All podcasts / Invest Like the Best / Summary

Henry Ellenbogen - Man Versus Machine (EP.452)

2025-12-16 - 106 min - source - Read full transcript

Patrick O'Shaughnessy (host)Henry Ellenbogen

Key insights

Only about 40 of roughly 4,000 US public stocks compound wealth at 20%+ annually over any rolling 10-year period, and about 80% of those started as small-cap companies.

Ellenbogen derived this by reading 50 years of New Horizons Fund shareholder letters and found the fund's entire multi-decade outperformance traced to roughly 20 stocks; this '1% valedictorian, mostly-small-cap' finding is the empirical foundation of Durable's investment philosophy and its preference for small-cap entry points.

compounding-philosophy

A single missed compounder can mathematically erase every other good decision a portfolio manager makes.

New Horizons sold its early Walmart stake; Ellenbogen calculated that had the fund kept it, the position alone would have been worth more than the entire ~$8B fund he was managing at the time - a lesson that the public market's daily liquidity makes selling a future giant a live risk every single day, not a one-time decision.

compounding-philosophy

Founders who already won once in an adjacent business ('Act 2' teams) carry a structural edge because they understand the hardest edge cases before they start over.

Workday's founders had built and lost PeopleSoft to Oracle before rebuilding HR systems of record for the cloud, and Max Levchin built Slide and PayPal before Affirm; Ellenbogen says Durable itself is an Act 2 firm and structures 5-10 new investments a year to lean into repeat relationships with these operators.

compounding-philosophy

Durable only buys an early-stage growth position if it can honestly write that it would want to buy more at a higher price should the company execute as expected - if the thesis instead depends on an eventual acquisition or a story not fully working out, the firm passes.

This 'dollar-cost-average up' discipline is paired with a willingness to buy more of already-owned compounders like Duolingo or Colliers when the price falls on macro fear rather than a broken thesis, which Ellenbogen says most rule-based or short-time-horizon investors structurally cannot do.

compounding-philosophy

An internal 'man versus machine' study concluded systematic/quant strategies dominate repeat, known-data investing problems, while fundamental investors retain an edge on judging people and discontinuous change.

Ellenbogen ran this study roughly a decade ago at T. Rowe after studying Two Sigma directly; the conclusion was to deliberately stop competing where quants have an advantage (e.g., trading on known cyclical signals like PMI reversion) and instead double down on people- and change-driven research, a split Durable has carried forward.

market-structure-and-machines

An estimated 80-90% of institutional trading flow now comes from strategies with one- to three-month time horizons, which Ellenbogen argues explains unusually high recent earnings-season volatility.

He points to the most recent quarterly earnings season being more volatile than any since the 2008 financial crisis, despite no comparable systemic stress, and attributes it to agency-constrained allocators and quant models reacting mechanically to price signals - a structural gap that a firm willing to hold through volatility, like Durable, can exploit.

market-structure-and-machines

AI represents a second, steeper wave of the Kaizen continuous-improvement discipline Danaher applied to physical businesses for 40 years, this time targeting IP- and knowledge-work-based headcount rather than factory processes.

Ellenbogen cites Affirm's Max Levchin using AI-driven process lean-outs (legal, compliance, merchant monitoring) to grow revenue without adding headcount, and frames it as the natural extension of the Danaher Business System (DBS) that Mitch Rales built starting from Japanese manufacturing Kaizen principles.

ai-and-kaizen

Durable's early view on robotics is that current use cases are already at or below the cost of equivalent human physical labor, despite being the worst the technology will ever be, and the cost-decline curve looks structurally steeper than prior automation waves.

Because general-purpose models (not narrow, task-specific ones) power current robotics, Ellenbogen estimates costs could fall 15-20%+ annually versus the 3-5% Amazon achieved with fulfillment automation over 20 years - meaning early movers in distribution and physical infrastructure could open a lead that is 2-3 years and widening for competitors to close.

ai-and-kaizen

In 2022, Durable distinguished 'free money era' investment patterns that would stop working from perennial patterns that persist across rate regimes, and used that framework to buy more of struggling growth names rather than exit.

The average loss-making Russell 2000 Growth stock fell over 70% in 2022; Durable's research found the number of viable public compounders roughly triples (about 120 vs. 40) in zero-rate environments, but that growth-plus-demonstrated-path-to-profitability still works regardless of rate regime - the basis for buying more Duolingo during that selloff rather than treating it as a broken thesis.

compounding-philosophy

Durable measures its 'make colleagues better' culture standard by requiring 360 reviews to cite specific investment instances, not general praise.

Feedback has to name concrete help - e.g., a colleague's domain knowledge changing a specific investment view or someone attending a key meeting and relaying useful color - because Ellenbogen believes being individually excellent and making colleagues better are both required, not substitutable, at Durable.

firm-culture-and-people

Durable was deliberately structured to survive its founders' departure, unlike firms Ellenbogen studied that had a great run but no succession plan.

The firm only hires and promotes people it believes could eventually run the investment organization, requires 360 reviews and mandatory three-year thesis lookbacks, and Ellenbogen frames the explicit goal as building a firm 'better the day I left than when we were the best while we were there.'

firm-culture-and-people

Public-market pressure can force a healthy business transition that private-company status might let a founder avoid or delay.

When Netflix's DVD-to-streaming pivot burned cash faster than management modeled, Ellenbogen (then at T. Rowe) personally flagged the risk to Reed Hastings and led a roughly $400M PIPE recapitalization as the stock fell from $280 to $70; he argues nearly all of the ~40 historical compounders went through at least one 50%+ drawdown during a genuine transition, and that daily public marks forced Netflix's team to realign incentives and capital allocation faster than private ownership would have.

public-vs-private-markets

Media referenced

The '4% study' (academic research on stock market compounder concentration) - paper - Ellenbogen references a widely cited study, which he recalls as coming out of Chicago, on how a small fraction of public stocks drive nearly all market wealth creation - the same conclusion Durable's own New Horizons Fund archive research reached independently.
OpenAI Dev Day - other - Cited in passing as the kind of topic that comes up in Durable's unprepared Friday lunch discussions, where the investment team shares lateral insights from the week.

Companies

Durable Capital Partners - Ellenbogen's firm, founded in 2019 after he left T. Rowe Price; built around long-term compounding, Act 2 founders, and roughly 10-15% of capital in private markets alongside a public book.
T. Rowe Price - Where Ellenbogen ran the New Horizons small-cap growth fund for years under mentor Jack Laporte, compounding 19% annually; reading the fund's 50-year shareholder letter archive shaped his 'only ~20 stocks drove 50 years of returns' thesis.
Walmart - New Horizons held Walmart from near its IPO but sold the stake; Ellenbogen calculated the foregone gains alone would have exceeded the entire fund's other assets, a formative lesson in the cost of selling a true compounder.
Amazon - Ellenbogen's largest position when running a global TMT fund and a regular research subject via lunches with Jeff Bezos; source of his core mental model - use a technology cost advantage to gain share, then reinvest it into a physical mote (fulfillment) competitors can't quickly replicate.
Costco - Along with Walmart, one of the three retailers (with Amazon) that ended up capturing 62% of all retail once they matched Amazon's technology curve while keeping their own scale advantage.
Domino's Pizza - Best-performing Russell 2000 growth stock of the 2010s despite modest ~10% growth; won by investing early in app/ordering technology to build a direct customer relationship, improving convenience on top of an already-scaled franchise model.
Workday - Ellenbogen invested around $100M revenue scale in 2012; founders Aneel Bhusri and Dave Duffield were 'Act 2' operators who had built and lost PeopleSoft to Oracle, giving them unusual clarity on HR systems' hardest edge cases when they rebuilt for the cloud.
Affirm - Max Levchin's fintech company; used as the lead example of AI applied to knowledge work - lean-out projects across legal, compliance, and merchant-monitoring processes that let the company grow without adding headcount.
Duolingo - One of Durable's largest holdings; CEO Luis von Ahn (former Carnegie Mellon AI/ML head) built the Duolingo Chess feature with a fraction of the historical headcount using AI, and Durable bought more of the stock during the 2022 growth-stock selloff on conviction in the underlying business.
Colliers / First Service - Both chaired/founded by Jay Hennick, a repeat Act 2 entrepreneur (mentored by Peter Drucker) whose real edge, per Ellenbogen, is disciplined capital allocation and decentralized incentive design rather than any single structural moat; Durable bought more Colliers when the market sold it off on rate-driven commercial real estate fears.
Danaher - Mitch Rales's Danaher Business System (DBS), the four-decade-old Kaizen-based operating discipline Ellenbogen treats as the direct ancestor of what AI is now doing to knowledge work; alumni of DBS have gone on to lead over a dozen Fortune 500 companies, including turning around GE.
Netflix - Ellenbogen, while at T. Rowe, personally called Reed Hastings to warn the DVD-to-streaming transition would burn cash faster than modeled; led a roughly $400M PIPE recapitalization when the stock fell from $280 to $70, an example he uses to argue public-market pressure can force healthy transitions.
Figma - Durable's Catherine covered Figma continuously from a private $30M-revenue company in 2020 through the 2021 growth round to today's public earnings calls - Ellenbogen's example of the private-to-public continuity his firm is structured to provide.
Two Sigma - Ellenbogen studied the firm's quant approach directly by spending time with its principal, concluding systematic strategies dominate repeat, known-data problems but not situations requiring judgment about people and discontinuous change.

Techniques and frameworks

The 1% valedictorian / compounder framework - Roughly 40 of ~4,000 US public stocks compound wealth at 20%+ annually over any rolling 10-year period (about 1%), and roughly 80% of them started life as small-cap companies - the empirical basis for Durable's small-cap-first, long-hold investment philosophy.
Act 2 teams - Backing founders or operators who already won once in an adjacent business (Workday's PeopleSoft veterans, Affirm's Max Levchin after PayPal and Slide) because they bring hard-won clarity on edge cases and can align an organization from day one.
Man versus machine - An internal study Ellenbogen ran roughly a decade ago at T. Rowe concluding quant/systematic strategies win on repeat, known-data problems, while fundamental investors keep the edge on judging people and discontinuous change - the framework behind Durable's deliberate avoidance of short-cycle, price-signal-driven trading.
Physical Kaizen to digital Kaizen - Framing AI as extending Danaher-style continuous-improvement discipline from physical, product-based processes (40 years of 3-5% annual cost deflation) to IP- and knowledge-work-based processes, with a correspondingly steeper projected cost-decline curve.
Dollar-cost-averaging up (and selectively down) - Durable only initiates an early-stage growth position if it can honestly write in the investment memo that it would want to buy more at a higher price if the company executes as expected over three years; it also buys more of proven compounders like Duolingo or Colliers when price falls on macro or narrative fear rather than a broken thesis.
Three-year investment memo lookback - Every Durable position held three years gets a two-page review comparing what the original memo predicted to what actually happened, designed to catch gradual thesis drift that a quarter-by-quarter view would excuse.
Good-to-great thesis - Durable's pattern of finding already-advantaged, non-technology companies (distribution, trucking, healthcare) positioned to use AI to widen an existing cost or scale advantage into a durable, unassailable one - versus chasing only first- and second-derivative technology plays.

Summary

Henry Ellenbogen, founder and managing partner of Durable Capital Partners, walks Patrick O'Shaughnessy through the investment philosophy he built first at T. Rowe Price's New Horizons Fund and then at Durable since 2019. The core empirical finding, which Ellenbogen derived by reading 50 years of New Horizons shareholder letters, is that only about 40 of roughly 4,000 US public stocks compound wealth at 20%+ annually over any rolling decade, and roughly 80% of them started as small-cap companies. A painful early lesson sharpened the stakes: New Horizons sold its early Walmart stake, and the foregone gains alone would have exceeded the entire fund's other assets - proof that in public markets, selling a true compounder is a live risk every single trading day, not a one-time mistake. This "1% valedictorian" framework, paired with a preference for "Act 2" founders who already won once in an adjacent business (Workday's PeopleSoft veterans, Affirm's Max Levchin after PayPal and Slide), anchors how Durable sources and sizes positions across both public markets and roughly 10-15% of capital held in late-stage private companies.

A recurring theme is how Durable positions itself relative to quant and short-cycle capital. Ellenbogen describes an internal "man versus machine" study he ran a decade ago after studying Two Sigma directly: systematic strategies dominate repeat, known-data problems, but fundamental investors retain the edge on judging people and discontinuous change. He estimates 80-90% of institutional flow now comes from one- to three-month-horizon strategies, which he says explains why the most recent earnings season was more volatile than any since the 2008 financial crisis. Durable's response is structural - own fewer positions, understand the businesses and people deeply enough to hold through volatility, and be transparent with its own investors that this means accepting short-term underperformance in service of long-term compounding.

On AI, Ellenbogen frames the technology as a second, steeper wave of the Kaizen continuous-improvement discipline that Danaher applied to physical, product-based businesses for four decades, this time targeting IP- and knowledge-work-based headcount rather than factory processes - citing Affirm's Max Levchin lean-outs and Duolingo's small AI-augmented team shipping Duolingo Chess. He extends the same logic tentatively to robotics, arguing current use cases already match or beat human labor cost despite being the technology's worst-ever state, and that general-purpose (rather than narrow) models could produce a steeper 15-20%+ annual cost-decline curve than the 3-5% Amazon achieved with fulfillment automation - opening a 2-3-year, widening lead for early movers in physical infrastructure.

The conversation also covers Durable's internal discipline: a "dollar-cost-average up" rule requiring the investment memo to justify buying more at higher prices if a thesis plays out (or the firm won't buy at all), mandatory three-year thesis lookbacks, and 360 reviews that require citing specific investments where a colleague made someone better rather than general praise. Ellenbogen argues Durable was deliberately built to survive its founders' departure, unlike firms he studied on a "listening tour" that had a great run but no succession plan. He closes with a defense of public markets: Netflix's DVD-to-streaming transition, where he personally warned Reed Hastings the pivot would burn cash faster than modeled and led a roughly $400M PIPE recapitalization as the stock fell from $280 to $70, illustrates how public-market pressure and daily marks can force a healthy transition and realign internal incentives faster than private ownership alone would.

Notable Quotes

"Let's go do less so we can do more." - Henry Ellenbogen

"The great investors to me actually are better at 70 than they are at 50... I'm in my 50s. I'm better than I was when I did this at 30." - Henry Ellenbogen

"We break every tie in pursuit of investment excellence." - Henry Ellenbogen

"If you're going to go make major decisions, you have to be thoughtful about them. People will support you, but you have to be able to be responsible for the consequences." - Henry Ellenbogen