Henry Ellenbogen - Man Versus Machine - [Invest Like the Best, EP.452]
What this covers
This is a conversation with Henry Ellenbogen, founder of Durable Capital Partners and formerly the manager of T. Rowe Price's New Horizons Fund. He describes his work as studying people and change, and tells the story through the founders and operators he learned from, including Jeff Bezos and John Malone. His case studies run across industries. Amazon shows a 3 to 5% annual shipping cost advantage reinvested in fulfillment for about 20 years (Cost Curve Reinvestment). Cable shows 20% growth and 20% returns on equity while the rest of media earned ordinary returns. Danaher, co-founded by Mitch Rales, shows Kaizen compounding at roughly 20% a year for nearly four decades.
The second half of the range is about market structure and what comes next. Ellenbogen estimates that 80 to 90% of institutional flow comes from one- and three-month performance horizons or from quant funds reacting to price signals. He says multi-manager platforms that cut capital after short drawdowns force managers onto horizons no longer than their own careers. From that he draws the Man Versus Machine split: machines win repeatable problems on known data, and people keep the edge in judging founders and transitions. He also argues that zero-rate years inflated the count of compounders to about 120 from roughly 40, a claim that is contested.
He also covers Durable's own practice: small initial positions, buying more as a thesis is validated, and a three-year review against the original underwriting. He admits the cost of selling Walmart too early. His forecasts on AI and robotics are also contested: AI as bigger than the internet, and robot costs falling 15 to 20% a year.
Ellenbogen argues that great investing rests on judging people and change, not on predicting markets: a rare 1% of companies drive long-term returns, and rule-based or short-horizon capital, now dominated by quant and monthly-incentive flows, leaves those companies mispriced and gives long-horizon investors an edge.
- About 40 of roughly 4,000 US public stocks compound at about 20% a year over rolling decades, and about 80% of them start as small caps
- Machines beat humans at repeat-actor problems on known data, so human judgment about people and change is the remaining edge
- Short-horizon capital creates volatility that long-term owners can exploit by buying more when quality companies sell off
This asset isn't compiled yet
You're seeing its claims, ranked. Compile it to build the argument threads, weight them, and check each claim against your library — the full view.
Amazon compounded a 3 to 5% per year cost advantage in shipping over about 20 years, reinvesting it into fulfillment infrastructure, so that rivals who lacked the same distribution investment could not catch up even after they realized what was happening.
“they took that 3-5% cost advantage of getting that box to you... and then they used that economic advantage to then go build fulfillment centers”
The old closed media system scheduled programming to maximize advertising and monetization rather than when viewers wanted to watch, so the best shows aired on Thursday and the worst on Sunday night, and that closed structure was vulnerable to internet distribution.
“I'm only gonna show you the product or the TV show that you most wanna watch when I can make the most money. Not when you wanna consume it.”
Physical-world competitive advantages such as well-located real estate, dense networks and reconditioning centers, as used by Carvana, are very hard to replicate because they require land, capex, systems and culture to be built together over time.
“You can't spin those things up, these things that are super messy.”
Danaher, co-founded by Mitch Rales and his brother, built a Kaizen-based operating system over 40 years that has produced more than a dozen Fortune 500 CEOs, including the leader who turned around GE.
“there's over a dozen Fortune 500 CEOs in the United States who started their jobs at Danaher, including the guy who just has turned around GE.”
During the zero-rate era, the number of compounders was about 120 versus roughly 40 in positive real-rate periods, which Henry says made compounding about three times easier and led many investors to ignore profitability.
“during the period of time of free money there was 120. So it was three times easier to do it.”
Over a rolling 10-year period, about 40 of roughly 4,000 US public stocks compound at about 20% a year, or grow about 6x, so roughly 1% of the market produces nearly all long-run wealth creation.
“over a rolling 10 year period, you have about 40 stocks that compound well with a 20% a year or go up a little bit over 6x. So about 1% of the stock market are the valedictorians”
Multi-manager platforms that cut capital after one- or three-month drawdowns force portfolio managers to operate on horizons no longer than their own careers, which pushes them toward short-cycle decisions.
“It probably means you can't have a time horizon longer than your career horizon.”
Daily public marks and deep investor bases can force a company to confront a required transition, such as Netflix's shift from DVD rentals to streaming, sooner rather than later, and realign incentives around it.
“it sent a signal to Netflix that actually you're under a real transition here and maybe your assumptions on your financial model are… you need to have a wider fan of scenarios”
Because market volatility is driven by capital with short-cycle incentives, a long-term firm should expect to be buying high-quality compounders when they are down, not selling them.
“because of market volatility and this agency principle that I think about that's forcing people to have short-cycle views of the public markets, actually probably you're buying more of them when they're down.”
Kaizen drove 40 years of physical-product cost improvement, and AI is beginning a parallel wave that applies process improvement to human work, creating the next Amazon-style winners who cannot be caught once ahead.
“in many ways I feel like we're just getting started on processes that are done by humans.”
In the early media era, cable networks were the best businesses because they grew about 20% for 20 years with roughly 20% returns on equity, while the rest of media earned average returns; John Malone backed most of the cable entrepreneurs.
“For 20 years, cable networks grew 20% with 20 ROEs. So they were compounders”
Henry estimates that 80 to 90% of institutional flow is driven either by firms with one- and three-month performance horizons or by quant funds reacting to price signals, which makes public stock prices move on short-term signals.
“we estimate somewhere between 80 and 90% of the institutional flow is driven either by the firms that have one month and three month agency, or the quants”
Henry believes AI will be more impactful than the internet, affecting nearly every company that relies on white-collar and IP-based employees, not just technology firms.
“I personally—and I'll lay out in a second—think it probably is more impactful than the internet was.”
Henry believes the public markets remain a proven path to building a compounder or generational company, though not for everyone, because the compounder pattern is clear: growth, profitability and a managed transition.
“I believe the path to building a compounder, or even what some people would say a generational company through the public markets is proven.”
Selling the Walmart stake early meant that, had it been kept, its value would have exceeded the total assets Henry was managing (about $8 billion), showing that one wrong sell decision can mathematically wipe out many good decisions.
“the stake in Walmart would've been greater than the sum total of everything that I was managing”
The closed media system optimized for advertiser value rather than viewer timing, and the telecom bubble's broadband build-out eventually enabled YouTube and Netflix, which broke the oligopoly, so owning the durable asset was riskier than buying the next standard.
“what broadband was enabling eventually was things like YouTube and Netflix, which would break down this whole closed system that was run like an oligopoly”
Durable reviews every investment it holds for three years against what it originally underwrote; small deviations excused quarter after quarter become obvious when they persist for about 12 straight quarters.
“Of course if it deviates a little bit for 12 straight quarters, actually it's staring you in the face.”
Companies can build durable compounding advantages through human capital, talent judgment, operating excellence and continuous improvement, even without data network effects, as demonstrated by Danaher compounding around 20 percent a year for nearly 40 years.
“for nearly 40 years you've compounded wealth at 20% in something that didn't have deep physical moats or didn't have data network effects”
Henry defines a true competitive advantage as one where the company reinvests a unit-economics edge so that, even if a rival matched its people and actions tomorrow, the leader remains too far ahead to be caught.
“if your competitor does a competitive moat attack, doing the exact same thing with people as well”
In some use cases robotics is already cheaper than the equivalent physical labor process, and because robots are improving through general-purpose models, their costs may decline 15 to 20% a year, possibly faster, creating power-law outcomes for early adopters.
“in certain use cases, it's pretty clear that already the cost is lower than the equivalent analog process”
Incumbent companies that already have physical distribution, scale, and cost advantages can use AI to lower relative costs, gain revenue scale, and reinvest, turning good businesses into great ones; these are often the best stocks.
“which of the companies that are in distribution, trucking, healthcare; who are already good and maybe investing on a different curve, have the ability to use AI to either substantially lower their relative cost advantage”
Durable buys early-stage growth companies in small initial positions and adds more only as the business de-risks and proves competitive advantage, so it buys more at higher prices as the thesis is validated.
“we would then wanna buy more as it, in our lexicon gets bigger, but also de-risks and then proves to be a competitively advantaged business”
Quant funds are strong at repeatable problems on known data, so simple data-based factor bets will be beaten by machines, while humans have the advantage in judging people and understanding change.
“if it's a repeat actor problem based on known data, actually the quants are pretty good”
Just as companies in the 2010s had to master China-based cost curves for physical products, companies that ignore AI-driven cost reduction in IP-based and spread businesses will be displaced, since AI now deflates the cost of intellectual work.
“I think the same thing applies to AI, but I think it's not product based this time, it's IP based.”
Danaher's Kaizen-based business system, which began in the factory and has produced more than a dozen Fortune 500 CEOs, is now being applied conceptually to human-performed processes, which Henry believes is only beginning.
“we're just getting started on processes that are done by humans”
Henry's favorite sources of long-term advantage are physical real estate and distribution moats, which are hard to replicate, and soft moats built on human capital, operating excellence and Kaizen culture, which are equally hard to imitate.
“I really love two and they're actually quite different. I love physical real estate.”
A small company does not need to be GAAP profitable or earn above its cost of capital to be a good investment, but it must demonstrate measurable progress toward those returns.
“if you are a small company, you don't have to be GAAP profitable and you don't even have to have an ROI that is above your cost of capital, but you do have to show progress”
Durable hires people without traditional finance backgrounds and develops them internally, because the firm needs analysts who can later lead the investment organization, and junior team members can offer valuable fresh perspectives.
“we really believe that you have to be an expert in what you do, develop into it.”
Durable measures whether colleagues make each other better through 360 reviews that ask people to cite specific investments where a colleague improved their work, rather than through vague impressions of teamwork.
“we actually ask everyone to give feedback on what their colleagues did to make them better... We ask people to basically point to specific investments”
The typical compounder among the 40 six-times companies experiences a drawdown of around 50 percent, but that drop usually comes during business transitions rather than broad market declines, which only cause about 20 percent drops.
“the average one of them has a period of time where the stock goes down 50% and it doesn't go down 50% only when the market is down 20. They go down 50% when they go through transition.”
Workday's co-founders had built PeopleSoft and were returning to HR systems; their prior experience with exception management let them use cloud technology to build a system of record that handled edge cases at scale.
“if you understand HR systems of record, there's a bunch of the stuff you have to build into the product that for lack of a better word, is exception management”
When a trader's capital is cut after a bad one-to-three-month period, the practical time horizon becomes their career horizon, so such firms cannot pursue investments that take longer than a career to pay off.
“It probably means you can't have a time horizon longer than your career horizon.”
Durable's writing culture and three-year lookbacks force the investment team to hold themselves accountable, because humans become too emotionally involved to judge outcomes; deviations that persist across twelve quarters are impossible to excuse.
“if it deviates a little bit for 12 straight quarters, actually it's staring you in the face.”
Growth and profitability patterns hold regardless of interest-rate regimes, but patterns that only work when money is free include buying low-quality returns with cheap debt and valuing companies without regard to profitability.
“what patterns only exist when money's free? So not surprisingly, everyone would imagine the pattern of driving growth and profitability is perennial and actually works, regardless of whether money is free or not.”
Domino's, one of the best Russell 2000 Growth performers of the 2010s, won by investing in an app and direct customer relationship to improve convenience, the third leg of pizza's quality-value-convenience equation, on top of a capital-light franchise model.
“they realized is if we really, really invest in technology, we can make convenience a lot better”
The typical 6x-in-10-years compounder experiences a 50% stock decline, not because of market downturns but during business transitions, so investors should expect messy transitions in great companies.
“the average one of them has a period of time where the stock goes down 50% and it doesn't go down 50% only when the market is down 20. They go down 50% when they go through transition.”
Durable only invests if its memo can credibly argue it would want to buy more at higher prices in three years; if the thesis depends only on the stock being bought by someone else or a talent deal, it does not invest.
“if we can't write the memo that we want to buy more at higher prices, we can't buy the shares”
In 2022, when the average loss-making Russell 2000 Growth company fell over 70%, Durable bought more Duolingo because it judged that companies with discipline and organizational fortitude could adapt to profitability requirements, unlike the market's blanket rejection.
“The average loss making company in 2022 in the Russell 2000 Growth went down over 70%.”
The market treated Colliers as a cyclical commercial real estate broker that suffers when rates are high, but Henry argues it is really a high-quality real estate asset manager with a growing consulting platform, so Durable bought more when it sold off.
“it's actually a really good real estate asset manager, emerging really strong consulting firm. So last year when people were worried about the weak commercial real estate market and interest rates, we bought a lot more of that company.”
Affirm's CEO Max Levchin believes the company can sustain high growth without adding headcount by using AI and a process team to lean out operations that were impossible before AI, with the lean-out starting from a process rather than cost perspective.
“In addition, they can do it without adding headcount.”
Duolingo CEO Luis von Ahn built a chess product with two people in six months and then four more people, shipping it in nine months, and believes the same work would have needed four to six times as many people and four times as long before AI.
“He's doing it out of his platform. That product is well over a million DAUs”
In 2022 the average loss-making company in the Russell 2000 Growth index fell more than 70 percent, and while the market was largely right, not every company lacked the discipline to adapt to higher interest rates, so some were mispriced.
“The average loss making company in 2022 in the Russell 2000 Growth went down over 70%.”
When Netflix's stock fell from about $280 to $70 during the Qwikster and streaming transition, Henry, then at T. Rowe Price, led half of a PIPE recapitalization at a roughly $4.5 billion valuation after pushing Reed Hastings to model subscriber-loss scenarios.
“we did that PIPE at four and a half billion dollars”
Companies that have not invested in distribution and technology infrastructure for robotics are likely at least two to three years behind leaders, and each day of waiting widens the gap.
“at minimum, is probably two to three years behind and every day that they wait, they're probably getting further behind”
Duolingo's stock falls when AI companies show live translation features, and Henry argues the market is directionally right about the risk but probably wrong about the magnitude of the threat.
“And actually I don't think the market's wrong. Now, it's probably wrong on the magnitude, but I think what the market is saying is there's a risk”
A company must already be operating well before it tackles AI-driven change, because a turnaround and a strong operating performance cannot be done at the same time.
“if you're not operating well when you have to deal with change, you're not gonna be able to do two things at once. You can't do a turnaround and do well.”
Founders who have already built and won one business have a higher probability of building another great company, because they re-solve a known problem with clarity and can align people, structure and investors from the start.
“if you've been one before, you have a higher probability of being one”
Durable is a writing culture because humans cannot maintain the executive distance needed to hold themselves accountable when they know and root for the people they invest in, so written memos force clear reasoning on why a company is advantaged.
“We're a writing culture, because at the end of the day, human beings are innately human.”
Durable only hires or promotes people who could one day lead the investment organization or manage the capital base, because the firm has few senior slots and refuses to compromise on that standard.
“we don't wanna hire you, we don't want to promote you unless we think you can actually one day lead the investment organization and drive the firm. And the reason I feel that's so important is we just don't have that many slots.”
The strongest early signal that a company could be one of the rare 1 percent is whether the people running it have already done it before, so Durable starts by studying past winners and their patterns.
“if you've been one before, you have a higher probability of being one”
The zero-rate decade made it rational for venture investors to ignore profitability, for public companies to buy low-quality businesses with cheap debt, and for boards to avoid hard questions about growth versus profitability.
“it was logical for venture capitalists to value companies and not care about profitability at all.”
A long-only firm with patient capital can own meaningful positions in volatile public stocks and buy more during drawdowns, which requires fewer but larger positions than a trading-oriented fund.
“if we're gonna do less, what it means to be able on the public side to be able to own meaningful positions”
When evaluating how leaders handle AI-driven change, Durable first checks that a company already operates well, then that it already won its first market, and then that its leader is resilient, humble and able to hold a perspective while learning.
“if you're not operating well when you have to deal with change, you're not gonna be able to do two things at once. You can't do a turnaround and do well.”
About 80% of the companies that become great long-term compounders begin their compounding journey as small-cap companies, which is why Durable deliberately focuses on small caps.
“about 80% of those companies start their compounding journey as small cap companies”
Reviewing the New Horizons Fund's 50-year shareholder letters, Henry found that only about 20 stocks drove the fund's long-run performance, which led him to study the pattern of great holdings.
“it was really only 20 stocks over 50 years that drove the performance”
Durable deliberately limits its activity and portfolio breadth so that it can understand companies deeply enough to own meaningful positions through volatility and buy more when prices are attractive.
“what we have said at Durable is really simple, let's go do less so we can do more”
At Durable, a durable competitive advantage is when a company uses technology to lower costs and drive revenue, gains 30 percent or more incremental market share, and reinvests that unit-economic gain into something a rival cannot easily copy even with equal people.
“they use it to lower costs and drive revenue that result in them gaining 30% or more incremental market share in their end market”
In the cloud and mobile era, Walmart and Costco, once they understood Amazon, used their existing scale and relationships to compete, so that together the three companies captured about 62% of retail.
“62% of all retail has gone to those three”
At the peak of the free-money era, about 30% of Treasury bills worldwide had negative yields, so borrowers were paid to borrow in real terms.
“30% of all treasury bills in the world actually had negative yields and relative to inflation, you were being paid to borrow”
Henry's former firm invested in Workday in 2012 when it had about $100 million of revenue, backing co-founders who had built PeopleSoft and saw cloud as the vehicle to complete their vision.
“At my part firm I invested in Workday when they were at about $100 million of revenue scale.”
Investing at a $1 billion or $2 billion valuation in private companies such as Workday and Twitter was once controversial, but private growth companies can now be valued at $100 billion while still losing money, and some believe they can remain indefinitely private.
“What has changed obviously in the private markets is that you can be a growth company that loses money and continues to, or be marginally profitable and not only be valued at a billion dollars, you can be valued at a $100 billion.”
Henry argues that, like organisms that persist by staying in balance with their ecosystem, healthy companies must balance returns to customers, employees, shareholders and communities to sustain themselves over long periods.
“why shouldn't investing mean that there should be a healthy balance between companies that invest in their customers, their employees, their shareholders, and actually support their greater communities?”
Durable's lookbacks of sold positions and past mistakes are conducted clinically and without blame, because intellectual honesty and self-improvement make the team better, while a blame culture would discourage the honesty needed.
“We don't do any of this in the spirit of “You made a mistake; this is an environment where you should critique yourself negatively.””
The main risk for Durable is owning a company that was advantaged and well run under the last technology generation but fails to adapt to discontinuous change, so the firm tries to find advantaged companies that can go from good to great.
“Mistakenly own a company, that based on last generation competitive risk was a great company”
At Durable, making colleagues better is a formal requirement measured in 360 reviews where people cite specific investments, and it is built into weekly gating meetings, Friday insight lunches and quarterly operating reviews.
“When we do 360 reviews at Durable, we actually ask everyone to give feedback on what their colleagues did to make them better.”
Durable performs a three-year lookback on every investment it owns for three years, comparing what it underwrote with what actually happened, and the preparation needs only about two slides.
“every single investment at Durable, if we own for three years, we actually do a three year lookback on what we underwrote and what it did.”
Once Ellenbogen's former firm recognized machines beat it at pattern-based investing, it chose to stop making those investments and concentrate on people and change, where the firm had an advantage.
“where we're investing where we're not advantaged versus these machines, probably shouldn't have done it anyway, but we're gonna stop doing it”
Durable writes down its AI views every six months and updates them, and Henry says its perspective is less wrong over time because the firm has time to read and talk to experts, which operating executives often lack.
“we at Durable write down our views on AI every six months and we update them.”
Durable has not actively marketed or sought new outside capital since 2022 because markets are full and it wants to accept more capital only when entry points are attractive, and it has made about 14 new private investments since 2023 at roughly five a year.
“We haven't really, you know, marketed ourselves or tried to get new investments since 2022.”
Durable keeps about 10 to 15 percent of capital in private markets and the rest in public markets, makes roughly five new private investments a year, and sized Duolingo as a potential future compounder rather than a basis-point position.
“we probably have 10-15% of our capital in the private markets and the rest in the public markets”
A CFO should be understood not as a policeman but as someone who sells standards that force sharp capital decisions, because profitability imposes efficiency and discipline across an organization.
“you should think about your CFO's function not as a policeman, but actually as someone who basically sells standards that force you to make sharp decisions”
In investing, the youngest person in the room can have the most valuable perspective, as shown when Anouk Dey's millennial insight shaped consumer investments such as DoorDash, Sweetgreen and Warby Parker.
“the youngest person in the room, on our investment team, actually can have the most valuable perspective”
Durable prefers a competitive culture where it roots for its competitors and partners to win, using the Warriors and John Wooden model of fun and character rather than the Michael Jordan model of driving others out, while still competing hard on the day.
“We believe everyone can win.”
Henry says the kindest thing anyone did for him was his mother's response when he dropped out of Harvard at 19 to become a congressional chief of staff: she supported him but required him to take responsibility for paying for his education.
“People will support you, but you have to be able to be responsible for the consequences.”
Durable's humility means assuming rivals and quant firms are smart and learning from them, rather than assuming Durable is right and the other side is wrong.
“we never look at a problem and assume we're right and the other person's wrong”
Mentor Jack Laporte believed small companies run by people who think like owners, who improve themselves daily, treat employees well and allocate capital well, are the ones that grow and sustain over decades.
“he believed that you invested in small companies and they were run by people who thought like owners”
Several of Durable's largest public holdings, including DoorDash, Affirm, Toast, Figma and Warby Parker, were private investments that Durable led and then held and added to through the IPO and public-market phases.
“Even if I look at Durable, among our largest positions in the public markets are DoorDash, Affirm, Toast, now Figma, Warby Parker. And we led those last private rounds”
The market is probably right that AI poses risk to Duolingo, though it may overreact in magnitude, so Durable treats Duolingo as a higher-risk holding that requires a higher discount rate offset by a larger opportunity.
“Now, it's probably wrong on the magnitude, but I think what the market is saying is there's a risk”
Durable holds an extra Monday meeting where the whole team decides which investment ideas deserve deeper work, so people learn which uses of time are valuable and colleagues can help with diligence.
“we do an additional meeting where we go through ideas that we're looking at and we gate them together.”
Durable is an Act 2 organization: its founders previously ran other businesses, and the firm was structured with a clean sheet of paper to align capital and people with its long-term compounding mission.
“we ourselves were an Act II team”
Duolingo CEO Luis von Ahn combines deep AI technical expertise with business clarity and communication skill, reminding Henry of Shopify's Tobi, a rare profile that led Durable to make Duolingo one of its largest holdings.
“Luis reminds me of Tobi from Shopify.”
Max Levchin exemplifies the Act 2 entrepreneur: technically fluent, able to recruit strong talent, a credible leader, and resilient, which Durable saw first at Slide and later at Affirm.
“Max to me represents the best of an Act 2 entrepreneur.”
The most recent public-market earnings season, covering Q2 results, produced more earnings volatility than any season since the financial crisis, which Henry attributes to short-horizon and quant-driven flows.
“If you study earnings volatility, it was more volatile than any earning season since the financial crisis.”
Great investors get better at 70 than at 50, and at 50 than at 30, because they learn patterns and develop better processes over time.
“The great thing about investing and the great investors to me actually are better at 70 than they are at 50.”
Durable builds a library of case studies on past winners, including a Columbia Business School class co-taught by partner Anouk Dey, so the firm learns recurring patterns of compounders.
“the students literally do about six case studies a year, but over time you build out a library of them.”