A buyback below intrinsic value pays the shareholders who stay

Berkshire's cash had five places to go

In the first six months of 2026, Berkshire Hathaway's insurers, its railroad, its utility business, its manufacturers, and its retailers together generated more cash than any one of them needed. Greg Abel, in his first two quarters as chief executive, and Warren Buffett, now chairman, decided where the rest went.¹ Free cash flow has exactly five destinations: back into the businesses that produced it, into a business the company does not yet own, into the company's own stock, out to shareholders directly, or against what the company owes.

Two of those five share the test this course has already built. A dollar reinvested in the business, or spent buying another one outright, is worth spending only when it earns more than that dollar costs to raise: return on capital measured against the cost of capital.² A dividend and a buyback both send cash out to shareholders, but only one of them is priced. A buyback trades cash for a fixed block of the company's own future earnings at whatever the seller will take that day, so the same purchase creates value or destroys it depending only on the price paid. Debt is neither: paying it down buys certainty, not a return, and the test is how much of that certainty the business still needs.

The table below sizes four of the five for the first half of 2026, all from the same filing.¹

The same test prices reinvestment and acquisition

Berkshire's railroad and its utility business alone spent $10.6 billion on their own property and equipment in the first half of 2026, more than the company spent buying back its own stock over the same stretch.¹ None of that spending is judged as a single lump sum. Each rail siding and each substation is funded because the unit running it expects a return on that project above what the capital for it costs, the identical comparison this course has used to judge every reinvestment decision so far.² A regulated utility's cost of capital is set largely by its own regulators; a railroad's by what investors demand to hold rail assets instead of something else. Spend past the point where the next dollar of capital expenditure clears that hurdle, and the business keeps growing while the reinvestment destroys value.

An acquisition faces the same test aimed outward. On January 2, 2026, Berkshire completed a $9.7 billion purchase of OxyChem, a chemicals maker, from Occidental Petroleum.³ Buffett had already set out the standard for a deal of any size, years before this one closed: whether the purchase is likely to leave shareholders wealthier per share than not making it at all.⁴ That standard does not ask what else the money could have funded. It asks only whether the business bought, at the price paid, earns enough to have been worth owning. Berkshire's largest bet of this kind is still the 2010 purchase of the BNSF railroad, $34 billion committed on Buffett's own description of it as "an all-in wager on the economic future of the United States," a commitment sized to a country rather than a company.⁵ Sixteen years and a change of chief executive later, the test that judged OxyChem has not moved from the one that judged BNSF.

Below intrinsic value, a buyback pays the remaining owners

A buyback is not a use of cash in the way the other four are. Reinvestment, acquisition, and debt paydown all change something about the business; a dividend sends cash to every shareholder at the same fixed rate per share. A buyback instead trades company cash for a fixed number of shares at whatever price the seller will accept that day, and that price decides everything else about whether the trade helped anyone left behind.

Buffett has said as much for over a decade, and the rule has only gotten more literal with time. In Berkshire's 2012 letter to shareholders, the buyback ceiling was still tied to a formula. It started at no more than 110 percent of book value, then rose to 120 percent once a large block traded near that level.⁴ "Value is destroyed when purchases are made above intrinsic value," Buffett wrote in the same letter, with the book-value ceiling standing in for an estimate the company was not yet willing to publish outright.⁴ In July 2018, the board dropped the formula and named the standard directly: repurchases would be made whenever the price sat below Berkshire's own intrinsic value, "conservatively determined," with no book-value ceiling stated at all.⁶ The 2026 version of that same clause governs the filing covering the first half of the year.¹ It hands the call to Berkshire's chief executive after consultation with its chairman. The standard itself is unchanged across a change in who holds each title.

The arithmetic behind that standard is short enough to write out. Call the board's own estimate of per-share intrinsic value V, before the buyback, with S shares outstanding. Spending C dollars buys back C/P shares at price P. The S - C/P shares that remain now claim what is left of the company's value once that cash is gone, VS - C.

The sign depends on nothing but the price. Below intrinsic value, the shares bought back cost less than the claim on the business they represented. The remaining shareholders end up owning a bigger slice of a company worth the same amount, worth more per share than before. Above it, the company overpaid for its own stock with the remaining shareholders' money, and the same purchase makes them worse off.² That is why Berkshire stopped. From mid-2024 into early 2026, the stock traded at 1.6 to 1.8 times book value, and Berkshire repurchased nothing for 21 straight months.⁷ In March 2026, with the price down to roughly 1.45 times book value, Abel restarted the program, and Berkshire spent roughly $4.4 billion buying back its own stock over the first half of the year, most of it in the second quarter.¹⁷⁸ Nothing about the test moved. The price did.

The two doors Berkshire almost never opens

Dividends and debt paydown are both simpler, and Berkshire uses both sparingly for a reason that has nothing to do with taxes or optics. Every dollar Berkshire keeps instead of paying out is held to what Buffett calls a dollar-for-dollar test. Does a dollar retained, over time, deliver at least a dollar of market value to the shareholder who let the company keep it?⁹ Berkshire has cleared that bar for decades, which is also why it has never adopted a dividend. "We relish the dividends we receive from most of the stocks that Berkshire owns, but pay out nothing ourselves," Buffett wrote, then worked the arithmetic behind it.⁴ In his own illustration, a shareholder wants cash from a company earning 12 percent on its net worth, with the market willing to pay 125 percent of that net worth for the shares. Selling a few shares a year leaves that shareholder further ahead than collecting a dividend of the same size, because each share sold carries that 125 percent premium with it.⁴ A dividend only helps a shareholder who cannot otherwise turn stock into cash as cheaply as that.

Debt paydown is not one dial either. In the first half of 2026, Berkshire's insurance and other businesses cut their borrowings by roughly $1.7 billion net. Its railroad and utility businesses added about $2.8 billion over the same stretch, funding part of that $10.6 billion of capital spending with borrowed money instead of cash on hand.¹ Both moves answer to the same test as reinvestment: debt is worth taking on only where the project it funds still clears its own cost, and worth retiring where the business no longer needs the cushion it provides. Across the whole company, Berkshire carries about $128.6 billion of borrowings against $365.5 billion held in cash and short-term Treasury bills, the balance sheet of a company that, in its own words, uses debt sparingly and would "reject interesting opportunities rather than over-leverage" itself.⁹