Berkshire under Greg Abel: Q3 2026 buybacks, cash and returns

Executive summary

From January 2001 to October 2026, Berkshire Hathaway's B shares compounded at 9.92% a year in US dollars, against 7.41% for MSCI World net, and lost less over the matched index peak-to-trough dates in four sell-offs studied. The third-quarter 10-Q, published on 7 November 2026, shows Greg Abel's Berkshire spending more on buybacks than on acquisitions, capital expenditure or listed-stock purchases: $9.6bn of buybacks, against $6.5bn for acquisitions, mainly Taylor Morrison.

Operating earnings fell 5.8% and cash and Treasury bills stayed above $350bn.

Copio's assessment: Berkshire combined financial resilience with faster historical compounding than the index. That supports studying its structure, but does not establish that its cash reserve caused the outperformance or carried no opportunity cost.

The qualification is recent. Over the last ten years its Sharpe ratio fell below the index's while its CAGR remained higher, 14.16% against 12.95%; in Abel's first ten months as chief executive the shares gained 7.9% against 10.9% for MSCI World. A balance sheet with more than a quarter of its assets in Treasury bills compounds at bill rates on that part, and bill yields in the quarter were lower than a year earlier.

This article follows our study of Warren Buffett's asset allocation, which covers the 90/10 instruction and Berkshire at 30 June 2026.

What does Berkshire own?

Berkshire is a holding company: insurance, a railroad, utilities and several hundred manufacturing, service and retail businesses, plus a listed-stock portfolio and a Treasury bill reserve. At 30 September 2026, its balance sheet showed total assets of $1,303.9bn. Cash and bills were 28.9% of that, equity securities 25.7%, and the operating businesses and other assets, carried at book value, about 42.4%.

The rise in cash is mostly an accounting timing effect. Berkshire bought $20.5bn of bills in September that it paid for in October (10-Q, balance sheet notes). Net of those payables, the insurance and other businesses held $352.3bn, down 1.9% from $359.2bn at 30 June (10-Q, Financial Condition).

The 10-Q names the five largest equity holdings as Alphabet, American Express, Apple, Bank of America and Coca-Cola, in alphabetical order and without weights. Together they were 65% of the $335.5bn equity portfolio (Note 4).

Holdings in banks, insurance and finance fell from $104.1bn at 31 December 2025 to $85.9bn with almost no change in their cost basis, which is consistent with lower valuations, but cannot isolate price effects from purchases and sales. The cost basis of "commercial, industrial and other" rose by $26.4bn over the same nine months, a net cost-basis increase that does not identify gross purchases or their timing.

The 13F holdings report for 30 September is due by 16 November and had not been filed when we checked on 9 November, so individual weights for the quarter are not yet public. Berkshire also owns 26.5% of Occidental and 27.5% of Kraft Heinz, carried in the equity method line. Abel's 2025 letter put its five Japanese trading-house stakes at $35.4bn at the end of 2025, with yen borrowing roughly equal to their $15.4bn cost, at an average cost of 1.2%.

Third quarter 2026: share repurchases $9.6bn; business acquisitions $6.5bn; capital expenditure $5.5bn; purchases of listed stocks $2.4bn; sales of listed stocks $2.8bn. Source: Berkshire Hathaway 10-Q for 30 September 2026, PDF pages 7-8 and 52 and 10-Q for 30 June 2026, page 7; figures with locators (CSV). The quarter's cash flows are our calculation: nine-month totals minus first-half totals.

Berkshire sold slightly more listed stock than it bought in the quarter. The acquisition line is Taylor Morrison, a homebuilder bought for $72.50 a share in cash, about $6.8bn of equity value, completed on 24 July. It joins Clayton's site-built homebuilders, and contributed about $1.1bn of home sales by 30 September (10-Q, PDF page 45). OxyChem, bought for about $9.4bn on 2 January, earned $192m before tax in nine months, after acquisition-accounting depreciation.

Operating earnings of $12.7bn were 5.8% lower than a year earlier (earnings release). Insurance underwriting earnings fell 43%, largely at GEICO, where claims became more frequent and more severe (10-Q, PDF page 36). Insurance investment income fell 14%, which the 10-Q attributes to lower short-term interest rates. BNSF, Berkshire Hathaway Energy and the manufacturing, service and retail group all earned more than a year earlier.

Why does Berkshire use this structure?

Abel's first annual letter, dated 28 February 2026, gives the reason for the cash in two parts. "Our cash and U.S. Treasury holdings now exceed $370 billion. While some of this capital is required to support our insurance operations and protect Berkshire against extreme scenarios, it also constitutes our dry powder." He adds: "We will always aim for ownership of productive businesses over U.S. Treasuries."

The balance sheet is meant to let Berkshire "invest when others are tentative or fearful, and stand firm when financial storms roll through" (same page). The 10-Q repeats a hard floor: repurchases stop if consolidated cash, cash equivalents and Treasury bills would fall below $30bn, and "Financial strength and redundant liquidity will always be of paramount importance at Berkshire" (PDF page 49).

On buybacks, Abel's letter says Berkshire will buy its shares "when they trade below our estimate of intrinsic value, conservatively determined, ensuring that repurchases enhance per-share value for continuing owners" (page 6). Under the current programme the chief executive decides after consulting the chairman. Since 18 September 2026 that is Howard G. Buffett; Warren Buffett became chairman emeritus and remains a director.

The third-quarter purchases averaged about $503 per B-share equivalent, 1.41 times book value per share at 30 September, by our calculation from the monthly repurchase table. They retired about 0.9% of the shares. Abel judged that price below intrinsic value; the filings give no estimate of that value.

On the stock portfolio, the letter calls the holdings in Apple, American Express, Coca-Cola and Moody's a concentrated approach that "will continue, with limited activity in these holdings" (page 15). Abel states that responsibility for equity investments "ultimately resides with me as CEO", with Ted Weschler managing about 6% (page 16). He is also plain about the limit of size: "At Berkshire's scale, the math of compounding works against us", with "a constant focus on managing downside risk" (page 17).

Historical performance versus MSCI World

A hypothetical $100 in Berkshire's B shares at the end of 2000 was worth $1,152.27 at 31 October 2026, against $633.27 in MSCI World net total return, both in US dollars. That is 9.92% a year against 7.41%, a gap of 2.51 percentage points, compounded over 25 years and ten months. Berkshire pays no dividend, so its share price is its total return.

Our cash-rate proxy, the monthly average three-month T-bill yield divided by 12, compounded at 1.81% a year over the same months. It approximates cash accrual rather than a traded bill portfolio's total return.

$100 at end-2000 became $1,152.27 in Berkshire B shares and $633.27 in MSCI World net by 31 October 2026; 2008: -32.1% and -40.7%; 2022: +3.3% and -18.1%; 2026 to October: +7.9% and +10.9%. Source: BRK-B closes from Yahoo Finance and MSCI World net USD month-end levels (index 990100); annual returns (CSV), monthly inputs (CSV).

The yearly bars show where the lead came from. Berkshire rose 7.3% in 2001 and fell 4.0% in 2002, when the index lost 16.8% and 19.9%. It lost less in 2008 and gained in 2022. It lagged in strong markets: 2003, 2009, 2019, 2020 and 2025 each favoured the index by more than ten points.

Berkshire's shares were more volatile than the index, yet earned enough more to post the higher Sharpe and Sortino ratios over the full period. The last row qualifies that. From November 2016 to October 2026, Berkshire still compounded faster, but its Sharpe ratio was below the index's; its Sortino was 1.16 against 1.12.

Berkshire B sits above the line of MSCI World and Treasury bill mixes: 16.9% annualised excess-return risk for a 10.9% mean annual return, against 15.3% and 8.4% for the index. Source: the same inputs plus FRED 3-month Treasury bill rate (TB3MS); frontier data (CSV). The curve is restricted to long-only MSCI World and bill mixes rebalanced each January, without leverage or costs; it is not an efficient frontier for Berkshire's own opportunities.

The crash record shows the shareholder's losses, without isolating what caused them. We measured each sell-off from MSCI World's daily peak to its trough, then took Berkshire's change over the same dates and its own worst fall within the broader calendar study window. Those windows are 2001-2003, 2007-2009, calendar 2020 and calendar 2022; Berkshire's own peak and trough need not match the index's.

Source: daily inputs (CSV) and episode results (CSV); the first window starts with MSCI's net series on 29 December 2000. In the 29 months when MSCI World fell more than 5%, it averaged -8.6% and Berkshire -4.9%; Berkshire fell in 23 of them. In the week to 6 November 2026, before the results were published, the index's worst day was -3.3% on 5 November, when Berkshire lost 1.2%.

Copio's reading

Is it optimal for geometric growth?

Berkshire's historical growth was strong; neither its optimality nor a growth sacrifice caused by its reserves is established. The objective here is expected log wealth for a long-term shareholder, over decades, subject to Berkshire's real constraints: insurance claims that must be paid in any market, regulated utilities and a railroad that need capital, and no dividend. The observed 9.92% CAGR is a result; it does not prove that no feasible alternative would have compounded faster.

Three tests from the evidence:

• Is the concentration paid for? Five stocks are 65% of the listed portfolio, but the listed portfolio is only a quarter of assets. Over 2001-2026 the shares returned more than the index with smaller drawdowns. Over the last ten years Berkshire still compounded faster despite a lower Sharpe ratio. Neither statistic establishes the return investors should expect from today's price.
• Does the reserve improve growth after its cost? The $352.3bn of net cash and bills outside the railroad and utilities was 1.97 times the $178.5bn insurance float at 30 September, by our calculation (10-Q, PDF page 39). The surplus earns bill rates, and insurance investment income already fell 14% in the quarter as rates fell. The reserve can support growth by avoiding forced sales, funding claims and enabling purchases. Its opportunity cost must be weighed against those benefits; the ratio to float alone does not measure the required reserve.
• Do costs and financing change the result? There is no management fee, and Berkshire's debt outside the railroad and utilities was $45.8bn, small against its $761.4bn of equity. Corporate tax is one friction, alongside operating costs, financing and acquisition prices: Berkshire's 10-K estimates that a 30% fall in its stocks would cut net earnings by $68.6bn after tax, on $297.8bn of holdings at the end of 2025.

The $9.6bn of buybacks at 1.41 times book is the clearest growth decision of the quarter. It raises each remaining owner's share of the businesses, and it is a bet on the same company the shareholder already owns. Whether it beats the T-bill alternative depends on the intrinsic value Abel estimates and does not publish. The decisive comparison still missing is Berkshire's return on newly deployed capital against bill yields over the next downturn.

Is it crash aware and protected?

Berkshire maintains substantial liquidity, but its shares have no contractual crash protection. The design examines the scenario that matters to an insurer: paying claims during a crash without selling assets or borrowing. The cash reserve, holding-company debt and $30bn buyback floor support that objective. Float is an accounting balance, not a schedule of claims: these figures alone cannot certify adequate liquidity in every extreme event.

The 10-K itself warns that losses "could be far worse due both to the nature of equity markets and the concentrations existing in our investment portfolio" than its 30% test (Form 10-K 2025, page K-61).

For the shareholder, the reserve does not pay out when markets fall. Berkshire B lost 53.9% from December 2007 to March 2009 and 29.6% in early 2020, close to the index's falls. The shares held up in 2001-2002 and 2022, but the filings examined here do not attribute that to the cash reserve. Our CalPERS study illustrates why milder losses must be judged alongside growth. Berkshire outgrew the index in this sample, but its record does not identify what it would have earned with a different reserve.

Is it diversified without sacrificing growth?

Its return sources differ from the index's, but most of them depend on the US economy. Over 2001-2026 Berkshire's monthly returns had a correlation of 0.49 with MSCI World and a beta of 0.55. That is imperfect co-movement, not independence or protection against a shared crash. Its earnings come from insurance underwriting, rail freight, regulated utilities, manufacturing and homebuilding, with technology mainly through its Apple and Alphabet stakes. October 2026 is an example: Alphabet fell 18.7%, MSCI World 0.8%, and Berkshire rose 8.9%.

November 2016 to October 2026: Berkshire B's correlation with MSCI World 0.63, with Alphabet 0.16, with Apple 0.39; American Express with Bank of America 0.73. Source: Yahoo Finance adjusted closes and MSCI World net USD; correlation matrix (CSV). These are observed monthly correlations in ordinary and stressed months together; they do not estimate how the stocks move in a crash.

The matrix shows two clusters among the top five: American Express and Bank of America move together (0.73), while Alphabet and Coca-Cola barely do (0.15). Berkshire's correlation with the index rose from 0.49 over the full period to 0.63 over the last ten years, so measured co-movement was higher in the recent window. Adding Berkshire would introduce single-company risk as well as exposure overlap; whether it improves a household's expected compounded growth needs a whole-portfolio comparison at current prices.

Copio does not grade Berkshire in its portfolio checker, because a single operating conglomerate holding individual stocks cannot be reproduced with UCITS ETFs. A household can assign each cash balance to a dated bill, emergency need or planned investment, then compare any remaining reserve with its long-term growth allocation. Include direct and ETF holdings when measuring the five largest company exposures, and replay 2008, 2020 and 2022 against spending needs. Berkshire's insurance liabilities do not supply a suitable cash percentage for a household.

For each proposed change, compare compounded wealth after fees, tax and withdrawals over the same horizon. Keep a reserve or diversifier when its funding or growth benefit justifies the cost; reject a lower-volatility alternative when it reduces expected growth without serving those constraints.

For comparison, Charlie Munger's Daily Journal portfolio was more concentrated, and Seth Klarman's Baupost also lets its cash rise when bargains are scarce. Our explainers on why a higher Sharpe ratio need not mean faster compounding and how long recoveries take after a crash give the background to these tests.