Cliff Sosin from CAS on Carvana and a bunch of other stuff $CVNA

Episode Summary

Executive Summary: Andrew Walker and Cliff Sosin revisit Carvana, subprime lending, and long-term investing. Sosin explains why he focuses on securitized and non-prime credit, why Carvana’s vertically integrated model can outperform, why alt data is useful but not a magic timing tool, and why threats like EV disruption or autonomous cars are less alarming than many assume. He also reflects on selling decisions, expert-network technique, and his evolving research interests.

Main Topics: Why Cliff focuses on securitizations and subprime lending (Priority: 5/5): Sosin argues his experience and skill set fit information-intensive lending businesses where underwriting, behavior modification, and structured finance matter. He sees systemic mispricing because the market paints too many non-prime lenders with one brush. Carvana’s lending and unit economics (Priority: 5/5): A major segment of the conversation centers on whether Carvana’s financing profits and loan performance are truly exceptional. Sosin says Carvana’s vertical integration, pricing, data advantages, and better collateral quality support stronger economics than skeptics assume. Alt data, timing, and the Carvana turnaround (Priority: 5/5): Sosin explains that alternative data helped him validate the business and understand trends, but it did not make the investment easy. He describes the 2022 downturn, the Omicron/logistics disruption, and how weekly data helped him track the recovery and deterioration. Carvana downside risks and competitive resilience (Priority: 4/5): Walker presses on used-car demand shocks, EV disruption, and stranded inventory risk. Sosin argues the used-car market is stable, Carvana’s margins and balance sheet provide resilience, and even disruptive scenarios likely unfold too slowly to kill the model quickly. Driverless cars and the future of car ownership (Priority: 4/5): Sosin pushes back on the idea that self-driving fleets eliminate used-car demand. He argues deadhead miles, wait times, personalization, and human preferences for ownership make shared autonomous fleets less economically compelling than critics think. Selling winners, tail risks, and lessons from Herbalife (Priority: 3/5): He uses Herbalife to illustrate the importance of reassessing thesis and monitoring specific risks. He sold as GLP-1 obesity drugs became real, and says the experience reinforces the need to weigh downside risks and not rely on stale growth narratives. How to use expert networks and think like an investor (Priority: 3/5): Sosin recommends asking open-ended questions, staying quiet, and letting experts talk. He frames investing as cumulative learning, pattern recognition, and waiting patiently for owned ideas to work out.

Key Arguments: Subprime/non-prime lending is an information business: lenders find deserving borrowers others overlook and create value through underwriting and behavior modification. Many credit blow-ups happen when signals are distorted, such as borrower churn in the late 1990s or rising home prices in 2004-2008, not because all subprime lending is bad. Carvana is not just a subprime lender; about two-thirds of its loans are prime, and the business also earns meaningful profit from vertically integrated financing. Carvana’s loan performance can be better because it has more data, better cars, lower LTVs, higher down payments, and a smoother online experience that improves selection and verification. Alt data is best for validating a broader thesis and tracking business health, not for cleanly timing quarters or avoiding every drawdown. Carvana’s 2022 weakness was partly due to Omicron-related logistics problems and partly due to a real demand decline; weekly data showed the deterioration even as the story evolved. The used-car market is more stable than many think, and Carvana could retain acceptable margins even in a severe demand slowdown by adjusting price and volume. Tesla or Chinese low-cost EVs are unlikely to rapidly destroy the used-car market because the existing car park is enormous and replacement happens slowly. Self-driving cars do not automatically eliminate ownership because shared systems suffer deadhead miles, wait times, and loss of personalization and utility. Investment selling decisions should be treated as seriously as buying decisions; otherwise investors may prematurely trim huge winners. Expert calls work best when the interviewer shuts up, asks broad questions first, and writes follow-ups instead of interrupting. The best investing skill is cumulative learning: each industry studied creates parallels that help in future situations.

Data Points: Podcast episode: 310 - The episode number for Cliff Sosin’s second appearance. Carvana profit per unit from financing: nearly $2,000 - Sosin describes the economics of Carvana’s financing business. Prime mix of Carvana loans: about two-thirds - He says Carvana’s originations over-index in subprime, but most loans are prime. Carvana delivery lead times during healthy periods: about 3 to 3.5 days - Used as a proxy for current operational health. Carvana delivery lead times during 2022 disruption: about 7 days - Sosin links longer delivery times to Omicron-era logistics problems. Used-car market decline in 2022: 20% - He cites 2022 as the worst decline on record for the industry. Great Recession used-car decline: teens - He compares the 2022 downturn to the prior recession. Price elasticity estimate for Carvana: negative 7 to negative 8 - His rough estimate of demand sensitivity to pricing. Carvana margins expected this year: 11 and change percent - He uses this to argue the business could still withstand a severe demand shock. Carvana inventory holding per car: about $4,000 - Used to explain how quickly inventory turns limit exposure to price shocks. Inventory turns: about 6 times a year - Sosin notes Carvana turns inventory roughly six times annually. Incremental EBITDA per unit: mid-threes going to four - He frames the scale of per-unit economics in a stressed scenario discussion. Annual U.S. cars sold: 40-some odd million - He uses this to size Carvana’s addressable market. Long-term per-car net income potential: $3,000 to $3,500 per car - His illustration of potential future economics at scale. Potential future total income at scale: $60 billion - He multiplies per-car profits by roughly 20 million cars. Self-driving fleet empty-mile estimate: 30% to 50% - He cites empty miles in Ubers, taxis, and trucking as a key cost problem. Herbalife growth period: teens for decades - Sosin says Herbalife grew in the teens for a very long time before slowing. Herbalife underperformance window: 2014 to 2021 - He says growth became elusive after 2013 and eventually drove him to sell. Multilevel marketer growth sensitivity: 2% growth vs. 2% shrink risk - He notes small changes in MLM distributor growth can be dangerous.

Pivotal Quotes: "It is a very noble business." — Cliff Sosin: His defense of non-prime lending as a legitimate and valuable financial service. "I think investing is such a great business because in order to be a great investor, in order to invest successfully, you kind of like need to understand everything." — Cliff Sosin: He explains why he studies many industries and uses broad learning as an edge. "I'm waiting around for my stocks to go up." — Cliff Sosin: His shorthand for the long-term, patient, concentrated-investing mindset.

Implications: For investors, the episode argues that underwriting quality, structure, and operational data matter more than simplistic labels like “subprime.” It also suggests that Carvana’s durability may be stronger than skeptics think, while future disruption risks are real but likely slower and less absolute than bears assume.