Berkshire Hathaway

IThe business

Berkshire is an insurer that used its float, the premiums it holds before claims are paid, to acquire a railroad, a utility, dozens of manufacturers and retailers, and a stock portfolio.⁸ In 2025 revenue was $371.4B and operating earnings were $44.5B.R (reported) In the twelve months to June 2026 operating earnings were $48.0B and net earnings $85.8B, which includes $42.2B of investment gains.C (calculated)⁴⁵⁶ The company has about 387,800 employees and 51 non-insurance businesses.⁵⁸

Two terms keep one meaning on this page. “The letter” is Berkshire’s 2025 shareholder letter, signed by Greg Abel.⁸ “Cash” is the cash, cash equivalents and Treasury Bills of the insurance and other businesses, net of unsettled purchases: $359.2B at 30 June 2026 and $348.5B after July’s repurchases and acquisitions.C (calculated)⁴

Operating earnings is Berkshire’s own after-tax measure. It leaves out investment gains and losses, which swung from a $62.3B gain in 2021 to a $53.6B loss in 2022, and impairments.R (reported)²⁶²⁷

The $48.0B comes from six places: insurance investment income $12.0B, manufacturing, service and retailing $14.7B, underwriting $7.4B, BNSF $5.7B, BHE $4.2B, and a line called Other $4.1B.C (calculated)⁴⁵⁶

How it makes money

• Insurance underwriting. GEICO, Berkshire Hathaway Primary and Berkshire Hathaway Reinsurance earned $88.9B of premiums in 2025 and made $9.5B before tax from underwriting alone.R (reported) GEICO’s premiums earned were $44.5B at an 84.7% combined ratio, the primary group’s $18.7B at 95.8%, and the reinsurance group’s property and casualty book $20.4B at 84.5%. The whole property and casualty book ran at 87.1%, against 90.7% over five years and 93.0% over ten.⁵⁸
• Insurance investment income. Interest and dividends on the cash and stocks the insurers hold: $12.0B in the twelve months to June 2026, down 8.5% in 2025 and 8.3% in the first half of 2026 as short-term rates came down.C (calculated)⁴⁵⁶
• BNSF. The railroad runs over 32,500 route miles in 28 states. In 2025 it had revenue of $23.5B, an operating margin of 34.5% and after-tax earnings of $5.5B, and it paid Berkshire $4.4B of dividends.R (reported)⁵⁸
• Berkshire Hathaway Energy. Revenue was $26.3B and after-tax earnings $4.0B. Its customers pay rates that average 24% below the national level, and it spent $10.6B on capital in 2025 against $4.2B of depreciation.R (reported)⁵⁸
• Manufacturing, service and retailing. $14.7B after tax in the twelve months to June 2026, from Precision Castparts, Lubrizol, Marmon, IMC, Clayton, Shaw, Johns Manville, NetJets and others.C (calculated)⁴⁶⁸
• The investments. At 30 June the insurance and other businesses held $359.2B of cash and Treasury Bills and $323.8B of stocks at market, bought for $106.5B.R (reported) Five names are 66% of the stocks: Apple, American Express, Coca-Cola, Alphabet and Bank of America.⁴

Seven of the businesses

Seven of Berkshire's businesses. Figures are for 2025, from the 10-K and the 2025 letter.

BNSFOver 32,500 route miles in 28 states, $5.5B of after-tax earnings in 2025 and $4.4B of dividends to Berkshire.⁵⁸

BHE: PacifiCorp$4.0B of after-tax earnings in 2025, on $10.6B of capital spending against $4.2B of depreciation.⁵

GEICO$44.5B of premiums earned in 2025 at an 84.7% combined ratio; third in US auto with 11.6%.⁵

NetJetsNearly 1,100 aircraft in over 150 countries.⁸

Pilot675 travel centers; $42.2B of revenue and $0.19B of pre-tax earnings in 2025.⁵

Dairy QueenAbout 7,800 franchised restaurants.⁵

DuracellBooked refundable production credits for 2023 to 2025 in 2025; Berkshire does not give the amount.⁵

What changed in 2026

Greg Abel became chief executive on 1 January. Berkshire resumed buying its own stock on 4 March after buying none in 2025.⁹¹³ It spent $9.4B on OxyChem in January and put $10.0B into Alphabet in June.⁴¹⁹ Sixteen issuers in the 31 December 2025 13F, among them Visa, Mastercard, UnitedHealth and Amazon, are not in the 30 June 2026 13F.¹⁴¹⁵ The table lists what happened after the last balance sheet.

Cash and stocks at market prices are 55% of the price; the other 45% buys the businesses.¹⁴⁵⁷⁶⁰

IIThe moat

I think Berkshire’s moat is a cost of capital: a cost advantage in the cost of funds. Its insurers hold $177.5B of float, other people’s money that they pay out slowly, and in eight of the last ten years they were paid to hold it, because underwriting made money before a dollar was invested.C (calculated)⁴⁵⁸ The U.S. insurers hold about $333B of statutory surplus, the capital that regulators measure, and are rated AA+ by S&P and A++ by A.M. Best.⁵⁸ The balance sheet is the second support: after July’s repurchases and acquisitions the insurance and other businesses hold $348.5B of cash and Treasury Bills against $20.4B of notes the parent issued, and the parent does not guarantee BNSF or BHE debt.C (calculated)⁴⁵

In my base case the float moat is worth about $99 a share, 21% of the base: $72 from not counting the $155.2B of float as debt and $27 from $4.5B of underwriting profit capitalised.E (estimate)⁴⁵

I think the float moat is steady, the rail moat is widening slowly, and the investing edge has narrowed. Berkshire’s per-share market value beat the S&P 500 with dividends by 1.0 point a year from 2006 to 2015, 8.3% against 7.3%, and trailed it by 0.5 point from 2016 to 2025, 14.3% against 14.8%.C (calculated)⁸ The float grew $5B in 2025 and about $1.1B in the first half of 2026 on the 10-Q’s unrounded figures, and the letter says Berkshire will write less property and casualty business, because pricing is softening and new capital has entered.⁴⁵⁸

At $359.2B the cash is flexibility in a crash and a drag when nothing is bought. A three-month Treasury Bill yielded 4.22% before tax in Q4 2026, 3.3% after tax, against the 9.0% I use as a cost of capital. The gap is about $20B a year, $9.51 a share, the return the cash must beat to justify holding it. No case deducts it, because I carry the cash at face. The option is a 12% return on $139B, worth $22 a share in the bull case, and I leave it out of the base.C (calculated)⁴⁵⁷

At Berkshire’s scale, the math of compounding works against us.

The float still costs nothing. The open question is use: $36B deployed this year and $348.5B of cash left.

IIIThe competitive advantage

Berkshire meets named rivals in two places where I can measure it: auto insurance against Progressive and Allstate, and freight rail against Union Pacific and CSX. In both it earns good returns, and in both a rival is ahead on one measure. Investing is a third place it trails, covered in the moat and capital allocation sections.

Where Berkshire has led

Underwriting margin comes first, and the lead has gone. GEICO’s 2025 combined ratio of 84.7% beat Progressive’s 87.4%, which includes commercial lines and so is not like for like, and was level with Allstate Property-Liability’s 85.2%.C (calculated)⁵⁵³⁵⁶ In the first half of 2026 GEICO ran 89.3% against 81.7% a year earlier, while Progressive ran 88.3% against 87.6%. GEICO’s losses rose 10.1% on premiums earned up 3.0%, and its pre-tax underwriting earnings fell 40% to $2.4B.C (calculated)⁴⁵³ I did not compare Allstate’s first half. The letter ties the record to capital: Berkshire keeps underwriting risk instead of buying reinsurance, which only works with $333B of statutory surplus behind it.⁵⁸

The second lead is over CSX. BNSF’s operating margin rose from 32.0% to 34.5% in 2025, while CSX’s fell 4.0 points to 32.1%. Union Pacific’s was 40.2%, up from 40.1%.C (calculated)⁸⁵⁴⁵⁵

Where it trails

GEICO has grown more slowly than Progressive, and the gap widened. Progressive’s premiums earned grew 20.7% in 2024, 15.3% in 2025, to $81.7B, and 7.1% in the first half of 2026. GEICO’s grew 7.6%, 5.3%, to $44.5B, and 3.0%.C (calculated)⁴⁵⁵³ The letter says rate increases restored GEICO’s margins at the cost of lower retention.⁸

BNSF is 5.7 points behind Union Pacific. The letter calls the gap too wide and says each point of margin is worth about $230M of operating cash flow, so the whole gap is worth about $1.3B a year.C (calculated)⁸ A Union Pacific and Norfolk Southern merger would link a western network to an eastern one; Berkshire says it wants full access to Eastern markets.⁸

IVManagement

Greg Abel has run Berkshire since 1 January 2026, after leading the non-insurance operations as Vice Chairman since 2018. He ran Berkshire Hathaway Energy before that.⁸⁹ Ajit Jain runs insurance, and the letter calls him peerless at pricing risk. Chuck Chang became chief financial officer on 1 June and Marc Hamburg retires in 2027. Ted Weschler manages about 6% of the investments; Todd Combs announced on 8 December 2025 that he was leaving to join JPMorgan Chase.⁸¹¹

Warren Buffett is Chairman Emeritus and a director, with about 13.2% of the economic interest and about 30% of the votes; his shares go to philanthropy over roughly ten years after his death.C (calculated)⁸⁹²² Howard Buffett is the non-executive chairman.⁹¹⁶

We will always aim for ownership of productive businesses over U.S. Treasuries.

Abel’s record is the businesses he ran. Their after-tax earnings, BNSF and BHE plus manufacturing, service and retailing, grew from $17.2B in 2018 to $23.1B in 2025, or 4.3% a year.C (calculated)⁷⁴⁰ That is a record of steady operators of mature businesses. He has run capital allocation for ten months.

Capital allocation

The best deals were stakes in public companies and the railroad. The recent operating purchases have not yet earned what they cost.

Berkshire repurchased $77.9B of its own stock from 2018 to 2024, $51.8B of it in 2020 and 2021, when the year-end price was 1.21 and 1.31 times book value.C (calculated)^{25262728293031} It bought nothing in 2025, resumed on 4 March 2026 and spent $4.76B by the end of June at 1.41 to 1.46 times the prior quarter-end book value.C (calculated)¹⁰⁴¹³ The rule is to repurchase when the price is below intrinsic value, conservatively determined, and to stop if cash and Treasury Bills would fall under $30B.⁴⁸ The prices it paid, $476 to $489 a share, tell me management’s estimate was above those prices on those days. At today’s price and my base value, each $10B of repurchases costs continuing owners about $1.4B of value.E (estimate)

In 2024 Berkshire sold $134B more stock than it bought; in the first half of 2026 it bought $11.6B more than it sold.⁵⁴ So far in 2026 it has deployed at least $36B: OxyChem, net stock purchases and repurchases in the first half, July’s repurchases, Taylor Morrison and Lennar.C (calculated)^4131720 At that pace, about $45B a year, putting $139B to work takes three years and the whole pile eight.C (calculated)

The longer record is good. From the end of 2015 to the end of 2025 Berkshire earned $516.3B and paid no dividend. Its market value rose $758.9B and it spent $77.9B on repurchases, so each dollar earned came with $1.62 of value to shareholders.C (calculated)²⁴³⁵³³ The filings do not give invested capital by segment, so I cannot state a return on tangible capital for the non-insurance group.

Incentives

Abel’s 2025 pay was a $22.0M salary and $17,500 of retirement plan contributions, with no bonus and no stock; his 2026 pay as chief executive is not yet disclosed.⁹ Berkshire’s compensation policy says neither its profit nor its share price is to be considered in any executive’s pay. Chang’s salary is $8.0M, and Buffett was paid $100,000 for more than forty years.⁹¹²

That removes the incentive to chase the share price, and it also removes any pay for return on capital. Abel is a trustee of a revocable trust that holds 249 Class A shares, and the proxy says he disclaims any beneficial interest in them. At $813,595 a Class A share they are worth about $203M. I do not count them as his stake, so I leave out the $15.3M the trust paid for 21 more of them on 4 March.C (calculated)¹⁹²¹ Since March, Chang bought two Class A shares for about $1.5M and General Counsel Mike O’Sullivan bought 1,024 Class B shares in two purchases for about $0.5M, and on 28 September Jain sold 39,700 Class B shares at $503.09, about $20M.C (calculated)²³ Officers bought about $2.0M and Jain sold about $20M, and I give it no weight.

The pay policy rewards steadiness: Abel’s cash pay does not move with the share price, profit or the return on a deal. I think that pulls toward waiting. I can’t say how much of his own money is in. The only purchase tied to him is $15.3M of Class A shares bought by a trust, and the proxy says he disclaims it.⁹²¹ Over the last full cycle each dollar earned came with $1.62 of value, and the big repurchases came at 1.21 and 1.31 times book.C (calculated)^2435332829

I rate the operators high; I cannot yet rate the capital allocator. The first test is what Abel does with the $348.5B of cash left after July, and what he pays for Berkshire’s own stock.⁴⁸

VFinancials

Operating earnings rose from $17.6B in 2016 to $44.5B in 2025. Insurance explains two-thirds of the rise since 2021: underwriting went from $0.9B to $7.4B and insurance investment income from $4.8B to $12.0B, $13.7B of the $20.4B increase to the twelve months to June 2026.C (calculated)²⁶⁴⁵⁶ A hard market and Treasury Bills at 4% or more drive it, and both are cyclical.

Berkshire pays no dividend and retained everything it earned in 2025. Operating cash flow was $46.0B and capital spending $20.9B, which leaves $25.0B. That understates the cash earned, because discount accretion on Treasury Bills, $12.0B in 2025, is deducted in operating cash flow and arrives later as investing inflows.C (calculated)⁵ BNSF and BHE spent $14.4B of the $20.9B and plan about $15B in 2026, against $6.9B of depreciation in 2025.⁵⁸ Most of it is growth: rate base at BHE, capacity at BNSF.

For the ten years to the end of 2025, book value per Class B share rose from $103.29 to $332.55, 12.4% a year, while the price rose from $132.04 to $502.65, 14.3% a year, so the multiple of book expanded from 1.28 to 1.51 times. Book value was $348.26 at 30 June 2026.C (calculated)^245433233

The stocks

The twelve largest lines in the 13F are $278.9B of the $299.3B total at 30 June. The eleven biggest holdings that Berkshire carries at market earned a share of about $11.5B in their latest fiscal years, paid it $3.2B, and kept $8.4B inside the companies.E (estimate)^{144243444546474849505152} That retained $8.4B is not in operating earnings, which is why I value the stocks at their market price.

Accounting to watch

1. What operating earnings leaves out. In 2025 it left out $8.3B after tax of impairments on Kraft Heinz and Occidental; net earnings were $67.0B against operating earnings of $44.5B.⁷²⁶
2. Underwriting. Underwriting profit after tax was $7.4B in the twelve months to June 2026 and averaged $4.5B over five years and $2.4B over ten.C (calculated)⁴⁵²⁶ Catastrophe losses after tax were $850M in 2025, $1.2B in 2024 and $725M in 2023. Unpaid losses are $122.9B, plus $30.0B on retroactive reinsurance contracts, which cover losses from events that have already happened.⁴⁵
3. Deferred tax. The balance sheet carries $90.2B of income taxes, mostly deferred.⁴⁵
4. Tax credits and one-offs. BHE’s energy production tax credits were $2.06B, $1.88B and $1.66B in 2025, 2024 and 2023, and its income tax line was a $1.8B benefit on $2.3B of pre-tax earnings in 2025. The 2025 law accelerates their phase-out.R (reported)⁶⁶⁵ Duracell booked production credits for 2023, 2024 and 2025 in 2025, and the first half of 2026 includes tariff refunds. The filings I read do not give the amounts, so I did not remove them from trailing earnings, which favours the cases. Every $1B of unwinding in manufacturing, service and retailing earnings is about $6 a share.C (calculated)⁵⁴

Cash taxes, a small restatement of 2021 and 2022, and currency on yen and euro debt are real, but none moves value.^252636375

VIValuation

How an owner loses money here

Before the numbers, I ask how an owner loses money here. I see three ways. The insurance cycle turns: underwriting is $7.4B after tax against $4.5B over five years, and GEICO’s combined ratio rose to 89.3% in the first half from 81.7% a year earlier.C (calculated)⁴⁵ The cash earns little: OxyChem earned $121M before tax on $9.4B in the first half, about 2.6% a year.C (calculated)⁴ The stocks fall: five names are 66% of $323.8B, and a 30% fall takes my base to $430.C (calculated)⁴

What an owner can take out is less than the earnings line. Berkshire pays no executive in its stock, so there is no stock pay to take off.⁹ Operating cash flow in 2025 was $46.0B, plus $12.0B of Treasury Bill accretion that arrives later, less $20.9B of capital spending, or only $13.5B if depreciation alone is needed to stand still. I can’t split that, so the range is $37.0B to $44.5B. That is $17.3 to $20.8 a share, 3.2% to 3.8% of the price. Net earnings were $67.0B, so an owner’s cash is 34% to 45% below them.C (calculated)⁵

I can’t know what Abel earns on the next $139B, or whether the float is debt, which is $72 a share of my base. So I show $394 to $561, not one number.E (estimate)⁴⁵

What the price implies

At $542.19, Berkshire’s market value is $1,161B. That is 24.2 times operating earnings of the last twelve months and 1.56 times book value of $348.26 a share, above the 1.21 to 1.51 range of the last ten year-ends.C (calculated)¹⁴⁶

Take the investments out and ask what is left. At Q4 2026 prices the cash is $348.5B, the stocks $336.0B, the Occidental, Kraft Heinz and Berkadia holdings $25.5B, and bonds and post-June purchases at cost $24.4B. Subtract $49.5B of tax on the gains as if everything were sold, the $20.4B of notes Berkshire itself issued, and $22.3B of retroactive reinsurance liabilities net of their deferred charge, and $642.3B remains, 55% of the market value.C (calculated)⁴⁵⁵⁷⁶⁰ The price pays the other $518.3B for the businesses, which is 17.8 times their $29.1B of normalised earnings.

I repriced the stocks from 30 June to Q4 2026 closes: the 13F positions are worth $316.0B against $299.3B, a gain of $16.8B, or $7.83 a share.C (calculated)¹⁴⁶⁰ The $45.1B outside the 13F stays at June values. The model deducts the cash paid after 30 June and adds none of the earnings from July to October, which would add about 1%.

Run the model backwards. Holding my base case for everything but the discount rate, the price needs 6.5% on the operating businesses, 1.5 points over the 5.00% ten-year Treasury and 2.5 points under the 9.0% I use.C (calculated)⁵⁷ The businesses are 45% of the price. Paying $542.19 for my base cash flows returns about 7.0% a year on the whole company, a blend of 6.5% on the businesses, 9.0% on the stocks, and 3.3% on the cash until $139B of it is put to work at 9.0%. That is 2.0 points over the Treasury and 1.4 points under the 8.4% the CAPM gives for the whole company.E (estimate)

The parts at their peers’ own multiples add up to $492 to $534, below the price, and my base is below the midpoint of all three cross-checks.⁶¹

My base case ends year ten at 11.4 times earnings, because new capital then earns only the discount rate, while Berkshire’s listed rail and utility neighbours trade at 20 to 32 times. The price needs 21.2 times, so the dispute is the terminal value. Growth is a weak lever because it costs capital: moving BNSF and the manufacturing group from 2.5% to 5.5% a year moves the base by $14.E (estimate)⁴⁵

I applied every listed peer’s multiple, including Allstate’s 6.2 and CSX’s 32.4. Applying the insurers’ multiples to underwriting profit alone is generous, because their earnings include investment income. The manufacturing group has no listed twin, so I leave it at my value. The market may pay more for a railroad and a utility than my 9.0% allows, and the Union Pacific and Norfolk Southern merger may be lifting rail multiples; I could not verify that.⁸⁵⁶⁶¹

Bear, base and bull

I value Berkshire in two parts and add them. The investments have market prices. The businesses I value by discounting their after-tax earnings, less what growth needs in capital, at 9.0%, with every cash flow dated from 30 June 2026 and not rolled forward to Q4 2026, which would add about 1% to each case. Every input is in the tables on this page; the case values come from my Q4 2026 model.⁴⁵

In every case I count the retroactive reinsurance liabilities, $22.3B net of their deferred charge, as owed, because they run off. I do not subtract the rest of the float, $155.2B, in the base or the bull, because it has risen every year for ten years and the insurers pay claims out of the premiums that replace it.⁵⁸ This is the largest judgement on the page: counting all of it as debt takes $72 a share, or 15% of my base, off the base.E (estimate) It matters because float grew only about $1.1B in the first half of 2026. In the bear case a third of the rest, $51.7B, runs off.⁴⁸

I capitalise only earnings that do not already sit in the investments, so I leave out insurance investment income and Other. Segment earnings are after the interest on the debt that funds each segment, so I do not subtract the $85.3B that BNSF and BHE borrowed, or the $22.9B that funds Clayton, leasing and other subsidiaries, a second time.⁴⁶

BHE retains all it earns, as growth in the rate base on which the regulator allows a return. I assume it earns exactly my 9.0% on what it retains, so its core earnings of $2.1B are worth $24.5B. The other $2.1B is energy production tax credits, mainly from wind. I assume they last ten more years at today’s level and then stop, worth $13.2B: a simplification of credits that expire from 2014 to 2035 at MidAmerican and are renewed by repowering.⁴⁵⁶⁶

I discount at 9.0%. The capital asset pricing model on Berkshire’s own ten-year beta of 0.74 against the S&P 500 gives 8.4%: the 5.29% risk-free rate Damodaran used when he set his 4.20% equity risk premium on 1 October 2026, plus that premium times the beta. Its five-year beta of 0.60 gives 7.8%.C (calculated)⁵⁸⁵⁹ That beta is low because the investments are 55% of the market value. With a beta of 0 for the cash and 1.0 for the stocks, the businesses carry a beta of 1.04 and a cost of equity of 9.7%. I use 9.0%, close to the midpoint of the two. At 9.7% the base is $454.E (estimate)

The next best use of the same dollar: the ten-year Treasury pays 5.00%, my base returns 6.6% a year to mid-2036, and Prosus, the closest holding company on this site, returns 9.8% a year on its base, from a price 22% below it.⁵⁷ Here the price is 16% above. The business is steady, and the price already pays for much of the good outcome. It needs four things together: underwriting at $7.4B, 5.5% growth on 18% returns, those returns lasting past year ten, and the cash earning 12%. At my guesses of 50%, 40%, 50% and 30%, that is 3%, perhaps 5% to 8% because they move together.E (estimate)

Two shocks sit outside every case, so I price them separately.

The base case starts underwriting at $4.5B, the five-year average, against $7.4B in the last twelve months. BNSF and the manufacturing, service and retailing group grow 4.0% a year, a little under the 4.3% the non-insurance businesses averaged from 2018 to 2025, and new capital earns 15%. That 15% is my estimate, and the filings bracket it. The two groups’ after-tax earnings rose $4.5B, from $14.6B in 2018 to $19.1B in 2025, on $14.2B of capital spending above depreciation from 2018 to 2024, a return of 32% before any acquisition. Charging $17.0B of acquisitions as well (2018 to 2024, without Alleghany in 2022), some of which belong to BHE and insurance, it falls to 14.6%. Price rises and Duracell’s catch-up credits are in the earnings, so the first figure is an upper bound and the second a lower one, and 15% sits at the low end.E (estimate)⁴⁵²⁶²⁹

The bear case is the insurance cycle turning and the cash earning little: underwriting at its ten-year average of $2.4B, BHE earning the 7% a utility earns when wildfire costs and rate lag bite, and the $139B of cash put to work earning 6%, nearer the 4.1% operating earnings yield on Berkshire’s own stock than to my hurdle. In the base that cash earns 9.0%, which is neutral. The bull case is what has to go right: underwriting holding at the trailing $7.4B, BNSF closing part of its margin gap to Union Pacific, 18% returns that last past year ten, and the cash earning 12%, the kind of return a downturn offers. The tax on the stock gains is all due in the bear and half its face in the bull.E (estimate)⁸

The value after year ten is 52% of the operating value in my base case, 50% in the bear and 59% in the bull.C (calculated) No case deducts a further PacifiCorp wildfire loss beyond the bear case’s lower BHE return; the filing says more are reasonably possible and cannot be estimated.⁴ The bear case does not include a stock-market fall. With a 30% fall in the stock book it is about $357, which is $7 above my margin-of-safety price.C (calculated)

Base value per share by discount rate and by yearly growth, years 1 to 10, of BNSF and the manufacturing, service and retailing group. No cell reaches the price. The discount rate moves the base by $19 to $24 a point, because only the businesses are discounted.C (calculated)⁴⁵⁵⁷

What moves the base case, one input at a time, per share, largest swing first:

Swapping the bull inputs into my base one at a time, in the order the chart shows, takes it from $467 to $561. In that order the price is covered only at the sixth swap, and after five the value is $540. The order matters: no three of the six swaps cover the price, and only one set of four does, underwriting at the trailing $7.4B, higher growth and returns, returns kept after year ten, and the cash at 12%. Take away any one of those four and the bull case ends between $534 and $541, short of the price.¹⁴⁵

Three cross-checks sit outside the cases. I weigh them against the sum of the parts here.

The sum of the parts is the lowest of the four. I rank it first because it is the only one that charges the businesses for the 9.0% I would require, and the other three take market prices as given. My cases are 1.13, 1.34 and 1.61 times book. Berkshire paid up to $489, 4.7% above my base of $467, so management’s value on those days was higher than mine. Measured against $489, the price is 11% above it, not 16% above my base, and 3% below my bull case.C (calculated)¹³⁴²⁴

The margin of safety

• Price$542.19Close, Q4 2026 [1]
• Base value$467My estimate E (estimate)
• Margin of safety below$35025% under my base value C (calculated)

I set the margin of safety at 25% below my base value, which gives $350. My rule is 25% for a business with net cash and no customer above a tenth of revenue, 30% when one cycle or two customers drive results, and 35% when fixed charges such as leases sit ahead of the shares. Berkshire has net cash, no customer above a tenth of revenue, and an insurance cycle that drives a minority of earnings.⁴⁵

Investments priced at the market are 64% of my base value, so most of the 25% falls on assets that already have a market price. At $350 the market would pay $107B for businesses that earn $29.1B, 3.7 times, which is a crash price. Applied only to the parts without a market price, 25% of the businesses’ $167 a share is $42, and a 30% fall in the stocks is $37 a share. That gives $388; applied to the businesses alone, 25% gives $425. I keep $350 because it is my rule; the base with all the float counted as debt, $394, is a downside it also pays for.C (calculated)⁴⁵

The price is 55% above that line, so the stock would have to fall about 35% to reach it. The lowest close of the past year was $465.40 on 22 April 2026, 33% above it, and the last month-end close below $350 was October 2023.C (calculated)³⁶⁵

Where this could be in 5 and 10 years

For the base to stand at five years, BNSF and the manufacturing group must grow 4% a year on 15% returns on new capital, underwriting must settle near $4.5B, and $139B of the cash must earn 9%. Value then grows about 8% a year, under my 9.0% discount rate, because the other $209B of cash earns 3.3%. At ten years the base also needs its terminal value, 52% of the operating value, which assumes new capital then earns only 9.0%.E (estimate)

The bear and the bull differ on a few numbers. The bear has underwriting at $2.4B, growth of 2.5% on 12% returns, and cash earning 6%. The bull has $7.4B, growth of 5.5% on 18% returns that last past year ten, and cash earning 12%. The gap grows from $167 a share today to $257 at five years and $395 at ten.C (calculated)

The most likely way the base is wrong is the cash. My base needs $139B of $349B to earn 9%, and that is a decision I do not control. At 6%, the base is $22 lower today and $51 lower at ten years.C (calculated)

These are estimates of value, not forecasts of the share price. The market price can sit above or below value for years.

VIIThe case for the price

What most people miss

Two facts in the filings cut Berkshire’s way. First, the cash can move. Regulators limit what insurers pay up to a parent, so the money looks trapped. In 2025 Berkshire’s insurers could pay $31B without approval, and they paid $29B. The 2026 limit is about $33B, near a tenth of the pile.⁸⁵

Second, the Japanese stakes were funded in Japan. Berkshire borrowed about what it invested, in yen, at an average 1.2%. The five trading houses cost $15.4B and were worth $35.4B at the end of 2025. They paid $862M of dividends against about $185M of interest.C (calculated)⁸ So a $20.0B gain, before tax, took very little of Berkshire’s own money.

The bull case

The bull case is that I charge too much for safety and give too little credit to a good operator. Chris Bloomstran’s firm, Semper Augustus, has owned Berkshire since 2000. His letter of 25 February 2026 averages four methods to $570 a Class B share at the end of 2025, above my bull case. He says BNSF and BHE may under-earn by up to $2B a year, and gives no credit for it. He expects Abel to commit the cash in size in a deep recession or crisis, or when a rare big deal appears.⁶⁷

MarketBeat lists Brian Meredith at UBS raising his target to $628 from $604 on 22 October 2026.⁶⁸⁶⁹ Invezz reported in July that he put intrinsic value near $800,000 a Class A share, about $533 a Class B, which is below today’s price.⁷⁰

Barron’s and CFO.com reported Bill Ackman on a podcast in February 2025 saying BNSF is probably the least efficient railroad to run and Abel is more of an operator.⁷³⁷⁴ In May 2025 Insider Monkey reported him telling CNBC that Abel is a strong operator and allocator, though perhaps not Buffett’s match at acquiring businesses.⁷¹ Warren Buffett wrote on 10 November 2025 that Abel “has more than met the high expectations I had for him”.⁷² Abel’s letter says Berkshire will be disappointed without a substantial BNSF margin gain over the next few years.⁸

Against reality

If four things happen together, the price is possible. Underwriting stays at $7.4B after tax. BNSF and the manufacturing group grow earnings 5.5% a year on 18% returns. Those returns last past year ten. The next $139B of cash earns 12%. In my model that comes to about $548, 1% above $542.19.E (estimate) Without any one of the four, the other bull inputs reach only $534 to $541.¹⁴⁵

Only the first is met today, and it may not last. Underwriting is $7.4B, against $4.5B over five years and $2.4B over ten. GEICO’s combined ratio, its costs as a share of premiums, rose to 91.2% in the second quarter.⁴⁵ The second is mixed. BNSF’s after-tax earnings rose 9.5% in the first half and the manufacturing group’s 15.1%, but OxyChem and tariff refunds helped the second. BNSF’s margin was 34.5%, up from 33.6% a year earlier, but level with all of 2025 and 5.7 points behind Union Pacific’s. In the second quarter it slipped to 34.6% from 35.2% as fuel costs rose. The filings do not show the 18%.C (calculated)⁴⁵⁴ The cash is the furthest away. OxyChem cost $9.4B and earned $121M before tax in the first half, about 2.6% a year, after acquisition costs. At about $45B a year, $139B takes three years.⁴ Returns past year ten cannot be checked yet.

What they have to hit

Berkshire publishes every row each quarter. Thresholds match the signposts in the next section.

Where I come down

Two things moved me: Berkshire paid up to $489 for its own shares, 4.7% above my base, and the insurers’ dividends show the cash is easier to reach.C (calculated)⁴⁸ The 12% did not, because I want to see it earned first. BNSF near 37% with repurchases at $540 or more would raise my values, which stay at $394, $467 and $561 for now.

VIIIRisks, and what would change my view

The risk I weight most is the one Abel names himself: that size makes good returns hard to find. The cash pile is the test of it.

1. Capital allocation. $348.5B of cash earns 3.3% after tax. The drag of about $20B a year, described in the moat section, is the return it must beat, and no case deducts it. The pile was $359.2B at 30 June, down from $369.0B at year-end.⁴⁸ Each point of return on the $139B I assume is put to work is worth $7 a share.C (calculated)
2. People. Buffett gave up the chair on 18 September, the chief financial officer changed on 1 June and the general counsel’s post is new. Capital allocation and insurance rest on Abel and Jain.⁸⁹¹⁶¹¹
3. The insurance cycle. The letter says pricing is softening in primary, commercial and reinsurance lines, and that Berkshire will write less. GEICO’s combined ratio rose to 89.3% in the first half of 2026. Underwriting profit after tax of $7.4B compares with a ten-year average of $2.4B, and it is the largest swing in operating earnings.C (calculated)⁴⁵⁸
4. Interest rates. Insurance investment income after tax fell 8.5% in 2025 and 8.3% in the first half of 2026. A one-point fall in yield on $324.9B of Treasury Bills alone is about $3.2B a year before tax.C (calculated)⁴⁵
5. A fall in stocks. Five holdings are 66% of $323.8B. The 10-K puts the effect of a 30% fall in equity prices at $68.6B of after-tax earnings, on the $297.8B held at 31 December 2025.⁴⁵ On the $336.0B at Q4 2026 the same fall takes $37 a share off my base, and the bear case with it is $357.C (calculated)
6. Utilities and tax credits. PacifiCorp’s insurance recoveries for the 2020 fires are exhausted, and the James verdicts total about $1.25B to 201 plaintiffs. The Oregon Court of Appeals reversed the Phase I verdict in April 2026; the Oregon Supreme Court hears argument on 3 November 2026 and the damages trials are stayed. About half of BHE’s after-tax earnings are production tax credits that are being phased out. HomeServices faces antitrust cases, two in Texas asserting about $9B.C (calculated)⁴⁵
7. Rail. A Union Pacific and Norfolk Southern merger changes access to Eastern markets, and BNSF’s margin is 5.7 points behind Union Pacific’s.C (calculated)⁸⁵⁴
8. Structure. Class A shares carry the votes and Class B has one ten-thousandth of them, and the new chairman is the founder’s son.⁴⁹¹⁶

Our investment in Kraft Heinz has been disappointing.

What would move me

The first tests come in early November: the Oregon Supreme Court argument on 3 November and the third-quarter 10-Q, which will show the 30 September balance sheet, any repurchases after July, and the first quarter with Taylor Morrison, which joined on 24 July.⁴