Warren Buffett: Compounding and Patience

Last week's game film was a man who built a machine that can only buy. This week is the opposite temperament: a man who built a machine that can wait.

Here's the thing nobody says about the record. Warren Buffett's company had a losing year eleven times in its first fifty eight. He wrote the word mistake in his own shareholder letters sixteen times in five years, and he counted, and he put the count in the next letter. This isn't a man who never lost. This is a man who kept a ledger of losing and published it.

And the record is the best sixty-year record in the game, by a distance that isn't close. So the question this game film answers: how does a man who admits his mistakes in writing every year end up with the best sixty-year record in the game?

Start where the record starts, not where the legend does. In 1962 a young fund manager in Omaha begins buying shares of a New England textile company at seven dollars and fifty cents. The mills are dying. He knows it and buys anyway, because the shares are cheaper than the company's own assets. By 1965 he controls it.

He calls it, in the letter he wrote sixty years later, an error he recognized immediately. In 1965 the company paid no income tax at all, and hadn't for most of a decade, which he describes as a blinking yellow light at a venerable pillar of American industry. He kept the name. He shut the mills in 1985. He used the shell as the vehicle for everything after.

The most famous compounding machine in history was built inside a mistake its builder named in public.

Now the other end of the story. February 2025. The letter for 2024 comes out, and it's the last one he'll write as chief executive. He's ninety four. The table on page fifteen, the same table every year, reads: compounded annual gain, 1965 to 2024, nineteen point nine percent. The index with dividends, ten point four. Overall gain since 1964: five million five hundred and two thousand percent against thirty nine thousand.

In the same letter he writes a section headed, in his words, Mistakes, Yes, We Make Them at Berkshire, and lists the kinds: capital allocation gone wrong on businesses bought whole and in part, and misjudgments of the people hired to run them. Then the count: sixteen uses of the word in five years, and a note that many other huge companies have never used the word once over that span.

And one more number from the same pages. The company that paid nothing in 1965 paid twenty six point eight billion dollars of federal income tax in 2024, and over a hundred and one billion cumulatively. He recasts it: a million-dollar check every twenty minutes, all year, and still owing in January.

Ask the tycoon's question. Who made the rule? He did, in writing, and kept to it for sixty years. Who gained? Anyone who held the shares through the eleven losing years. Who paid? Anyone who sold in one of them. Keep that. It's the board.

Strip the play to its gears. There are five, and the fifth is the one everyone leaves out when they copy the first four.

Gear one is the frame. A share is a piece of a business, not a number on a screen. The letter says it in passing, as if it were obvious: marketable equities are viewed as partial ownership of businesses. Everything else follows from that sentence. If you own a piece of a business, you care what it earns, not what someone will pay you for it on Tuesday.

Gear two is the horizon. The letter puts it plainly: the horizon for these commitments is almost always far longer than a single year, and in many cases the thinking involves decades. The year-by-year numbers, it says, will swing wildly and unpredictably. That's not a warning. It's the design. A holding period of decades is a decision to be wrong for years at a time and not act on it.

Gear three is the engine, and it's the part the retail imitators miss. Berkshire's core is insurance. Premiums arrive today; claims are paid years later. In between, the money sits in Berkshire's hands and gets invested.

In 2024 the insurance operations alone produced over twenty two billion dollars of earnings, underwriting and investment income together. Growth funded by customers paying in advance, not by lenders. Last week's machine borrowed to buy. This one gets paid to wait.

Read the difference against the money course. Borrowed money has a price and a date; the lender can pull it in the stall. Money paid in advance by a customer has neither. It stays as long as the business keeps its promises, which means the engine runs through every gear of the short cycle instead of stalling with it. That's why the eleven losing years never became the eleven forced sales.

Gear four is the powder. At the last report the company held well over three hundred billion dollars in cash and Treasury bills.

The letter is careful about what that means. It says Berkshire will forever deploy a substantial majority of shareholders' money in equities, and will never prefer ownership of cash-equivalent assets over the ownership of good businesses.

The cash isn't a view. It's ammunition, held until a good business is available at a price. And the same letter says what it thinks of the cash itself: paper money can see its value evaporate if fiscal folly prevails, and fixed-coupon bonds provide no protection against runaway currency. A man holding three hundred billion in bills wrote that. The bills are a waiting room, not a home.

Gear five is the mistake ledger. The word is used, the count is published, the categories are named. Munger, he writes, called the alternative thumb-sucking: problems can't be wished away; they require action, however uncomfortable. The other four gears only survive sixty years because the fifth one keeps them honest, and it's the only gear that costs nothing to install.

One sentence for the whole machine: own pieces of businesses, hold them for decades, fund the holding with money customers pay in advance, keep powder for the day a good business goes on sale, and write down every mistake by name.

Nineteen point nine against ten point four. Nine and a half points a year, for sixty years. That's the whole gap between the two overall figures, and it's why the second number has three commas and the first has one.

Draw the share price at the milestones and you get the shape compounding actually has. Seven dollars fifty in 1962. A thousand in 1983, twenty one years later. Ten thousand nine years after that. A hundred thousand in 2006. Half a million in 2022. Seven hundred thousand in 2024.

For the first two decades the line is flat against the later scale. Everything a viewer would call the result happens in the last third. Compounding looks like nothing for longer than most people can stand, and then it looks like everything.

Now the number the legend leaves out. Eleven losing years in fifty eight. Roughly one year in five, the shareholder who checked his statement saw less than the year before. The record wasn't built by avoiding those years. It was built by the people who didn't sell in them, and by the one man who kept buying through them.

The tax line is the other quiet one. Zero in 1965. Twenty six point eight billion in 2024. The letter presents it as a point of pride, and the desk reads it as a measurement: that's the size of the earnings the machine now throws off every year, counted by the one auditor who never rounds in your favor.

The incentive map. Who made the rule? A founder who wrote it down and never changed it. Who gained? Long holders, and the founder most, because he held the longest. Who paid? Sellers in the losing years, and the managers he names by category when they fail. Who's still paying? Nobody, which is the strangest line on any board in this course. The machine funds itself.

Here's the reveal. It isn't the nineteen point nine. It's the sixteen. The best record in the game belongs to the one operator who published his own error count, and the letter says most of his peers never used the word once; the gap on the scoreboard and the gap in the vocabulary are the same gap.

Game film is for taking one play. Here's how the office reads Omaha.

First, the frame is available to you in full. Every position the office holds is a piece of a business, or a claim on one, or a unit of money. Write down which, and what it earns or does. A ticker is a name for a thing, not the thing.

Second, the engine isn't available, so don't pretend. You don't have insurance float. The nearest thing a digital tycoon has is a business whose customers pay in advance: subscriptions, retainers, deposits. That's your float. Build it before you build a portfolio, because it's the money that lets you hold the portfolio through the eleven bad years.

Third, the rule the desk took. Not the horizon; the fund's rules already write a reason on every move. The rule is the mistake ledger. Every quarter the office writes down what it got wrong, by name, and keeps the count. Sixteen in five years is the benchmark. If the office's count is zero, the office is lying to itself, and the letter says exactly what happens next: you start believing your own baloney.

Our paper fund reads it by its rules, and the score it keeps against simply holding Bitcoin is the office's version of the table on page fifteen: one comparison, published every time, no excuses attached.

• Buffett's partnership began buying Berkshire Hathaway in 1962 at $7.50 a share and gained control in 1965; the textile operations ceased in 1985. Wikipedia, Berkshire Hathaway.
• Class A share milestones: $1,000 on August 26, 1983; $10,000 on October 16, 1992; $100,000 on October 23, 2006; $500,000 on March 16, 2022; $700,000 in August 2024. Wikipedia.
• Compounded annual gain 1965 to 2024: Berkshire 19.9 percent, S&P 500 with dividends 10.4 percent; overall gain 1964 to 2024: 5,502,284 percent against 39,054 percent. 2024 Shareholder Letter, performance table.
• Between 1965 and 2023 Berkshire compounded at 19.8 percent against 10.2 percent for the index, with negative returns in eleven of 58 years. Wikipedia.
• "In 1965, the company did not pay a dime of income tax, an embarrassment that had generally prevailed at the company for a decade." Cash income tax payments to the U.S. Treasury "now aggregate more than $101 billion"; $26.8 billion was paid for 2024. 2024 letter.
• "During the 2019-23 period, I have used the words 'mistake' or 'error' 16 times in my letters to you. Many other huge companies have never used either word over that span." 2024 letter.
• On equities: "we view these as partial ownership of businesses." On horizon: "Our horizon for such commitments is almost always far longer than a single year. In many, our thinking involves decades." 2024 letter.
• "Berkshire will never prefer ownership of cash-equivalent assets over the ownership of good businesses, whether controlled or only partially owned." 2024 letter.
• 2024 earnings as the letter presents them: insurance underwriting $9,020 million; insurance investment income $13,670 million; operating earnings $47.4 billion in total. 2024 letter.
• Cash and Treasury bills of $365.5 billion at the latest report cited; total assets $1.222 trillion in 2025. Wikipedia.
• Greg Abel succeeded Buffett as CEO on January 1, 2026; the letter notes "At 94, it won't be long before Greg Abel replaces me as CEO." Wikipedia; 2024 letter.