In 1984, a portfolio manager named Robert Kirby wrote one of the most important articles in investing history. Almost nobody read it. Funny enough, that included me - until now, of course.
Kirby managed money for a woman whose husband had quietly copied every buy recommendation Kirby's firm made. But the husband ignored every sales recommendation. He put roughly $5,000 into each stock (roughly $16,000 today), threw the certificates in a drawer, and never touched them again. When he died, his wife inherited the portfolio, and Kirby finally got to look inside.
The results embarrassed him for one big reason. And you probably know where I'm going with this. The neglected portfolio had crushed the professional portfolio. A few positions had gone to almost nothing. It did not matter. One holding, a small position in a company called Haloid, had grown into more than $800,000 worth of Xerox (XRX). That position alone was worth more than the entire professional account.
Kirby called it the Coffee Can Portfolio, after the Old West habit of stuffing valuables into a coffee can under the mattress. His conclusion was uncomfortable for everyone who's paid to make decisions, which is that many decisions are wrong.
I have been thinking about that article a lot lately because two of my largest holdings have run hard. GE Aerospace (GE) is up close to 50% over the past year and trades at a forward earnings multiple north of 40x. By almost any conventional assessment, that is expensive. The old me would have at least trimmed. The current me is not selling a single share.
That is not laziness, and it is not denial. It is a framework. In this article, I want to build that framework from the ground up: the academic evidence, the practitioners who proved it works, the math that makes it so powerful, and the honest risks that come with it. Then I will apply it to four stocks I own and do not intend to sell, potentially ever.
I believe this is one of the most important pieces I have written this year because it deals with the single biggest mistake I see investors make, which is selling too early.
And, as I'm always honest, I have also sold too early. Comfort Systems USA (FIX) has been one of my best calls ever, which I sold at roughly $1,000 per share last year. This year, it broke $2,000. While I like that I made good money and invested it in other promising investments, it's one of the reasons I had to write this article.
Now, let's dive in!
The Brutal Math Nobody Wants To Hear
In 2018, Hendrik Bessembinder, a finance professor at Arizona State University, published a paper with a deceptively boring title: "Do Stocks Outperform Treasury Bills?" I have brought up this article before, and I promised to keep mentioning it, as the findings are incredibly fascinating. And I am very honest when I say that I cannot stop thinking about them.
And, by the way, the answer to the question in the title is "no."
Bessembinder studied every U.S. common stock from 1926 through 2016, roughly 26,000 companies. He found that the majority of individual stocks underperformed one-month Treasury bills over their lifetimes. Think about that for a second. Most stocks, over their entire existence, lost to the most boring asset on earth. Roughly half of all stocks delivered negative lifetime returns outright. The single most common outcome for an individual stock was a loss of essentially 100%.
This is one of the reasons why I spend way more time on finding good stocks than on the ones to avoid.
So how does the stock market beat everything else over time? The answer is the concentration of outcomes.
Bessembinder found that just 4% of companies accounted for the entire net wealth creation of the U.S. stock market above Treasury bills. The other 96% collectively matched T-bills. Even more extreme, roughly 90 companies, about 0.3% of the total, generated more than half of all the wealth ever created in the market. When he extended the study globally to more than 60,000 stocks, the skew got worse.
This paper is usually quoted by the indexing crowd, and fairly so(!). After all, if you cannot identify the 4% in advance, you should basically own everything. That is a legitimate conclusion.
But if you ask me, there is a second conclusion hiding in the data, and it is the one that matters for stock pickers.
If nearly all long-term wealth comes from a tiny number of extreme winners, then the most expensive mistake in investing is not holding a loser too long, as a loser can only cost you 100%. The most expensive mistake is selling a winner too early, because a winner you sold can cost you 1,000%, or 10,000%, in forgone compounding. There's a difference in the downside of selling a winner too early and not selling a loser at all. Very different things.
Hence, in a normally distributed world, trimming winners and rebalancing into laggards is disciplined. In a positively skewed world, it is systematically transferring capital from your future 100-baggers into your future bankruptcies.
To add another name to the discussion, Thomas Phelps understood this back in 1972, when he published "100 to 1 in the Stock Market." He studied every stock that had turned $1 into $100 and found hundreds of them. The common thread was not genius stock selection, but simply keeping stocks without selling them. Phelps distilled it into four words that should hang above every brokerage account: "buy right and hold on."
Chris Mayer updated the study in 2015 in "100 Baggers" and found the same thing. The average 100-bagger took roughly 26 years to get there. The return was available to thousands of investors. Almost nobody collected it, because almost nobody sat still for 26 years.
And that makes sense, as 26 years is a very long time. That's coming from someone who turned 26 in 2021. So, it's more than 80% of my entire time on earth, and roughly twice the entire duration I have spent following markets so far.
This is a great segue into the next part of this article.
Why We Cannot Sit Still
If holding is so profitable, why does nobody do it?
The answer is that our brains are wired against it. In 2000, Brad Barber and Terrance Odean published "Trading Is Hazardous to Your Wealth," a study of tens of thousands of retail brokerage accounts. The households that traded the most underperformed the market by more than 6% per year. The gap was almost entirely self-inflicted: costs, taxes, and the reliable human tendency to sell whatever just went up and buy whatever feels comfortable.
Odean's related work found something even worse. The stocks investors sold went on to outperform the stocks they bought to replace them. Every trade was, on average, a downgrade.
And to be completely honest with you, once you understand your own psychology, you'll see it reflected in markets. Usually, markets peak whenever I get messages from people I haven't spoken to in years who ask me if now is a good time to buy. And single stocks often bottom when people start cursing and calling them the worst stocks they've ever bought.
On that note, I often add to my favorite holdings when they start to annoy me. LandBridge (LB) is a prime example. As much as I love that company, its volatility isn't always easy to stomach. But buying great companies on weakness and holding them is how we build wealth. Doing the opposite is how we destroy wealth. And as I spent my college years trading the market, I know all about that as well.
This is why I treat a never-sell list as a behavioral device as much as an analytical one. It is a pre-commitment. By deciding in advance that certain businesses are not for sale at any reasonable price, I remove my own worst instincts from the equation.
It sounds so simple, but it has made my life much easier.
Nick Sleep, whose Nomad Investment Partnership compounded at roughly 21% per year before he closed it, ran his fund this way in its final stretch: a handful of businesses, held with near-total inactivity, including an Amazon (AMZN) position he simply refused to touch for over a decade. His letters describe the job not as finding new ideas but as protecting old ones from himself.
And then there are taxes.
Every sale of an appreciated position in a taxable account hands a slice of your capital to the government, permanently.
An unrealized gain keeps compounding in full, in most jurisdictions, anyway.
It functions like an interest-free loan from the Treasury that you only repay if you choose to. Over one year, the difference is trivial. Over 30 years, deferral alone can add meaningful percentage points to your annual after-tax return, purely as a reward for doing nothing.
What "Never Sell" Actually Means
Before I go further, I need to define terms, because "never sell" gets used to describe three very different strategies, and two of them I want no part of.
The first is the pure coffee can I just brought up: buy it, bury it, never look at it again. It worked for Kirby's client, but it only worked because one Xerox outran a graveyard of zeros. I am not willing to run a portfolio where I refuse to even watch the holdings. And, given my job, I have to watch my holdings on a very frequent basis. And it's fun, too.
The second is never-sell-as-loyalty: holding a ticker because of what it used to be, because of the dividend history, because grandpa owned it. This is how people rode Eastman Kodak, Sears, and many others.
The third is my version, and the distinction is everything: valuation-agnostic, but never thesis-agnostic.
I will not sell a great business because it has gotten expensive. I absolutely will sell it if the business breaks. The moat erodes, the capital allocation turns reckless, the industry structure shifts against it, and management starts empire-building. Those are sell signals. A high multiple is not. Price is what other people think. The thesis is what I think. I only act on the second one.
The reason I sold Comfort Systems is its valuation. It got so expensive that I disliked it for a company that usually runs very low margins outside of a hyperscaler data center boom. If a company sees such an aggressive spike in its valuation that it may take decades to justify (this is in general, not about Comfort Systems), I am willing to sell. Even the late Murray Stahl, who held Texas Pacific Land (TPL) for decades, said that there was a price he would sell at. But for obvious reasons, he never revealed that price.
In general, all of this means every stock on my never-sell list has to re-earn its place, every single year. Never-sell is my strategy, but there are obviously exceptions.
The Duration Test
So what earns a business the label? I run everything through what I now call the duration test. It rests on one observation about how returns work over different time horizons.
Over 1 to 3 years, your return is dominated by the multiple. Buy at 40x earnings and watch it de-rate to 25x, and it barely matters how good the business is. Valuation is roughly 80% of the short-term story.
Over 25 to 30 years, the relationship inverts. If a business compounds intrinsic value at 12% to 15% annually for three decades, the entry multiple gets amortized into a rounding error. Pay 40x instead of 25x for a true multi-decade compounder, and you give up maybe 1.5% to 2% of annualized return over 30 years. Pay 25x for a business whose moat dies in year eight, and you lose the whole thesis. Terminal value dominates here. Charlie Munger put it best: over the long run, a stock cannot deliver much more than the business itself earns on capital, almost regardless of what you paid.
So the question "is it too expensive to hold?" is really the question "how long does the moat last?" A never-sell stock is one where I believe the competitive advantage will outlive my holding period. Concretely, I am looking for five things, and readers of my work will recognize the DNA of my TOLL+M framework in them:
Irreplaceable assets. Not just hard to replicate. Impossible, or so close that nobody rational would try. Land, networks, certified installed bases, regulatory positions.
Structural rather than cyclical advantage. The business can have cyclical earnings. Plenty of my holdings do. But the competitive position itself must not depend on the cycle.
Decades-long demand visibility. I need to be able to describe, in plain language, why the world still needs this exact company in 2050.
Low incremental capital intensity, or pricing power that overwhelms it. The best compounders grow without consuming their own cash flows.
Alignment with the macro regime. My view remains what I call Physical Stagflation: constrained physical infrastructure meeting persistent industrial demand, producing sticky inflation and durable pricing power for asset-heavy incumbents. A never-sell stock should be a beneficiary of that world, not a victim of it.
There are obviously great non-TOLL stocks, but in these cases, valuation becomes a bigger factor.
The Part Where I Argue Against Myself
Now, the uncomfortable section, because every strategy has flaws.
Start with survivorship bias. Every never-sell article, including this one, opens with Xerox or Amazon. Nobody opens with the investor who coffee-canned Nokia at the top, or Intel (INTC) a decade ago, or Citigroup (C) in 2006. The winners write the story. Bessembinder's data cuts both ways: if only 4% of stocks create all the wealth, a concentrated buy-and-hold portfolio is a high-stakes bet on selection. Hold the wrong four stocks forever, and "forever" ends badly.
Then there is the most instructive example I know, and I own its descendant. The old General Electric conglomerate was, for decades, the definitional never-sell blue chip. It was the most admired company in America. Yet, it destroyed several hundred billion dollars of market value, cut the dividend twice, and got ejected from the Dow after more than a century.
The people who held it through all of that were practicing "never sell," while GE Capital had quietly turned an industrial champion into a leveraged financial company, and the moat everyone assumed was there had already drowned.
That failure is precisely why I can own GE Aerospace today with conviction. The engine franchise buried inside the old conglomerate was always one of the greatest businesses on earth. The wrapper was the problem. The lesson is not "GE burned people, avoid GE." The lesson is: re-underwrite the actual business every year, ignore the logo, and the moment the thesis breaks, the never-sell label dies with it.
This overview may be of help:
One more risk worth naming: concentration creep. Let winners run for 20 years, and they will become enormous positions. My own book is the proof. My top three holdings are roughly 40% of my portfolio, and I have written before about the correlated Permian risk sitting inside that number. I accept it deliberately, because trimming winners to feel diversified is exactly the Bessembinder mistake. But accepting a risk on purpose and being blind to it are different things. Know which one you are doing.
Also, note that my largest positions haven't even been with me for more than five years. They are so large because I invested a lot (high conviction) and because they performed well. Imagine what my portfolio could look like ten years from now if my thesis is right. Without trimming and/or the ability to deploy a lot of new capital, I will mostly depend on 4-5 investments.
Four Stocks I Do Not Plan To Ever Sell
With the framework built, here is how it looks in practice. These are four businesses from my own portfolio that currently pass the duration test. Note what is missing from each write-up: a price target. That is the point.
GE Aerospace. Roughly 70% of revenue comes from the aftermarket, serviced from an installed base of about 50,000 commercial and 30,000 military engines. Every LEAP engine shipped today, at thin or negative margin, becomes a 25-to-40-year annuity of high-margin spare parts and shop visits. The commercial services backlog stands near $170 billion, and services revenue grew close to 40% in the most recent quarter. Certification barriers make displacement nearly impossible, as an airline cannot swap engine suppliers on an existing fleet, and a challenger would need decades and tens of billions to matter. At a forward multiple north of 40x, the stock is expensive on any screen. While I am neutral on a short-term basis, the valuation is not nearly high enough for me to sell. I cannot think of a compounder I would replace it with.
Union Pacific (UNP). Nobody will ever assemble this asset again. The right-of-way alone is irreplaceable at any price, which is why the only way American railroading can create a transcontinental network is the pending merger with Norfolk Southern (NSC), an $85 billion attempt currently grinding through the Surface Transportation Board. I want to be honest about both branches. If approved, UNP becomes the first true coast-to-coast single-line railroad, with pricing power and share gains from trucking that could compound for decades. If rejected, I still own the best-run western franchise in a two-player duopoly, and the never-sell case never depended on the deal.
Texas Pacific Land. This company has around 882,000 surface acres and a massive royalty position in the Permian Basin, assembled from a railroad bankruptcy in 1888. TPL does not drill. It owns the ground and collects royalties, water revenues, and easement fees while operators deploy the capital and take the risk. Margins are extraordinary because there is almost nothing to spend money on. This is the low-capital-intensity pillar taken to its logical extreme: the asset literally cannot be replicated, the reinvestment requirement rounds to zero, and every increment of inflation in energy and land values flows to the owner.
CME Group (CME). CME is the world's biggest futures exchange operator. It dominates interest rate products, agriculture, metals, stock indices, and other products. It has a moat so wide it can be seen from space and a top-3 margin profile in the S&P 500 due to its low-cost operations and efficient scalability.
Here's how I would summarize these four investments:
And here's my takeaway:
Takeaway
I believe one of the biggest mistakes investors make is selling their best businesses far too early.
While valuation always matters, I've become increasingly convinced that truly exceptional companies deserve to be judged by the durability of their competitive advantages, not just by the multiple they're trading at today.
That's why I focus so much on identifying businesses that can keep compounding intrinsic value for decades, even if they occasionally look expensive.
Of course, "never sell" doesn't mean ignoring reality. If the thesis breaks, I'm out. But if the moat remains intact, I would much rather let compounding do the heavy lifting than outsmart myself by constantly trading.
Let me know what you think and how you approach the "when do I sell?" question!