip showing that the company genuinely was not performing well.
The 2023 turn appeared through several imperfect signals together. Headcount was falling, units finally stabilized, and a delivery line had emerged; meanwhile, the regional-banking crisis supported a strong supposition—before the data arrived—that lending markets were no longer firming rapidly and would normalize. Sosin saw the setup and told Walker, but a poison pill prevented him from buying more.
8. Carvana’s long runway supports an extreme—but explicitly distant—upside case
Some 40-some-odd million cars are sold annually in the United States, and Sosin considers most of that market addressable, with new cars offering another potential avenue. Scale improves the experience, and “there’s nothing like it,” so he expects Carvana eventually to sell “many, many millions” of vehicles.
Margins are already well in excess of 2.5-3 times those of the industry despite rapid growth. Fixed-cost leverage remains available, as do granular improvements such as deciding which parts to replace during reconditioning by balancing immediate cost against warranty and vehicle-service-contract expense.
His illustrative destination is $3,000-$3,500 of net income per vehicle multiplied by 20 million units, or $60 billion in today’s dollars. Sosin places that outcome perhaps 10, 15, or 20 years away and emphasizes the cash generated along the path; it is a framework for magnitude, not a near-term forecast.
Walker’s challenge is that famous “never sell” winners—Google, Amazon, Facebook, and Apple—developed major businesses their original investors could not anticipate. He also raises Walmart, Costco, and Home Depot. Sosin responds that technology, including robotics and self-driving trucks, is working in Carvana’s direction and could make its advantage over dealerships bigger.
9. Margin advantage now gives Carvana room to absorb an industry shock
Sosin says Carvana’s resilience is “far greater than it’s ever been.” Competitors generally cannot sustain losses indefinitely, while Carvana’s higher margins let it choose how to trade off price and volume; an evenly apportioned industry shock should hurt results without recreating the old solvency risk.
Used-car demand is more stable than many assume: 2022’s roughly 20% contraction was the worst decline on record, while the Great Recession produced a decline in the teens. Walker nevertheless asks Sosin to stress a 30% fall combined with reverse fixed-cost leverage.
Sosin estimates Carvana’s price elasticity at perhaps negative 7 to negative 8, while admitting it is uncertain. In that framework, a 4% price cut could recover roughly 30% lost demand; allowing another couple of percentage points for industry price compression produces about six points of concession.
Against an expected margin of “11 and change” percent for the year, that scenario leaves Carvana near the average competitor’s present margin. It would be unpleasant, but competitors would simultaneously be “losing lots of money and disappearing at breakneck speed,” limiting how long the adverse pricing environment could persist.
10. Cheap EVs threaten long-run market size more than today’s inventory
Walker imagines a compelling $20,000-$25,000 EV making Carvana’s gasoline inventory obsolete and leaving too little value for used-car intermediation. Sosin separates an immediate inventory shock from sustained real-price deflation and considers the overnight stranding scenario implausible.
The United States has roughly 300 million vehicles, and even a spectacular manufacturer cannot replace a meaningful share quickly. A superior low-cost EV might lower expectations and used prices at the margin, but production constraints would make the transition unfold over many years.
Carvana holds only about $4,000 of inventory for every annual vehicle sold because it turns inventory roughly six times. A 10% price shock therefore costs about $400 per annual unit, versus roughly $4,500 of incremental margin and prospective per-unit EBITDA in the mid-$3,000s, eventually approaching $4,000.
Sosin concedes the limiting case: if new cars fell to a few thousand dollars and became disposable, “that would be bad for Carvana.” A future with $30,000 new cars would merely reduce industry size at the margin, partly offset by households owning more vehicles or replacing them more frequently.
11. Autonomous driving may strengthen personal ownership instead of destroying it
The standard robotaxi case says idle privately owned cars incur depreciation and capital costs, so shared autonomous fleets should reduce cost per mile. Sosin’s rebuttal is deadhead mileage: Uber, taxis, and even long-haul trucking can run 30%-50% of miles empty because vehicles must reach passengers and reverse directional commuter flows.
Empty driving is more expensive than stationary ownership. Once deadhead miles, cleaning, payments, and fleet overhead enter the calculation, pooled autonomy may be slightly cheaper or slightly more expensive, but the supposedly overwhelming cost advantage “largely disappear[s].”
Average economics also hide customer heterogeneity. A price-sensitive driver can buy an eight- or ten-year-old Toyota with minimal depreciation and capital cost; a new BMW 7 Series buyer knowingly pays more for quality. A shared fleet cannot simultaneously beat the Toyota on cost and satisfy the BMW buyer’s preferences.
Waiting several minutes for a 15-20 minute journey carries meaningful time and planning costs, while cars store child seats, toys, tools, gym bags, and shopping. Self-driving may turn the car into “a personal room,” strengthening ownership; it could even reduce Uber usage when one’s own car can handle drinking, parking, or an airport drop-off and then drive itself home.
12. Selling, interviewing, and idea generation all require resisting premature closure
Sosin uses Stanford’s sale of Google near its IPO to argue that sell decisions deserve the same work as buys. Mechanical trimming may reduce risk, but anyone who automatically diversifies above 30% can never experience a Berkshire-scale winner; Carvana’s concentration therefore remains an underwriting judgment, not a rule.
Herbalife shows the opposite decision. Sosin bought after Bill Ackman’s pyramid-scheme attack, which he calls “wildly” and “entirely incorrect,” but by 2021 the sideways 2014-2021 record deserved more weight than decades of teens growth. His unproven explanation is that gig work weakened MLM recruitment, where losing 2% of distributors compounds through the network.
Effective GLP-1 weight-loss results then activated a tail risk Sosin had explicitly identified years earlier, so he sold Herbalife—while ruefully failing to buy Novo Nordisk or Eli Lilly. Experiences with Celanese and Ashland similarly increased his respect for forecasting difficulty and the need for valuation’s margin of safety.
His expert-call method follows Robert Caro: ask open-ended questions, write down follow-ups, and write “STFU” in his notepad while the source keeps talking. The same openness governs research—health insurance, life sciences made more approachable by AI, or even a mistakenly downloaded 10-K. “One brick at a time,” the objective is simply to learn something every day while “waiting around for my stocks to go up.”
Full transcript
You're about to listen to the Yet Another Value podcast. Today's episode is Cliff Sosin. It is episode 310. Cliff was one of the first five people on the podcast. He returns for his second appearance. It's a super interesting conversation about Carvana and a whole lot of other stuff on investing. Before we get there, a word from our sponsor AlphaSense and then Cliff Sosin.
This podcast is brought to you by AlphaSense. Uh those of you who've been following the podcast and the blog know that I've been doing a lot of work recently on shareholder engagement, particularly at these busted biotechs, you know, I've done recently a podcast on Sage and Kos. You guys can go and look in the show notes. I've posted a lot on the blog about busted biotech, why this time is different, all of these companies that are trading at enormous discounts of cash. So given my focus on corporate governance and shareholder engagement, I worked with AlphaSense and they said, "Hey, let's do a free webinar. we'll get you in touch with the corporate governance experts. Let's do a free webinar and kind of bring some uh bring some information and get you a little smart on it. So, we did the free webinar. It was really excellent. It was with a professor at the University of Chicago who actually teaches corporate governance, has been on a 100 boards, including almost 20 public company boards. We had very differing views, but it was really interesting to get his insights as someone who's been there, his insights and kind of his differentiation from how I'm viewing corporate governance. So, you can go check out that webinar. I'll include a link in the show notes or you can just go to alpha-sense.comyavvp to get a free trial and kind of show them thank you for the support of this podcast.
All right. Hello and welcome to the yet another value podcast. I'm your host Andrew Walker with me today. I'm happy to have on for the second time my friend Cliff Sosin. Cliff, how's it going?
Very well. Thank you for having me.
But look, super excited to have you. Uh before we get started, disclaimer remind everyone nothing on this podcast in investing advice. Uh there's a longer disclaimer at the end. everyone can hop into it. So, Cliff, look, I'm super excited to have you back on. You were one of the first podcast guests, back when I literally had no clue what I was doing—a fog-induced COVID podcast. It’s been 5 years, so I’m happy to have you on for the second time. I’m going to send you an invite for the 2030 pod at the end of this thing. We’re talking about all sorts of stuff today: Carvana, anything. Where should we start?
Well, first, it might be interesting to talk about why I came back after all these years.
Go ahead. Is it because I’m so handsome?
It’s actually because I like you so much, believe it or not.
The history here, in case people don’t know, is that over the last 5 years, Andrew and I have chit-chatted all the time. I count him as a friend, and I make fun of him mercilessly for being incredibly slender and unable to lift anything heavy.
There was a bit of public exposure that I created for myself 3 or 4 years ago, and it’s sort of a 1-way ratchet. As you do it, it feels good, but what I learned was that it’s really not helpful in a variety of ways. Having people follow you into ideas makes it harder to buy things and makes it harder to sell things. You suddenly feel like you have to explain yourself.
Also, when things go wrong, it really sucks to have a big crowd rooting for your demise. So, I deliberately decided that I wasn’t going to do any more of these. That’s what I kept telling you every time you would ask.
But then the whole Carvana saga unfolded, and it was such a crazy thing. I could feel it gradually disappearing into the mists of time, and I wanted to memorialize it. So, I reached out to my good friend and very successful podcast host with a big audience, Patrick.
But I couldn’t do Patrick’s podcast and then, after all of our relationship, not do yours. So, here I am. My hope is that we’ll find great stuff to talk about. This is a different audience, one where we can get into the finer points of the details of retail margin or something, and so it’ll be exciting.
Look, I was there in college, and I’ve never had a problem with sloppy seconds. So, I’m completely okay with it.
You did—I listened as prep and read a bunch of things. You did a podcast on Invest Like the Best with Patrick. People can listen to that for the whole Carvana story. I’ve got some unique takes on it and everything, but let me start with a broader question.
People can go to CAS Investment Partners and review your 13F over the past, call it, 13 years or whatever, since you’ve been following 13Fs. When I look at your 13F history, I’d say somewhere between 1/3 and 1/2 of your investments—and there are not a lot of them; you run a very concentrated portfolio—in some way relate to 1 of 2 things: securitizations, such as Carvana, with a lot of securitizations and auto loans, that type of stuff; or subprime lending.
I think about World Acceptance and Credit Acceptance, which I don’t believe you’re long now, but you’ve been historically. Capital One, which I think you are long now. There are others.
When I look at that, is there something in your skill set that screams securitizations and subprime loans are right up your alley? Or do you think there’s a systemic mispricing in those types of opportunities?
Good question. Maybe a bit of both. I certainly spent a lot of time thinking about subprime lending over—gosh, I first started thinking about it back in the financial crisis, when you were worried that you would be a subprime borrower.
No, that’s what I meant.
Yeah, yeah, yeah. I’m old now. I’m dating myself, but this is back in 2007 and 2008. That was the first time I was thinking about it.
I’ve sort of been around it for a long time, and I think it’s a fascinating industry to study. It’s distinct from—one way to frame the world is that you can broadly divide the role of lending. You’re trying to give loans to people who deserve them.
There’s a broad swath of people we call prime. They basically have a reputation that they pay their debts, which is what you could call a FICO score, credit score, or something like that. Lending to them is fairly straightforward. You identify that they’re prime, and that means they’re likely to borrow only what they can afford to pay back and make every effort to pay it back.
Barring loss of a job, death, divorce, or disease, they will pay it back. That business generally is about distribution. It operates on thinner margins and tends to run with more leverage because the loans are, in many ways, safer than loans that aren’t prime, and people certainly perceive them as such.
Those businesses, by virtue of being easier to do, are actually in many ways more competitive. By virtue of the competition, people tend to run them with more leverage. In some sense, you’re selling disaster insurance: when there’s a recession, these prime borrowers lose their jobs at the same time, and you’re stuck.
Non-prime lending is different in that you bring information to markets and find the good borrowers among the ones who were thrown out by the prime market. You also bring behavior modification to the market. You engage with borrowers in a way that makes them more likely to repay.
In doing that, you actually produce a lot more value by finding the deserving borrowers among the many. You get paid for that because you’re able to charge a much higher rate. These businesses can have very good economics.
They come with some risk, and there are a variety of types of that risk. One is that, in general, it’s an information business. Capital One would talk about itself as an information business. You’re sorting the good borrowers from the bad, and it’s a process of figuring out a mechanism to do that.
False or misleading signals can really screw you up. If you look at the big blowups in subprime lending—the 1998 blowup in auto and card lending, and the 2008 financial crisis—these were situations where the signals were screwed up.
The way that happened was that, in the 1990s, the non-prime auto lending market and the unsecured lending markets were growing, and people were essentially refinancing their debt from 1 lender with another lender. If this were just 1 lender constantly rolling borrowers into more and more debt, it would obviously be able to see that the borrowers weren’t actually able to pay.
But in this case, because each borrower was being handed off like a hot potato from 1 lender to another, it looked to each lender like its borrowers were performing.
But in fact, they weren't. They were just rolling their debts, and as long as the industry grew, that meant the lenders had great returns because the borrowers were performing. That drew capital to the business, which in turn provided the money to roll the borrowers. You can see how this doesn't work great until it doesn't.
And that's exactly what happened in 2008. What happened was, from 2004 to 2007, the same thing but with house prices. As house prices rose, that created a false signal that these borrowers were repaying, which in turn created incentives for people to lend to them. That made it possible for them to buy houses, which in turn helped drive up home prices and created the signal.
If you get these feedback loops, it can be very dangerous. But there are a lot of really good businesses in the mix. If you find a way to build relationships with borrowers where you've found the good borrowers amongst the bad ones, established a relationship with them, and they're performing, that can be very valuable.
You're providing them loans, and these people are often operating where their income is very roughly equal to their expenses. If there are disruptions to their income or unexpected expenses, they need to borrow, and the ability to do that is very important. As long as you've underwritten that correctly, you can be a valuable partner for them, get paid for it, make good money, and make their lives better. I thought it was worth saying: I think it's a very noble business.
My sense is that the blowups of the past in different sectors and the complexity of it have generally kept people out. There are a lot of things you have to be cautious of, but it's a space I spend a lot of time in, and it's also just interesting. It's human behavior, finance, and banking. There's a lot of things that come together.
But yeah, you're right. I've found myself in that space in part; it's one of these things where you build a circle of competence and keep growing around it. It probably is an area where I have a bit of differentiation.
I think what I heard from you there is that the market paints all subprime lenders with a broad brush, and you think there are some hidden gems in there based on everything you just laid out. There's alpha to be generated by the companies if they can figure out a way to lend to someone whom the market paints as subprime but who might be slightly better than subprime, and you think you've got the skill set to figure them out.
Let's bring it to Carvana. I did the podcast with Patrick. I did the podcast earlier with Recurve. A frequent point of bear contention over Carvana is the lending stats, right? They always say they're going to blow up, and I think you guys say, "Hey, look at the stats. Their ABS is actually better than other subprime lenders."
What is Carvana's unique niche? I understand everything about the Carvana cycle. I understand a lot of the pieces of the Carvana flywheel. I don't understand why their borrowers would be particularly better than your average subprime borrower. I actually think it would be worse because you're buying over the internet versus going in person and getting that old, crisp handshake when you buy a car in person.
Yeah, there's a lot there. First, Carvana's whole financing business generates a couple thousand dollars of profit per unit. It's nearly $2,000 of profit per unit.
When you say nearly $2,000, are you saying that in a dismissive way, like it's not a lot, or are you saying that in an extremely important way?
I'm going to start by describing that there's a prime mix and a non-prime mix, and there's a part that all dealers get, and so forth. All dealers generally get paid to originate loans. You can look at the financials for any car dealership, and they get paid to originate loans.
A lot of the profits in Carvana's financing business are more similar in character to the typical gain on sale you'd get if you look at CarMax. Now, CarMax doesn't sell the loans; they hold them and collect the money. But if CarMax were to sell its loans, they would get a gain on sale, and that gain on sale is similar in character to what Carvana gets.
I frequently hear Carvana skeptics say Carvana is a subprime originator in an online car-sales platform. It's really a subprime-originating business.
Well, it is, but it's also a prime-originating business. Carvana does over-index in subprime originations, but depending on where you draw the line between subprime and prime, and depending on when exactly you look, something on the order of two-thirds of Carvana's loans are prime. I'm just pointing that out to level-set.
In the prime business, Carvana does make more money than CarMax. Some of that—and a lot of that, actually—is because they charge a bit higher rates. That's basically the bulk of it. There are other things too, maybe on the margin, but that's the big piece.
There's a piece of Carvana's business that's a lending business: the non-prime originations. In the case of CarMax, those loans still get made; they just get made by Westlake, I think. So that's not totally unique, because Carvana is vertically integrated into doing that, whereas most dealers are not.
You can get the performance data for Carvana's non-prime ABS issuance from KBRA or whatnot. You can see what they earn in interest, what the charge-off expectations are that Kroll has for those pools of loans, what they pay in servicing, and what the cost of funds are that they borrow at. The residual of that is excess profits, and you can see how much that is.
If you know that the average length of these pools is about 2 years, you can capitalize that on a discounted basis and work out that the roughly 9% gain on sale that Carvana gets on these loans isn't out of line with the economics of the loans.
The reason I keep pointing this out is that we're getting the point of debate down to a smaller and smaller piece of the total pie. We started with $2,000 profit per loan. We probably reduced that to roughly $800, and now we're talking about some subset of that $800, which is Carvana's ability to outperform the industry in generating these loans.
The whole Carvana thesis is that, because of its scale, because it's online, and because of its business model, it's so much better than your local used auto dealership. So it's going to literally—and it has been, with 2022 excluded—eat the market, right? It's a superior value proposition.
Part of that is, yes, I agree with you. The $2,000 per loan, I think if you went to the used auto dealership across the street, it'd be $1,200 or $1,300, but that $800 difference is material. That's more than 50% better, actually. I'm wondering why that's the one place where it jumps out to me. I'm not sure why Carvana should be literally almost 50% better than the dealership.
Yeah. So let me answer that in brass tacks. For one, Carvana is vertically integrated into the whole lending stack. Dealerships are typically selling the loan, and then there's a finance company that's making money there.
There's also a lot of efficiency created by being vertically integrated. Think about the loans that the finance company evaluates but doesn't actually originate because they go to a different finance company. The other one is that, in the prime part of the business, Carvana does charge a bit more. Their prices are lower, so it's still a better deal for consumers all in, but that's worth pointing out.
In the non-prime piece, I do think the loans perform somewhat better in the end. Some of that is going to be better underwriting, from the fact that Carvana is just very good at this, and they have a lot more data available to you when you're originating loans on the internet versus when you're originating them in a store.
A lot of it is also going to be that Carvana's cars are in great shape and somewhat lower priced, and the experience is really great. The primary reason people default on their subprime loans is that the car breaks down.
So that’s less common with a Carvana car. Once someone does default, obviously, the lower LTVs on Carvana’s loans mean that the loss given default is lower. That’s a function of Carvana having lower prices, as well as a higher down-payment mix in their subprime book.
Can I go back to one thing you said? Why does originating a loan online have better statistics than originating a loan in person?
Well, to be honest, I’m not 100% sure of that. I think that’s true, and I think it’s just because you get a lot more information about people when you do something online. I’ll give you a concrete example.
In originations, one of the things that you’re always trying to balance is the amount of verification you make a borrower go through. The greater the amount of verification, the more certain you are about the features of the loan, but the more adverse selection you’re going to face, in that the good borrowers will just borrow elsewhere.
Yep.
Because Carvana has this really sleek experience and everything is really efficient, and because they’re already doing the whole loan transaction, they’re able to get more verification done more easily, in a click, as part of the whole transaction, than other lenders might be able to do. They can do that more conveniently, so I would imagine that would be the sort of thing on the margin.
There are also great examples. I don’t think this is a Carvana example, but in general, in the subprime lending industry, if you were to take loan applicants and examine the amount of battery life left on their phones when they apply, you would find that the ones with little battery underperform the ones with a lot of battery. Do you want to know why?
I’ve heard stuff like this before. Do you want to know why this is concerning to me? Every time I look at my wife’s phone, it has 4% battery, and I’m like, “If my phone goes below 60%, we’ve got to charge this thing. We’ve got to get it full.” So I know I’m trustworthy. I’m a little worried about her.
Right—about people’s sense of responsibility and their discomfort with the risk that things could go wrong. There are a lot of these things that are correlated. There was a long time—I don’t know if it’s still the case—when people who used Internet Explorer had worse credit than people who used Chrome. Why? Because you had to download Chrome. It was the sort of person who knew about this and who bothered to do it.
There are many little things like that where Carvana’s knowledge about your whole journey, and their vertically integrated stack, should give them more information to make better underwriting decisions. In fairness, that’s a tough thing to assess, and I’m not 100% sure that’s the case, but I’m giving you a lot of reasons why I think it could be.
Collectively, between price and underwriting, and in general, when you give non-prime borrowers a prime experience—which is what Carvana really does—you get positively selected for. Consumers are more likely to stay with the transaction if you give them a great car at a great price and treat them well than if you don’t.
I think all of those things play a role. In addition, of course, they’re vertically integrated, and that helps. Some of it is just what you get paid for doing the work.
Let me switch to a completely different track. I’ll give you a couple of little puff pieces during this. I had trouble prepping for this episode, and one of the reasons is that when I was prepping, I was going over our conversations. Every now and then, when you and I would talk on the phone, I’d take notes on Carvana, reread your letters, or listen to Patrick.
In hindsight, it sounds so easy. For those who don’t know, Cliff basically rode Carvana from $300 to $3 and back to—let’s round it up—$300 right now, though I’m sure you would prefer the actual $300 rather than the rounded $300 right now. It was hard because, in hindsight, I look at my notes and your letters and think, “Damn, it seems so easy. It was so obvious.” I know it was not obvious or easy at the time, but one of the things that jumped out at me is that when I listen to the Patrick podcast or read your letters, I see a lot of alternative data about how Carvana is performing intra-quarter.
I’ve heard from retail investors, some of whom have made a fortune on Carvana, and one literal quote I’ve heard is, “The easiest money I’ve ever made is trading Carvana alt data.” I guess my question is this: We’re taping this on May 6, and tomorrow is May 7, when Carvana reports earnings. If you want to spoil the earnings, we can, but when I see all this alt data on Carvana that people are using to talk about it, how was the easy money, and how are you using alt data with Carvana?
Why was there this easy money in the alt data? A lot of the people I’ve heard say this rode Carvana from $200 to $3 to $300. I’m thinking, “Why didn’t you avoid the drop from $200 to $3 with alt data and then take the move from $3 to $300?” I think you were a Carvana bull, and congratulations—you were absolutely correct and made a lot of money. But don’t tell me it was easy money trading the alt data when you rode it down 95%. The alt data should have helped you avoid that a little bit. Does any of that make sense?
Yeah. It’s interesting. There’s a lot of alt data people can buy. I have a team—a consulting firm that I’ve hired—that’s built a lot more. They may be coming on the podcast at some point in the near future. I’ll send this particular clip to them at some point.
I have a lot of visibility. Carvana lends itself particularly well to this sort of analysis. But it turns out, first, I think other people have a lot of this information, and second, it’s always very difficult. You know this: This week’s sales are light, but what does that tell you about next week’s sales? It’s sort of like nothing.
If the company is going to grow to eventually sell all the cars, then the fact that this week’s sales are a little off is sort of immaterial in the scheme of things. It’s not obvious that this week’s sales being off is predictive of next week’s sales being off. In many cases, there’s a meaningful amount of variation in high-frequency data because of weather and all kinds of nuanced seasonality that you discover when you start really stressing over everyday sales.
I find that the most useful thing about the alt data is really to step back and validate the broader hypothesis. We did an analysis asking the question: Suppose you’re looking for a particular make, model, trim, year, and mileage. What are the odds that Carvana could have it?
Let’s say that you find it on Carvana and CarMax, and you live in a randomly selected MSA. Once I factor in shipping fees and such, if I need the car within 2, 3, 4, or 5 days—whatever the number is—which one provides me with a cheaper all-in solution? You can imagine doing that analysis, tracking it over time, and producing metrics. The answer is that Carvana varies over time, obviously, as Carvana moves its prices around and CarMax does the same, but Carvana wins the vast majority of the time.
That’s the sort of analysis you can do. But the experience during 2022, just to go back to the alt-data thing, was that sales were weak at the beginning of 2022 for the very good reason that the Omicron part of the pandemic had screwed up their whole logistics system.
Their delivery lead times, which are about 3 days now on average—3.5 days—and which reflect a pretty healthy system, were about 7 days, which is super long. They were only showing people about half of the inventory, whereas now that’s about 90%, and that’s, again, a healthy system. Unsurprisingly, they had very low conversions.
That was the explanation: Once Omicron goes away, this will get better. But it turned out that Omicron was also masking a decline in underlying demand. As they fixed their logistics system, demand didn’t bounce back. By this point, of course, the stock had gone down a lot, and you were like, “Okay, well, now demand is clearly softer. They’ve got to work through this issue or whatever—something of a lost year.”
But then what happened was that every week, demand was a little lower. I could see why—not exactly why, because you never really know exactly why—but I had a bunch of reasons that made sense, and you could back-test them. We could see that the interest rates the competitors were charging made no sense.
But we could track that. Then you're like, well, sure, but that should go away, right? So how do you trade that? Do you sell your stock because something that shouldn't be happening, that can't last, is screwing things up? Or do you do nothing, or do you buy more on the theory that the thing that's been around for a long time, that should never have been there in the first place, is going to go away?
And so you sort of find yourself—what ends up happening is you end up getting some resolution, but then, in terms of the short term, the next thing down is actually sort of harder to predict. You also get weird stuff, like the company beat a number of times in 2024, and each time I expected it or whatever. We should have had a reasonable sense of what earnings would be, but there's this difficult thing of knowing, okay, great, that's likely to mean the company's doing really well, but how is the stock going to react?
I remember reading a blog about someone who hacked a law firm and stole all the earnings releases, and they were trying to trade on that. I think their trading history was that they got 65% or something of their trades right. It turns out that even if you have a reasonably good sense of what earnings are going to be because you've done all this analysis, it's not obvious necessarily how the stock's going to react. Their Q4 results were quite good and the stock went down. Their Q3 results were also quite good and the stock went up.
So it's quite tricky. I don't buy this idea that alt data makes life easy. If anything, when stocks go crazy and there's no connection to the underlying data, it makes living with it easier. You feel more connected to the business, you understand that things are fine, and you can just deal with it. If you have some extra money around, you can buy more.
Where it's tricky is when, in late 2022, the company really wasn't doing well. The sales were weak, and they kept getting weaker. It was water torture, because instead of having to suffer through just 1 bad quarter, 1 bad release a quarter, I had to deal with another drip of weak sales every day.
So anyway, that's kind of a long answer.
I think the impressive thing with Carvana is—again, it was hard for me to prep for this podcast because I had you telling me at $20, at $40, at $50, and the stock today is at $200 except it has an extra digit in there. Carvana was a great opportunity, but on the way up, actually, the story there is that I couldn't buy more.
Yeah, and on the way up, because of the poison pill, by the way, I couldn't buy more. But 1 thing you can look at is the average delivery times. You can track them, and in general they trend down as the company gets more and more efficient. There's all kinds of noise in them, but think about that as the line to buy a car.
Each period, the company builds to a certain amount of expected demand for each month or each week, and then demand happens. If it's more than expected, 1 of the things that happens is that the lines get a little longer, because people take up the delivery dates that are closer and are pushed to the delivery dates that are further out. Conversely, if demand's a little soft, the line gets a little shorter because of the opposite effect.
What you could see in 2023 was that headcount was falling and units had finally stabilized, although we were always unsure whether they could destabilize the next week. For the first time, it looked like there was basically a line. In addition to that, the banking crisis of early 2023 meant that we didn't have data on it yet, but it was a strong supposition that the lending markets were no longer firming really fast and were going to normalize.
That was why it was set up at that point. I couldn't do anything about it, but I did tell you, and then you didn't do anything about it either, but that's okay.
No, I was going to say it was hard for me to prep because I was thinking about that the whole time. The thing I'd give you flowers for is that this was a company that tapped its ATM—which I hate ATM tapping—and was down 99%. The base rate for anything tapping its ATM or going down 99% is not exactly great, but you had done all the work. You had the conviction, and those are the 2 most important things.
I talk all the time to people and say, "Investing is not an idea game of original ideas, for the most part. You can go steal someone's ideas in half a second. It's generally the ability to develop conviction." Sometimes, when a stock goes down, that conviction lets you know, "Hey, I need to buy more of this," or, "The thesis is completely broken. I need to get the fudge out of this."
So let me ask you a different question. People can go look at your 13F. You've got a huge concentration in Carvana. What keeps you up the most at night right now with Carvana?
Yeah. It's interesting. I sort of think that, from a risk-adjusted return perspective, you could make the case—and I have made the case—that Carvana has never been a better, more straightforward investment.
Why don't you walk us through the upside?
The upside is self-evident. There are 40-some-odd million cars sold in the U.S. every year, and the vast majority of that market is addressable for them. As Carvana gets bigger, it gets better, and there's nothing like it. There's also the new-car opportunity, and so they're going to sell many, many millions of cars—tens of millions, 20 million, 30 million. These are tough questions to answer, but many, many millions.
The company already has margins that are well in excess of 2.5 to 3 times those of the rest of the industry, despite growing at a fast pace. They also have meaningful fixed costs that they can leverage over time. There are plenty of opportunities to improve the core unit economics of the transactions, to be smarter about exactly which parts on which vehicle you replace in the reconditioning process, and to manage warranty and vehicle service contract expense versus cost of goods.
The upside is enormous. If you do some math, you could work out for yourself that they could maybe make, in today's money, $3,000 or $3,500 of net income per car. You multiply that by 20 million cars and that's $60 billion. That's in today's money and in a future year—maybe 10, 15, or 20 years out. If you discount that back, it's a tremendous investment opportunity. Of course, they generate lots and lots of cash along the way.
The point I wasn't going to make was the upside. The point I was going to make was the downside, because I think the company's resilience now is far greater than it's ever been. This is why, when you asked what keeps me up at night, I actually find that I'm not kept up at night about it very much. Certainly, you can worry about the stock or whatever, but that's just wasted energy.
In terms of the business, when you have margins that are so much higher than everybody else's, you have the ability to absorb shocks. You compete with an industry where people don't have vast resources to operate at losses over long spans of time. They have to make money, or at least break even.
To the extent there are shocks to the industry, provided those shocks are roughly evenly apportioned, Carvana should be able to manage its own economics by choosing how it trades off between volume and price. That makes the business very resilient. When things go wrong, they might make less money or grow less, but they won't lose money. They're also well-capitalized and all that.
When I was going through 3 possible things that might keep Cliff up at night, number 1 was, "Is Andrew fitter than me?" Obviously, that's what's going to keep you up at night the most.
Actually, number 1: Carvana in 2022 faced, as you've laid out, literal triple snake eyes—rolling the dice and getting snake eyes 3 times in a row. It was just completely crazy and unlucky. I was thinking, hey, used-car sales go down 30% for an 18-month period, for—choose your reason. Carvana, while it does have higher margins, absolutely a lot of those margins are based on fixed-cost leveraging that a lot of its competitors do not have.
So I was thinking, hey, could you have a scenario where used-car sales are down over the medium term, so Carvana’s leveraging of its fixed costs actually goes in reverse, and that’s the disaster scenario? Now, as you pointed out, the balance sheet is way better and all this sort of stuff, but that was one of the 3 things I thought might keep you up at night.
Well, first, the used-car industry is far more stable than most people realize. The 20% decline that we experienced in 2022 was the worst decline on record, and the decline in the Great Recession was in the teens. So 30% would be bigger, but sure, let’s go with that.
I think we’ve done a lot of work to try to estimate price elasticity of demand for Carvana, and we certainly don’t know it, but I don’t think it’d be crazy wrong to guess at sort of negative 7 to negative 8. So one thing you could imagine is that if there was a 30% decline in demand, if they lowered their prices 4%, all else equal, they would arguably take that back.
Now, maybe, of course, industry prices would be lower, but industry margins are generally pretty thin. Maybe industry prices come down another couple hundred basis points, so you have to give back, say, 6 points of price in that scenario. That’s no fun, but their margins are going to be 11% and change this year, and so they would still have margins comparable with the average competitor today.
Obviously, that would be in an environment where all of their competitors would be—I mean, the scenario I just outlined has all of their competitors losing lots of money and disappearing at breakneck speed.
Another tail risk I could think of—and again, I was trying to think of tail risks—is China. The EVs in China look incredible. I can’t claim to have personally driven one, so I don’t know, but the EVs in China are sub-$25,000 for cars that, to me, look better than a lot of the $50,000 gas guzzlers that we’re driving.
I don’t know if it’s, “Hey, the U.S. lowers auto tariffs and China imports EVs.” You could label that the tail risk for national-security and trade reasons, but that’s probably not it. But what if I told you Tesla comes out with a $20,000 killer electric vehicle?
The reason I point this out is that electric vehicles are cheaper to make because there are just a lot fewer parts than in gas-powered cars. The reason I think it’d be bad for Carvana is twofold. One, in this hypothetical world, Carvana is stuck with a heck of a lot of old gas-guzzling inventory that is completely stranded. It’s higher-cost and worse than this hypothetical $20,000-to-$25,000 brand-new electric car that’s killing it.
Number 2, eventually Carvana takes a one-time hit: “Hey, we’re writing off all our legacy inventory.” But if the new cars are priced at $20,000 per unit, there’s not a lot of margin left for Carvana to step in and sell used cars in this world. I realize I’m dreaming up a different world, but there are good cars getting sold in China around cost. That’s the other risk I was thinking of.
No, this is a good one. By the way, just for people listening, I specifically asked Andrew to try as hard as he could to come up with the hardest questions he could. So I hope you guys appreciate that.
There are 2 things that you said there. One was sort of a shock price—some sort of price shock to used-car prices—and the second was the risk of lower vehicle prices. I’m not going to disagree. It’s not just a price shock, though, in this case. You’re left with stranded inventory, right? You have $30,000 used gas-guzzling cars, new cars are now $20,000, and they’re way better EVs. All of that’s a write-off.
I guess I find that to be fairly implausible for the reason that there are 300 million cars in the car parc in the U.S. No matter how great Elon Musk is—and he’s great in a lot of ways—or how capable the Chinese are, the ability to produce enough cars to replace even a meaningful portion of those over any span of time shorter than many years is constrained. Cars require a lot of stuff.
Even if you had some new car that was obviously way better than every other car, it would take many years before it could get the majority of new-car sales, and many, many more years before that would filter meaningfully into the overall used-car market. There’s obviously some price volatility in used cars based on a bunch of stuff, and that could drive them down on the margin over time or whatever, but—
Can I push back there? I don’t disagree. It’s not like you’re going to replace the used-car parc overnight, but if, for some reason, Tesla introduced a $25,000 killer EV and it was new, yes, they can’t sell 100% of them, but isn’t that going to destroy the demand for used cars?
Everyone is going to be saying, “Hey, my first choice is to get this new, much cheaper EV.” Only after every last one has been bought am I going to even think about getting a used car. And, by the way, it’s got to come down, because if not, I’ll just wait on this Tesla killer EV that I’m framing.
Yeah, a lot of people in the used-car market really can’t wait 5 or 10 years for Tesla to ramp that much. Certainly, I think if people saw Tesla ramping some super-low-cost car at some high rate, they might have rational expectations, and that would probably drive down the price of used cars on the margin, but nothing like some sort of calamitous decline.
To put it in perspective, for every car Carvana sells per year, it has to hold about $4,000 of inventory because it turns the cars about 6 times a year. So if you think about the risk, a 10% shock to car prices is sort of $400 a car for Carvana, in the context of a $4,500 incremental margin per unit, or mid-$3,000s eventually going to $4,000 overall EBITDA per unit. I’m not particularly worried about the one-time shock.
Your better question is, let’s just say that there’s significant deflation in the real price of cars over a span of time due to technological innovation. I think that’s fair. If cars are cheaper, all else equal, the used-vehicle market will be smaller. Some of that would be offset by people having more of them, and some of that might be offset by people being richer and therefore more likely to exchange cars.
In some limit case, if you drove the price of cars down to a few thousand bucks or something, then they could be disposable, or there would be no used market in the sense of the way we treat cell phones or old TVs or whatnot. However, at least in the sort of work I’ve done, it doesn’t seem likely that new-car prices are going to go so low that people would still not be interested in a discounted used car.
It may be that that lowers the price and shrinks the market somewhat over time. Another offset would be that, as people get richer—as you know, the history has been that while we’ve gotten ever more efficient at making cars, people have asked for more and more out of them. Even as you lower the price of manufacturing an ICE engine, people want more horsepower, air conditioning, and everything else.
So it’s not obvious to me that there’s a future where cars are all like $5,000 a pop, but that would be bad for Carvana. A future where new cars are $30,000 a pop would be a marginal hit to the industry size. Keep in mind, it wouldn’t just be the price of cars offset by the number of cars; the net effect would be smaller than the price deflation.
Yeah. Look, cars, I think they’re pretty much at—not at the curve, but they’re close to the curve of efficiency, given the limitations of union contracts and all that sort of stuff. I’d be surprised if you were talking about new cars going for mid-$30,000s today and, 10 years from now, a new car costs $5,000 just because there’s so much metal in there, right? There’s so much metal, so much energy. It’s hard for me to imagine that world.
I think a lot of people who haven’t read your letters or listened might wonder, “Hey, Andrew and Cliff haven’t talked about the risk of driverless cars,” right? And you’ve actually—this is like, I remember 18 months ago, I was in New Orleans and talking to you on the phone, and I was like, “Driverless-car risk?” You broke it down beautifully for me.
So you can break that down now if you want to. I do have a follow-up question on driverless cars, though. I guess we might as well. We’re on a podcast. Might as well not leave the listeners hanging. Let’s do it. It’s fine.
So, this is a long answer.
I think you have to try and keep it within a reasonable time limit, because I do have a couple more non-Carvana questions I want to ask you.
Okay. And by the way, you don't need to limit this to an hour. You're welcome to, but I don't have anything until 4, so you can keep it brief.
So, people are worried that—and in general, the concern with self-driving cars, I think, goes something like this. If you look at a car, it spends a lot of its time idle, and while it's idle, it accumulates depreciation and capital costs. If you could increase the utilization of cars by sharing them because they're self-driving, you could materially reduce the cost per mile driven.
The theory would be that if we introduce self-driving cars, what will happen is we'll all end up ultimately not owning our own cars, but participating in basically large shared self-driving fleets. Without car ownership, of course, there's no need for a used-car market. That's a valid perspective, but I think it has a number of mistakes.
The first is that it is correct that if you imagine a pooled system of self-driving cars, such a pool would have lower average costs per mile driven. But what that misses is that by introducing a pooled system, you would add deadhead miles into the system. While a car sitting idle and incurring capital and depreciation costs is costly, a car driving empty is very costly.
If you look at the portion of miles driven in Ubers that are empty, or New York City taxis that are empty, or even long-haul trucking that is empty, you get numbers like 30% to 50% as the portion of miles driven that are empty. The reason why this is so high is that there are a lot of them, and this doesn't go away. As the system scales, you might think the system gets denser. You might think these costs go away, but they don't.
The reason is, first, that people live a certain distance apart, and cars have to travel a certain distance in order to get to their fares. The other is that there are net flows of people. People travel in certain directions on balance, and that creates a need to have basically empty cars drive back in order to make repeated trips in the same direction.
An example: people live in the suburbs. They want to drive into the city in the morning, and they want to drive out of the city at night.
Exactly—to the central business district or home, all that stuff.
And so, if you try to factor that in, what you discover is that there are other overhead costs for a shared system—cleaning and the various payments and so forth—but the biggest is the deadhead cost. When you factor that in, what you discover is that a shared system's cost advantages largely disappear.
Obviously, there are a lot of assumptions as you're doing this sort of analysis, but a shared system might be a little more expensive, or it might be a little less expensive. It's not obvious, and there certainly aren't large savings. But that actually doesn't begin to capture the problems with the theory.
In doing these analyses, people often think about buying a new car and using it over its life, and comparing that to the cost of shared vehicles in a shared system. They compare the average cost per mile a driver faces with some sort of shared system. But the reality is that there's an enormous amount of heterogeneity in the cost per mile driven that most drivers face.
Many drivers face a situation where, if you want to have a car that has no capital costs and no depreciation costs, that is basically available to you. You just have to buy an 8- or 10-year-old Toyota, because that vehicle has a much lower cost per mile than a shared system could ever hope to achieve. For drivers who are very price-sensitive, a shared system would never be cheaper than buying their own older used car.
Conversely, the buyer of a brand-new BMW 7 Series faces much higher costs per mile than a shared system would offer, but they're getting something else for it, which is the quality of the vehicle. Once you factor in the mix of costs faced by different users, you discover that this would appeal to a much narrower set of people than one might think.
Beyond that, there are significant hedonic reasons why people would prefer owning their own cars. To be clear, I'm comparing owning your own self-driving car to participating in a shared self-driving vehicle fleet.
If you look at how people use their cars, for one, wait times suck. The average trip length, I think, is 15 or 20 minutes. If you imagine one of these shared systems, for it to be cost-efficient, you're going to have wait times. Those wait times might average a couple of minutes, but people will need to plan around the two-standard-deviation wait times, or whatnot.
You can imagine users needing to factor in 5 minutes of wait time for a 15- or 20-minute trip. You can work out that the implied wage on that, or whatever, more than eliminates any savings.
I used to be skeptical of that. Then you have a kid, and you're like, “Look, if I check out of the store and I've got a kid, and you add 5 minutes to the thing instead of me being able to walk right up to the car and put them in—to say nothing of the fact that I've got a young kid and you have to install the car seat yourself—if you add 5 minutes to that, I might as well just go throw myself off this building right now like 100 times a year.”
No. No chance, good sir. There are going to be temper tantrums. It's going to be a disaster.
Well, I also find that the mental energy of needing to remember to call an Uber is a tax on you.
That's right. But then there are the other hedonic reasons. Your point about car seats and stuff is really important. There's a reason why we have a variety of different cars. People own different cars to say different things about them and for different reasons.
Parents might own a minivan. They have kids' car seats installed in it, and they're leaving their kids' toys in it. People who work in various trades might have a pickup truck; they use it to hold tools. Some people might drive a sports car because they want to show off to the opposite sex. There are a lot of use cases for cars well above and beyond just transit.
More importantly, people leave stuff in their cars. They might go to the gym in the morning and leave their gym bag in their car while they're at work. Or maybe they'll pick up their dry cleaning, go into the grocery store, and come back. There are a lot of use cases that really start to add up and make owning your own self-driving car far more appealing to most people.
I actually think that with self-driving, you're sitting unoccupied in this space, and the shape of a car might evolve—I suspect it will—to be less driver-centered. It becomes a room, like a personal room that you occupy. I actually think that will increase people's hedonic reasons for owning their own car.
Nobody really loves sitting in a shared waiting space in other people's space. People want their stuff there. Those are all strong hedonic reasons why people would be willing to pay a premium to have their own self-driving car.
In fact, across many markets, we see that people own their own boats, they own their own RVs, they own their own evening gowns, and they own their own second homes. There's a whole host of businesses and industries where there's an economic argument.
When I really think about this, I think the argument that we're all going to have these self-driving vehicles—this shared pool of homogeneous self-driving cars moving us around—is some engineer's vision of us all sharing the same golf carts. It ignores the fact that we're people, and it flattens all of our humanity.
All that being said, I don't think it would be much more efficient for us all to have shared self-driving vehicles. It could be less efficient, and owning our own cars has a lot of advantages in many use cases.
I might even make a somewhat controversial point: if you survey people outside of a few big cities about why they use shared fleets such as Uber, by far the dominant answers are drinking, going to the airport, or being concerned about parking. If self-driving vehicles are available, you might make the case that in many of these use cases, demand for shared fleets would actually go down.
Instead of taking an Uber to the airport, I would take my own car, and it would drive itself home at the end. The broad point is that there would certainly be use cases for a shared vehicle fleet.
It’s a sort of auxiliary-car type of situation, much the way people use Ubers in the suburbs, for example, for the other 10% of uses. This includes people who live in cities where they don’t want to park a car, although with your own self-driving car, you can imagine having it parked away and coming to get you when you want it. There are reasons why a shared fleet would exist, but I don’t see it as threatening the family car as a means of vehicle ownership.
You actually hit most of my tangential questions on the things that keep you up at night.
Oh, one more example. Sorry: India. In India, because wage rates are so low, you could make a case that cars are already economically self-driving, and people in wealthier families in India don’t all share a giant Uber system where they have their own car or their own driver. Maybe it’s different—a lower-trust society, whatever—but my point is just that I don’t see people giving up. People like to own things.
Let me ask a different question, and I’ll bring this back to Carvana in a second. Do you want to tell your Google-Stanford story?
I wrote a letter once where I talked about how you might make the case that the people—Stanford among them—who sold Google around the time of its IPO and shortly thereafter made, in some sense, one of the worst investment decisions ever. The premise of the letter at the time was that treating your sell decision as seriously as you treat your buy decision can be underappreciated. A lot of the time, enormous effort goes into making a buy decision, but then the sell decision can just be rule-based: “Well, you know what? We’re done with this now.”
The premise at the time was that I was talking about the investment in Carvana, and that there’s a lot of well-founded wisdom in the idea that once something gets to be a certain part of your portfolio, you should trim it. I understand that has benefits, but if you do that, then you will inevitably never own something that really does tremendously well for you. Another example would be the people at the Berkshire Hathaway investor day who made hundreds of millions or billions of dollars owning Berkshire Hathaway. They definitely did not follow the rule of diversifying once something went over 30% of their portfolio.
That’s one of the things I’ve thought about in considering whether it’s appropriate to own so much Carvana. That’s been a framing I’ve found helpful.
I love that framing, but I do have a question on it. When you talked about that framing, you said, “Hold the winners and hold the multibaggers.” The names that would most frequently come up would be Google, Amazon, and Facebook, right? A few people maybe bought Apple in the ’80s and held it through the post-Steve Jobs crisis until he came back and rode it to glory. Those are the examples that really jump out.
Four of those are among the greatest tech companies we’ve ever had, and all of them evolved into businesses that, if you bought them originally, you would have never seen coming. You would have never seen AWS coming with Amazon. Google’s core business is great, but you wouldn’t have seen the incredible YouTube acquisition, along with tons of other businesses that have created value. With Apple, you bought it for computers in the ’80s; you never saw the iPhone coming. The only one that would qualify differently was Berkshire, where it was kind of just betting on the singular genius of Warren Buffett.
Can I pause there?
Yeah. Walmart, Costco, Home Depot—just thinking off the top of my head here—Fastenal. All those are incredible winners. But with the exception of maybe Walmart, I’m not sure I hear people say, “Oh, you know, this group of people bought Home Depot and Walmart and rode them all the way to a gazillion.” I’m just biasing myself when I instantly thought of the big 4 tech companies and Berkshire as the prototypical examples of this.
Well, maybe they didn’t, but that doesn’t mean someone couldn’t have, I guess. Certainly, the Waltons owned a lot of Walmart all the way up. I guess you’re trying to say these companies all had to reinvent themselves, and I selected those because I don’t think those did.
That’s a great point. You were going exactly where I was going, because I was going to say, “Hey, Carvana is dominating one used-car market,” but everything else is either a tech giant that evolved into a new thing or the singular genius of Warren Buffett, who also bought a heck of a lot of things along the way. How does Carvana fit into that frame? I think you very successfully jumped ahead of me.
I do think that when I put on a very long-term lens, it’s interesting to contemplate what Carvana could be over a long period of time. How much better can this system get, right, with self-driving trucks and maybe, eventually, the introduction of robotics? It does seem to me that technology is working in Carvana’s direction, and as all these things evolve, it’s more likely than not that we’re going to see that it’s relatively more advantaged than its competitive set of dealerships as technology continues to evolve.
Carvana, of course, is going to achieve its goal of basically helping customers move cars between each other—helping people move cars between each other—in different ways in 10 or 20 years. But I think that the role of doing that, and their advantage in doing that, should only get bigger based on what I know about how technology is trending downward.
A completely different question, moving off Carvana. This is not a specific commentary on this company; it’s just something I think about a lot. I think your history with the Carvana situation and other things is relevant here, but if I looked at your 13Fs, Herbalife used to be a huge position for you. I believe, based on your 13Fs, you exited in 2022, and that was a position I think you did well in.
Again, it’s not about that position, but if you looked at Herbalife today, the stock, for a variety of reasons, is far below where you sold it. I always look at these companies where I sold and then, a year or 3 years later, however long, the stock is down 50%, 80%, whatever percent, and think, “Oh, well, I dodged a bullet there.” But what do I learn from that kind of bullet-dodging? Was I lucky to get out right before? Did I see it?
Famously, Warren Buffett sold Fannie Mae and Freddie Mac about 5 years before the crisis because he saw a lot of cockroaches creeping around. Did I see that? Was I lucky to sell? When you see a company crash like that after you exit, how do you think about it in your investing process and your ongoing investments?
Herbalife is an interesting story. I invested in Herbalife when Bill Ackman did his whole “It’s a pyramid scheme” presentation. I’ll just take a moment to point out that at the time, Bill Ackman was wildly incorrect. He was entirely incorrect and remains entirely incorrect.
He was psychologically short it for 10 years, though.
Yeah. It was manifestly simple to show that it was not a pyramid scheme. It never was.
When I bought it, the company had grown in the teens for a very long time. I expected it might slow some. I kind of thought the past would be prologue. There were decades of growth in the teens, and it had penetration in markets that was way higher than in other markets. It was a viral business in the sense that you could think about it multiplying, so there was no reason that if they’d achieved some level of penetration in Los Angeles, they shouldn’t eventually get there in Minnesota.
What happened? A lot of things happened, but basically, over the next 5, 6, 7, or 8 years, there was a whole series of setbacks. Some of them were directly caused by Ackman. They had people standing outside Herbalife facilities telling them not to do it. There was a lot of distraction, and then they made changes—and changes are always disruptive to these organizations. All this stuff created seemingly one-time-type setbacks in this global business. Invariably, there’s always somewhere in the world where something goes wrong.
For a while, it looked like we had to decide whether to focus on the last year or 2 of poor performance or think about the 30-year body of work. By the time we got to 2020 and 2021, I was increasingly of the view that we had more and more evidence that, over time, growth had proven to be elusive. The company wasn’t necessarily shrinking; it was just going sideways.
Then it was generating cash and buying back stock, and the investment had been fine from a returns perspective, but it had underperformed what I’d hoped. To be honest, I don’t really have a great explanation as to why it was so successful for so long and then became less successful after 2013.
The best theory I have is actually that the gig economy worked against them. If you look across multilevel marketers, it’s been a hard life since 2012. My hypothesis—which I can’t really prove, or haven’t been able to invalidate—is that people who might have needed some side income used to work their way into one of these MLMs, and then some percentage of them would ultimately be good at it and build a business. With the choice of driving for Uber or something, that funnel got reduced a little bit.
One of the things about these businesses is the virality coefficient. A business that’s growing 2% is dangerously close to shrinking 2%. In a retailer, if you lose 2% of your sales, you lose 2% of your sales. Your store is still there; you keep going. With a multilevel marketer, if you lose 2% of your distributors, the effect compounds.
And so I began to worry about that. In general, I would say that my view of Herbalife by 2021 was that I wasn’t exactly sure why, but it was time to start thinking about the growth rate between 2014 and 2021 as maybe what I should think the company could do, versus the growth rate from 1982 until 2013.
I was already in a smaller position at that point. Other things had gone up. It had been okay, but sort of lackluster. Then it turned out that I was interested in health and wellness and biotech, and I read the GLP-1 agonist results, and I basically thought, “Well, that’s that.” So I sold it.
In fairness, if you’d asked me in 2018 what would have really changed my mind on Herbalife, I would have said, “If there was a really effective medical treatment for weight loss, that’d be really bad.” It just didn’t seem like that sort of thing. Then, of course, I sold it, and I had the intelligence to look at Novo Nordisk and Eli Lilly. Of course, I didn’t buy those because I’m just not that bright. I got half of it right: I sold Herbalife.
It’s tough to pick the winner, though. It’s tough to pick the winner. Now, all of them won, but it’s tough to pick.
Let me approach this question one more way. You sold Herbalife, and that’s great. What I heard there is, A, you reassessed your premise, and B, one of the tail risks that you had worried about came to fruition. You saw it early and got out, so that’s actually a really successful example of a lot—
It turns out that the tail risk was in an area I was constantly monitoring.
But let me ask it slightly differently. That’s a great example of it, but have you taken anything away from that example when you’re researching your investments? I know one thing I’ve done is, as these little tail risks of companies I’ve followed or invested in have hit, I start building out and thinking, “Hey, does this company have X, Y, Z tail risk? Maybe I shouldn’t be investing in it, or maybe I need more upside if I’m investing in something with this tail risk that kicked me in the balls 2 years ago.”
I don’t quite understand the question.
So, Herbalife—ah, forget it, whatever. You’re asking the question basically—
I mean, there are other examples I’ve had in the portfolio where I’ve sold things and the company has done lackluster afterward. In my very early days, I was invested in Celanese and Ashland, and I sold both of those for valuation-type reasons, and it turns out that they’ve underperformed. Celanese really made some terrible mistakes in the last couple of years and really got itself into a bit of trouble.
I don’t know. I think the game is hard. The reason why you need the margin of safety in your valuation is because things will invariably sometimes go right and sometimes go wrong.
I also think that, in the case of Celanese and Ashland, I have a greater appreciation for how difficult chemical companies are, in a way that I maybe didn’t before. When you’re 27, you have a little bit more confidence that you can predict the future will be different from the past than when you’re 43.
When you’re 27, you read a 10-K and the business description says, “The petroleum goes in, they put it through a thing, and it goes out.” You think, “Cool. I understand chemicals.” Then you follow chemical companies for 5 years and you’re like, “I understand 0.1% about chemicals.” It’s the most complicated business.
All right, I have 2 more questions. This is one of the reasons I do this podcast: I can ask questions I’d be embarrassed to ask in person.
Expert networks. If you’d asked me 18 months ago—2 years ago, probably before I met you—I would have said, “I use expert networks a lot.” I probably read an expert call on Tegus, AlphaSense, or whatever network every other day. When I’m researching an investment, I probably do several calls. I probably average out to more than 1 a month.
Remember, I’m not running enormous amounts of money or something, so there’s a budget constraint. But I would have said I’m a pretty frequent user—not as much as I used to be in my private-equity and consulting days, but pretty frequently on the scale of public-market investors.
Then I met you, and I learned how very, very wrong I am. I’m not going to ask you about your budget or anything, but what’s 1 thing that, when you talk to other investors, you think investors fail to do with expert networks that you do? Or what’s 1 way that people can use expert networks better?
Yeah. What’s his name? There’s an author who wrote these super-long books. I can’t believe I can’t remember his name right now. Brandon Sanderson? No, no, no. They were about power. Anyway, he wrote the Lyndon Johnson books, and he also wrote The Power Broker.
Robert Caro—yes. His first book was about the guy who built New York. What’s his name? Robert Moses. Yes. Robert Caro, who wrote books about Robert Moses and Lyndon Johnson, wrote a book called Working, where he talked about writing and interviewing.
In it, he described how important it was to be quiet and let other people talk when you’re interviewing. He would actually sit there and write “STFU” over and over in his notepad—which means “shut the fuck up”—while he was waiting awkwardly for the person to eventually throw out that last fact they’d been debating saying but hadn’t.
I’d say that when I was younger, I used to go to interviews with questions. I asked them, listened to the answers, and asked my next question. Increasingly now, I have questions and want to get the answers to them, but I’ll generally start by trying to ask open-ended questions and let them talk—and shut the fuck up. It’s a little tough to do because it makes you seem like you don’t know anything, and I think people like to think they know things.
But what happens is they talk about things you didn’t know about, and you get to learn how they’re thinking about things. Then you write down your follow-ups. You don’t ask your follow-ups; you write them down.
After they’re done, you go through your list of follow-ups and ask a follow-up. As you do that, you’ll follow that thread for a while. Eventually, you’ll get to the end of the call, and then you look at your list of questions. Lo and behold, if it was a decently run call, you’ve covered them. If not, maybe there are a few left, and you throw them at the guy or gal.
By the way, this also works really well for management teams, because what you learn with a management team—in addition to learning the things you didn’t know to ask about with the experts—is what they’re thinking about. Lord knows, if what they’re thinking about isn’t what you’re thinking about, that’s a really interesting thing to know.
I think that’s the biggest thing. I read Tegus transcripts, too, and they’re very mixed. There are a lot of people who go on these things who do a lot of talking.
It’s funny, because the whole time you were saying the STFU thing, I wanted to jump in and interrupt you and tell you some stuff. But now you’ve sent me the audio of a few of your expert calls, and as you were saying it, I thought, “The Cliff who I talk to on the phone is so much different from the Cliff who I listen to in the expert interviews,” because you do let the person go on and on and on.
So let them go on well past the point where they’re talking in tangents. Just let them go. It’s weird because you’re on the clock and it costs money, but eventually you can redirect them. You’d be surprised: they’re rambling on and on, and then they’re like, “And of course, everyone hated the people at Ops.” You’re like, “Do tell me more about why everyone hated the people at Ops.” Then they’re like, “Oh, everyone knows the CEO made that decision because his wife was divorcing him and he needed a quick inflow of cash.” You’re like, “What?”
Okay, last question here. Here’s my second puff thing for you. I do this thing I call my trite Munger series and my trite Buffett series. They’ll say something, and when I first read it, I’ll groan. It’s so corny, so obvious, so hokey. I’ll groan, and then 5 years later, I’ll have a couple of extra grays in my beard and hair, and I’ll reread it and be like, “Yes, it’s really hokey. It’s really silly.” But there’s a lot of wisdom in there, and I think you’re the only person I know who’s got what I’ll call a trite Cliff saying. You’re the only other person I’ve talked to who’s got a trite saying.
Every now and then, when I call you—I think it happened about 2 years ago—I’ll say, “Hey, what are you up to?” and you’ll say, “I’m waiting around for my stocks to go up.” The first time you said it, I groaned because it’s so silly and so arrogant, and it’s probably both of those. But when you think about it a little bit, you’re like, “What is any investor who’s doing long-term, concentrated investing doing aside from buying their stocks, waiting for them to go up, and probably reading a lot to make sure that the stock’s a great opportunity or that there isn’t the GLP-1 risk that we had at Herbalife?” So there’s my trite Cliff saying right there. Here’s my question right now: What are you researching while you’re waiting around for your stocks to go up?
Oh, well, I think investing is such a great business because, in order to be a great investor—to invest successfully—you kind of need to understand everything, right? There’s virtually nothing that’s off-limits or isn’t relevant in some way to understanding how the world works. I’ve actually, of late, spent a bunch of time studying health insurance.
You were going to say that. Yep.
I’ve actually found that with AI tools, seemingly old, historically inscrutable life science and biotech companies now feel more approachable. I’m less interested in the biotech stuff, although I thought yours were more interesting than they used to be, but I’m more interested in the life science stuff. Who knows if I’ll get there or whatnot?
I spend a fair bit of time keeping abreast of changes in AI. It’s obviously super interesting. One of the nice things about being involved in these companies for a while is that you get kind of into the industry. After this, there’s a really talented executive who owns a timeshare company, someone I have a relationship with, and he and I are going to catch up. We don’t really have an agenda, I don’t think, but it turns out that you get to further your network and just keep learning. This morning, I had a call related to the auto space—things that you get to do to further your network and just keep learning.
I think the top of my funnel is very unstructured. I spend a lot of time looking at a lot of different things. Sometimes it can be very happenstance. There was 1 investment—I forget which one it was right now—but I remember I didn’t actually make it, yet I got really interested in it and spent time on it because I had accidentally downloaded the wrong 10-K. I was on the plane, so I read it anyway. It really can be very happenstance.
The key is that all I try to do is learn things incrementally every day. Then I feel like I’m doing my job. I’ve never been in this position where I have so much invested in a single stock and am sort of waiting for so long, and it is actually harder than maybe people appreciate. It is really rewarding and fun to do new things, and it certainly makes you feel useful. Not doing them is actually harder.
What do I do? I spend my time learning stuff across a wide variety of things and hoping that the next thing will be the next big thing. These things are all cumulative, right? Hopefully, even if it isn’t something useful, it’ll be useful in another context. It’s the same job you do every day.
No, look, I’ve drilled some dry wells before, and then 18 months later, the company has a hiccup—their main plant explodes or something—and all of a sudden, the dry well turns out to have been a very fruitful well in hindsight.
I was just asking what you’re researching, but, yeah, even if you study a company and learn about it, it ends up being analogous to another situation, right? Or you learn about an industry, and later on you’re talking to a friend who’s dealing with a supplier in that industry. At least now you have some context; you’re not just starting from scratch. It’s all one brick at a time.
I’ve argued that a lot of investing is being able to quickly recognize parallels, and you can apply that too broadly. Cliff saw the parallels between Carvana and Amazon when he invested in it. You find a company in distress, like Buffett’s American Express distress, and you can kind of see how it is. Every hole you drill gives you another parallel to draw.
Cool. Cliff, anything else we should be talking about?
No, sir. I really appreciate you taking the time. It’s been fun.
I’m sure. You’re an in-demand man. I really appreciate you coming on. It’s good to know our friendship beseeched you to come on after the Patrick pod. I’m going to send you an invite now. Look, May 15, 2030—I’ve got a plan. So we’ll have to do it May 16 or May 14, 2030. But I’ll send you the invite now, and we’ll go from there.
And this is it. I go back into hiding after this. 5 years—for 5 years.
And that’s why May 14 or May 16, 2030, is when you’ve got the invite. By the way, I said this before we were recording, but you’re looking strong. You’re looking good. You lost some weight. I saw you hold your arms up, and there was something on there for the first time.
Well, you know, this is silly. I wanted to lose weight for a lot of reasons, but 2% of the reason was that I saw Cliff a couple of months ago. He’s like, “Do you even lift?” And I was like, “Cliff, the next time you see me, I’m going to have dieted so much. You’re not even going to be able to joke. You’ll know that I lift, my friend.”
All right. Hey, Cliff, it’s been great. We’ll chat soon. Cheers. Bye.
A quick disclaimer. Nothing on this podcast should be considered investment advice. Guests or the hosts may have positions in any of the stocks mentioned during this podcast. Please do your own work and consult a financial adviser. Thanks.