Dan Sundheim - The Art of Public and Private Market Investing - [Invest Like the Best, EP.460] — Summary & Takeaways · Colossus (Invest Like the Best / Business Breakdowns)

At a Glance

Dan Sundheim, founder and CIO of D1 Capital Partners, manages over $30 billion in assets across public and private markets, with investments in SpaceX, OpenAI, and Anthropic. Core view: In the AI space, he favors Anthropic, arguing it is not OpenAI's "Lyft," and compares the LLM business model to Netflix and Spotify. Key takeaways: He shared the extreme pressure his firm faced during the 2021 GameStop event and identified the biggest tail risk for the global economy. He also recounted an early career short of Orthodontic Centers of America, which helped him land his first job.

Theme 1: Core AI Investment Judgment — Anthropic Is Not "Lyft," LLM Business Model Analogous to Netflix and Spotify

Dan Sundheim believes the early comparison of Anthropic to OpenAI's "Lyft" was a mistake; its business model is more like a combination of Netflix and Spotify.

Historical Context and Mechanism Breakdown: When Sundheim invested in Anthropic, many warned him it was an "Uber vs. Lyft" dynamic, where the second player struggles to win. However, he argued that in the early stages, it is extremely difficult to determine who will be the ultimate first or second. His judgment was not based on the model's differentiation at the time, but on CEO Dario Amodei's clarity of thought and communication skills. He compared this to Jeff Bezos' 1997 letter to shareholders, believing Amodei's writing and speaking demonstrated "unusually clear thinking and focus," a key signal for identifying great companies early on.
Data Chain and Business Model Analogy: Sundheim breaks down the LLM business model as "Netflix + Spotify":
Netflix Component: LLMs require massive upfront capital to train models (fixed costs), but once a model is complete, it can be sold at a very high incremental marginal cost. This mirrors Netflix's flywheel of investing in content and amortizing costs through user growth.
Spotify Component: The models themselves (like music) tend toward commoditization, but personalization and data stickiness are the core moat. Spotify's music is essentially no different from Apple Music, yet it has pricing power due to its personalized recommendations. LLM competition will ultimately depend on who can better understand users and build data barriers.
Extrapolation and Uncertainty: Sundheim believes the debate over the LLM business model has shifted from "whether there will be economic returns" to "capital intensity and the speed of returns." Core uncertainties include: 1) Whether Scaling Laws will continue to hold, determining the return on capital investment; 2) The pace of enterprise adoption; if slower than expected, the financial leverage from high capital intensity poses a significant risk. He predicts 4-5 LLM companies will survive long-term, and the competitive landscape is already relatively clear.

Theme 2: The GameStop Crisis and D1's "Darkest Hour" — From "Home Runs" to "Singles"

Sundheim detailed the extreme pressure D1 Capital faced during the 2021 GameStop event and shared his key coping strategy: shifting from high-risk, high-reward pursuits to a "single" style of steady investing.

Historical Context and Pressure Description: In early 2021, D1 suffered massive drawdowns due to GameStop, dragging the firm's reputation and performance "through the mud." Sundheim described it as the "worst moment" of his career, feeling "lonely" as only a few peers faced similar difficulties. Despite external rumors of the firm's impending collapse, he never considered giving up, firmly believing the team still had the capability.
Mechanism Breakdown and Key Decision: The most critical turning point was the LP dinner in June 2022. At the time, the firm's drawdown was at its nadir. The president suggested canceling the dinner, but Sundheim insisted on holding it. At the dinner, he candidly told LPs that he would change the portfolio construction approach: abandon the "home run" high-risk strategy in favor of "singles" and "doubles" — reducing risk exposure and pursuing steadier returns. He acknowledged this might take longer to return to highs, but the team could not withstand another similar shock.
Extrapolation and Signals: Sundheim believes that after a massive drawdown, it is impossible to prove the negative narrative wrong in the short term; it takes "years" of consistent, stable performance to rebuild trust. He emphasized the importance of "having a plan": when you have a plan you believe in, even in a terrible external environment, you can shift your mindset from "helplessness" to "mission." He holds no grudge against LPs who redeemed, stating "capital follows returns," and is deeply grateful to those who stayed.

Theme 3: Market Efficiency and Investment Philosophy — Long-Term Value Opportunities Amid Short-Term Noise

Sundheim believes the current market is inefficient due to the dominance of short-term traders, creating significant opportunities for investors with a long-term horizon, and he particularly favors shorting.

Mechanism Breakdown: He points out that the decline in market efficiency stems from changes in participant structure: the rising share of passive investing, retail investors, and quant funds, while investors focused on long-term intrinsic value have decreased. Short-term traders (e.g., multi-manager funds) compete fiercely on short-term information, making that arena extremely efficient. However, once the time horizon is extended to "what is a company worth," competition becomes scarce.
Data Chain and Comparison: He contrasts today's market with 20 years ago: mutual funds and long/short hedge funds were the main players then; now, a large volume of trading is non-fundamental. This causes short-term stock price movements to often exaggerate the true change in a company's intrinsic value, creating mispricing.
Extrapolation and Personal Preference: Sundheim explicitly states he "loves shorting" because the market is full of investments based on "stories" rather than fundamentals, providing "endless shorting opportunities" for patient, fundamental investors. His investment framework is: completely ignore short-term events and focus on a company's long-term cash flows and moat. He believes most people cannot predict things three years out, but it is this uncertainty that creates arbitrage opportunities for investors with a medium-term horizon.

Theme 4: The Biggest Tail Risk for the Global Economy — Semiconductor Supply Chain Crisis in Taiwan

Sundheim believes the biggest tail risk for the global economy is a conflict over the semiconductor supply chain centered on Taiwan, with potential impacts comparable to the "Great Depression."

Mechanism Breakdown and Data Chain: He notes that Taiwan produces over 90% of the world's most advanced semiconductors, and semiconductors are the foundation of everything in modern society, as critical as oil once was. This supply chain is extremely fragile and difficult to replicate, but easy to destroy. A disruption would lead to a "catastrophic economic recession on the scale of the Great Depression."
Extrapolation and Scenario Analysis: Sundheim analyzes three possible paths:

1. US Replication of Supply Chain: Rebuilding semiconductor manufacturing capacity in the US, which could take 10-20 years. During this period, the US might reduce its defense commitment to Taiwan due to domestic supply, increasing the incentive for China to unify by force.

2. Diplomatic Resolution: The US and China reach some understanding, with China pledging no aggressive action during the supply chain rebuilding period, eventually leading to peaceful reunification.

3. Conflict Eruption: The worst-case scenario, directly causing a global economic collapse.

Falsification Conditions: He believes Chinese leaders' repeated emphasis on the Taiwan issue should be taken seriously (analogous to Putin's statements on the Soviet Union). The development of AI further increases the strategic value of semiconductors, heightening conflict risk. What he hopes to see is the US accelerating domestic supply chain replication while reaching some "time-for-space" agreement with China to avoid economic catastrophe.

Positions Mentioned

Position	Analyst Stance	Key Data
Anthropic	Bullish	Early comparison to "Lyft," but Sundheim believes CEO Dario Amodei's clarity of thought rivals Jeff Bezos; has taken a leading position in coding and enterprise markets.
OpenAI	Bullish	Invested in a $125 billion valuation round; successful in the consumer market, but its "do everything" strategy (hardware, robotics, enterprise, science) carries risk.
SpaceX	Bullish	Considers its engineering achievements "the most astonishing"; Starship will reduce launch costs by ~97%; Starlink's TAM is the global telecom market.
Rivian	Risk Warning	Investment returns did not meet expectations; manufacturing ramp-up difficulties, high capital intensity, failed to achieve scale effects in time.
Netflix	Bullish (as analogy)	Used as an analogy for the LLM business model: high upfront capital investment, high incremental marginal profit later.
Spotify	Bullish (as analogy)	Used as an analogy for the LLM business model: product commoditization, but stickiness and pricing power built through personalized data.
Amazon	Bullish (as analogy)	Misunderstood by the market for early losses, but Jeff Bezos' shareholder letters were a key signal for identifying a great company.
Costco	Bullish (as analogy)	Used as a model for a "low-cost, high-efficiency" business model.
Moody's / S&P	Neutral (as analogy)	Acknowledged as "great businesses," but not his most admired type.
Orthodontic Centers of America	Short (historical case)	Sundheim analyzed financial data and discovered it was capitalizing expenses that should have been expensed, constituting accounting fraud; the stock price halved after the short.

Judgments Worth Remembering

1. Anthropic is not OpenAI's Lyft; determining who is first or second in the early stages is extremely difficult. Support: Sundheim read CEO Dario Amodei's writing and compared it to Jeff Bezos' shareholder letters, believing his "clarity of thought" was a key signal for identifying great companies, not the model's differentiation at the time.

2. The LLM business model is a combination of "Netflix + Spotify." Support: The Netflix component refers to massive upfront capital investment to train models (fixed costs) and high incremental marginal profit later; the Spotify component refers to model commoditization, where personalized data is the core moat and source of pricing power.

3. The core debate on LLMs has shifted from "can it make money" to "the speed of capital returns." Support: Models have proven economic value, but capital intensity is unprecedented. Core uncertainties are whether scaling laws will persist and whether enterprise adoption is fast enough; otherwise, high financial leverage poses a significant risk.

4. AI will lead to a deterioration of the software industry's business model, but "system of record" software is hard to disrupt in the short term. Support: Sundheim believes software companies must integrate AI like Walmart adapted to e-commerce, a painful process. However, core system records like ERP, due to complexity and risk, will not be replaced by "vibe coding" in the short term.

5. The market is inefficient due to the dominance of short-term traders, creating significant opportunities for long-term fundamental investors. Support: The rising share of passive investing, retail investors, and quant funds makes short-term information competition extremely efficient, but once the time horizon extends to intrinsic value, competition becomes scarce, and mispricing is frequent.

6. After the GameStop crisis, D1's strategy shifted from "home runs" to "singles." Support: Sundheim announced at the June 2022 LP dinner that he would reduce risk exposure and pursue steadier returns, as the team could not withstand another similar shock. He acknowledged this would take longer to recover.

7. The biggest tail risk for the global economy is a conflict over the semiconductor supply chain in Taiwan, with impacts comparable to the "Great Depression." Support: Taiwan produces over 90% of the world's most advanced semiconductors; the supply chain is fragile and hard to replicate. A disruption would lead to a catastrophic economic recession. He hopes the US will replicate the supply chain in 10-20 years and reach a temporary agreement with China.

8. Over a 5-10 year investment time horizon, leadership matters more than the business model. Support: Sundheim believes great leaders make the right decisions and attract top talent, especially in tech. He looks for CEOs with "genuine passion, intense competitiveness, a desire to win, and deep understanding of details," and believes people want to work for those from whom they can learn the most.