All podcasts / Invest Like the Best / Summary

Dan Sundheim - The Art of Public and Private Market Investing

2026-02-24 - 76 min - source - Read full transcript

Patrick O'Shaughnessy (host)Dan Sundheim

Key insights

Public markets are more competitive on pure analytical skill, but private markets are more competitive on access to the best deals.

Sundheim argues that in private markets, once investors agree a company is excellent, the constraint becomes convincing the company to let you invest, not out-analyzing rivals, since private investors are mostly all 'trying to get at the same answer.' Public markets have far more participants but they are 'playing a different sport' - some short-term, some quant, some fundamental - which paradoxically makes public markets both more crowded and less efficient at the same time.

public-vs-private-markets

Frontier AI labs behave like a hybrid of Netflix and Spotify as businesses.

Sundheim told LLM executives directly that their business model resembles Netflix (spend enormous fixed capital upfront to train a model, then sell access at very high incremental margin) crossed with Spotify (the underlying product, unlike Netflix's differentiated content, is largely similar across providers, so personalization and accumulated user data - not raw model quality - become the source of pricing power and switching cost).

ai-business-models

Anthropic overtook the 'Lyft to OpenAI's Uber' narrative by staying focused on enterprise and coding rather than chasing every market simultaneously.

Sundheim says most people he spoke to initially dismissed investing in the presumed number-two AI lab. OpenAI is pursuing consumer, enterprise, hardware, robotics, and science all at once - a strategy Sundheim says historically rarely succeeds (even Amazon didn't move into enterprise until seven years after its consumer IPO) - while Anthropic tried consumer, found no traction, and went all-in on enterprise, becoming what he now calls the 'Uber' of the analogy.

ai-business-models

Hyperscalers (AWS, Azure, GCP) are structurally a worse long-term business because their largest customers - the LLM labs - will eventually insource compute.

Sundheim expects hyperscaler revenue to keep growing for years as AI workloads scale, but argues that once labs like Anthropic and OpenAI turn cash-flow positive (within 5-10 years, in his view), it will make more economic sense for them to build and run their own GPU infrastructure rather than rent it, the same way Meta already insources compute rather than using a hyperscaler. He also argues LLM labs are already better than hyperscalers at GPU inference specifically, since running GPU clusters differs materially from the CPU-cluster expertise hyperscalers built their businesses on.

ai-business-models

Software as a sector will likely see compressed margins and forced AI integration, similar to how e-commerce disrupted Walmart, rather than outright extinction for incumbents with strong systems of record.

Sundheim expects the first wave of AI-driven market disruption to hit software specifically (citing the post-Claude-Code software sell-off), but distinguishes between vibe-codeable point tools and deeply embedded systems of record like ERP and CRM, which he thinks survive because companies are reluctant to rebuild mission-critical infrastructure even when they could - evidenced by AI labs themselves still buying rather than building their own ERP systems.

ai-business-models

D1's near-collapse during the January 2021 GameStop short squeeze was recovered through deliberately reduced risk-taking communicated directly and vulnerably to LPs, not through a fast high-risk rebound.

Sundheim describes the trough of the drawdown in May 2022 and a pivotal LP dinner in June 2022 where, against his president's advice to cancel, he told investors D1 would 'hit singles and doubles' going forward - accepting a slower path back to the high-water mark rather than taking the higher-EV risk that the position warranted, because the firm could not emotionally withstand another cycle like GameStop. He notes it is impossible to disprove a negative narrative quickly: even a strong quarter right after a blowup reads as 'volatile and crazy' rather than proof of recovery, so the only real lever was years of slow, methodical performance combined with direct LP communication, not any single decisive move.

resilience-under-drawdown

Emotional discipline in investing is at least partly learnable, not purely an innate trait.

Sundheim, who is described as maintaining a narrow emotional band ('a four to a six') regardless of market conditions, says he has seen hedge fund managers who started out visibly volatile - 'throwing things at people on the trading floor' - go on to become generationally great investors by training themselves not to let emotion influence trading decisions.

resilience-under-drawdown

A short thesis anonymously posted online as a job-interview exercise crashed a public company's stock and launched Sundheim's hedge fund career.

In 2002, while barred from trading as a bank employee, Sundheim wrote a short case on Orthodontic Centers of America identifying capitalized expenses that should have been expensed. He posted it anonymously on Value Investors Club before his follow-up interview; the stock fell 20-30% within a day as hedge funds and mutual fund holders (T. Rowe Price, Fidelity) picked it up, and the write-up became his calling card across firms even though the interviewing fund passed on him.

investment-philosophy

The single biggest tail risk to the global economy, in Sundheim's view, is a US-China collision over Taiwan's near-monopoly on advanced semiconductor production.

He compares Taiwan's ~90%+ share of leading-edge chip manufacturing to a hypothetical single-country oil monopoly, arguing the supply chain is 'easy to destroy, hard to replicate' and that no outcome exists in which the US, China, and Taiwan are all simultaneously satisfied - someone loses, either economically (a depression-level shock if the supply chain breaks) or geopolitically (Taiwan's sovereignty).

geopolitical-risk

The best businesses are durable low-cost producers with a reinforcing cost/scale feedback loop, not simply monopolies.

Sundheim's aesthetic framework for 'the most beautiful business': providing a durable product or service sustainably at low cost, where lower cost drives more volume which drives lower cost still - citing SpaceX in launch and Costco in groceries. He notes that outright monopolies often get lazy and see returns deteriorate, so pure market dominance is not itself the ideal.

investment-philosophy

Sundheim explicitly does not want D1 to have enterprise value as a business and treats money purely as a scorecard for being a good investor.

He contrasts hedge funds ('tons of cash flow, no terminal value') with the portfolio companies he invests in ('tons of terminal value, no cash flow') and says he has no ambition to grow D1 into a large, multi-hundred-employee firm with succession value, because his motivation is intellectual and competitive rather than building a saleable asset.

investment-philosophy

Loyalty in hiring only works paired with an extremely high competence bar; longtime friends get hired at D1 because they clear that bar decisively, not because of the relationship.

Several of Sundheim's closest partners (including Jeremy, D1's president, a friend since childhood) predate any signal that Sundheim would be financially successful. He says the trust that comes from having known someone before he had money is valuable specifically because it is combined with competence that 'clears the bar by a lot,' not as a substitute for it.

investment-philosophy

Media referenced

1997 Amazon Shareholder Letter - article - Sundheim says reading Jeff Bezos's letter was the one signal that could have flagged Amazon as a missed early investment; he says Dario Amodei's essays struck him the same way when D1 first backed Anthropic
Real Dictators - podcast - Asked (on behalf of his son Zach) what he likes about this history podcast; Sundheim says he values learning about historical failure modes so he can recognize the pattern (communism, fascism) recurring today

Companies

D1 Capital Partners - Sundheim's firm, founded ~2018; runs concentrated public and private books ($30B+ AUM) including SpaceX, OpenAI, and Anthropic
Viking Global - Sundheim's employer for ~14 years before founding D1; became CIO managing over half the firm's capital before leaving to start his own fund at 40
OpenAI - D1 invested at the $125B round; Sundheim discusses the bet on LLMs as a business model and OpenAI's 'do everything' strategy (consumer, enterprise, hardware, robotics, science)
Anthropic - D1's contrarian early bet when peers called it 'the Lyft to OpenAI's Uber'; Sundheim credits Dario Amodei's writing and Anthropic's enterprise/coding focus for its current market-leading position
SpaceX - Major D1 private holding; Sundheim calls the launch business a low-cost-producer moat and Starship a game-changer for global broadband economics via satellite bandwidth
Rivian - D1 made a similarly sized bet alongside SpaceX on the thesis that EVs were software-defined cars; manufacturing ramp problems meant the return fell short of plan, unlike SpaceX
Tesla - Cited as the EV maker that won by getting up the manufacturing scale curve fastest, the capability Rivian lacked
Value Investors Club - Anonymous idea-sharing site where Sundheim posted his short thesis on Orthodontic Centers of America as a job-interview case study, launching his hedge fund career
Orthodontic Centers of America - Subject of Sundheim's 2002 short case (capitalizing expenses that should have been expensed); the stock cratered after he posted the write-up anonymously
Bear Stearns - Sundheim's employer in private equity when he wrote the VIC short case that got him fired up about markets and led to his first hedge fund job
AWS - Named as a hyperscaler Sundheim expects to grow but face margin pressure as LLM labs eventually insource compute once cash-flow positive
Azure - Same hyperscaler-insourcing thesis; Sundheim flags Azure may not accelerate as much as AWS or GCP
Meta - Cited as the precedent for insourcing: not a hyperscaler itself, already runs its own compute rather than renting from a cloud provider
Netflix - Business-model analogy for LLMs (heavy upfront fixed-asset spend, high incremental margin at scale); also cited for eventually adopting ads despite years of public resistance
Spotify - Second half of Sundheim's LLM business-model analogy: a commoditized underlying product made sticky and monetizable through personalization
Walmart - Analogy for how legacy software businesses may need to absorb margin-compressing AI disruption the way Walmart absorbed e-commerce
Costco - Cited alongside SpaceX as an example of a durable low-cost-producer business model Sundheim considers 'beautiful'
Moody's - Named as a great business but not in the same tier as SpaceX/Costco-style low-cost-producer moats

Techniques and frameworks

Netflix-Spotify LLM business model framework - Sundheim's framing for frontier AI labs: Netflix-like in spending heavily upfront on a fixed asset (the model) sold at high incremental margin, Spotify-like in that the underlying product is commoditized but personalization creates switching costs and pricing power
Hit singles and doubles - D1's post-2021/22-drawdown portfolio construction shift: deliberately lower position sizing and risk-taking to rebuild track record slowly rather than swinging for a fast recovery
Low-cost-producer moat - Sundheim's definition of a 'beautiful business': sustainable low-cost provision of a durable product/service with a positive feedback loop (lower cost drives more volume drives lower cost), cited via SpaceX and Costco
Primary-source, six-page short write-up - Sundheim's early investing discipline learned from the VIC/Orthodontic Centers case: build the model from financial filings until nothing reconciles, then document the thesis rigorously before acting

Summary

Patrick O'Shaughnessy interviews Dan Sundheim, founder and CIO of D1 Capital Partners, across public and private market strategy, the firm's near-death experience during the January 2021 GameStop squeeze, and the origin story of his career. The conversation opens with a detailed contrast between public and private investing: Sundheim argues private markets are less analytically competitive (once investors agree a company is great, the fight is over access, not insight) while public markets, despite having far more participants, are actually less efficient because those participants are "playing a different sport" - short-term quants, multi-manager funds, and retail flows rather than long-term fundamental investors.

A large portion of the episode covers D1's private AI bets. Sundheim explains why he backed Anthropic when peers dismissed it as "the Lyft to OpenAI's Uber," crediting Dario Amodei's essays with the same signal quality he retroactively wishes he'd caught in Jeff Bezos's 1997 shareholder letter before missing Amazon. He lays out a framework for LLM business models as a hybrid of Netflix (heavy upfront fixed-asset spend sold at high incremental margin) and Spotify (a largely commoditized underlying product made sticky through personalization), argues OpenAI's "do everything" strategy across consumer, enterprise, hardware, robotics and science is historically a difficult path to pull off even for the most talented teams, and predicts hyperscalers like AWS and Azure will structurally weaken over 5-10 years as their largest customers - the LLM labs themselves - eventually insource compute the way Meta already has. He also expects software as a sector to face Walmart-style margin compression from AI, distinguishing between easily vibe-coded point tools and deeply embedded systems of record that survive longer.

The emotional core of the episode is Sundheim's account of the GameStop-driven drawdown of early 2021 and the year-plus recovery that followed. He describes the isolating experience of being "top of the world" to widely assumed insolvent within weeks, the discipline of not quitting despite genuine adversity, and the pivotal June 2022 LP dinner where - against his president's advice to cancel it - he told investors D1 would deliberately "hit singles and doubles" rather than chase a fast high-risk recovery, because the firm could not emotionally withstand a repeat of that period. He frames trust-rebuilding after a blowup as something that cannot be accelerated by any single good quarter and instead requires years of consistent, lower-risk performance communicated directly and vulnerably to LPs.

The episode closes on Sundheim's origin story and worldview. He recounts posting an anonymous short thesis on Orthodontic Centers of America to Value Investors Club as a 2002 job-interview exercise; the stock cratered within a day, and the write-up (despite the interviewing fund passing on him) became his entry ticket into the hedge fund industry. He discusses SpaceX and Rivian as parallel big private bets with very different outcomes, describes his aesthetic framework for "beautiful businesses" as durable low-cost producers with reinforcing cost/scale feedback loops, and names a US-China collision over Taiwan's near-monopoly on advanced semiconductors as the single biggest tail risk to the global economy. He ends on why he deliberately keeps D1 without enterprise value or succession ambitions, treating money purely as "a scorecard," and on loyalty in hiring: longtime friends only get hired at D1 because they clear an extremely high competence bar, not because of the relationship itself.

Notable Quotes

"Money to me is a scorecard. I want to have the best score." - Dan Sundheim

"You have to almost not think like an investor. You have to think like somebody who's into science fiction." - Dan Sundheim, on forecasting AI's economic impact

"It's better to go through life being an optimist and be proved wrong than the pessimist and be proved right." - Elon Musk, as quoted by Dan Sundheim

"Emotionally, I would not be able to go through this again." - Dan Sundheim, on the 2021-22 GameStop drawdown

"The bad ones tend to be more obvious faster. The great private tech investments are sometimes slower to prove how great they are." - Dan Sundheim, contrasting Rivian and SpaceX