BRK-B · Forward model · Abel's price case

The Abel's price case, 20 quarters out

Each segment is projected from its own operating driver, rolled up into consolidated cash flow, and discounted back to a fair value per share. The assumptions below are editable — change them and every number on this page moves with them.

WHAT THIS PAGE IS: a two-column sum-of-the-parts wearing an enterprise-value engine's labels. READ THE LABELS SCEPTICALLY. The page says 'Enterprise value' where it means the present value of the OPERATING businesses only, and it says 'Net cash' where it means a $671,241M INVESTMENT PORTFOLIO. Fair value = PV(after-tax operating earnings) + the investment portfolio, which is the standard two-column way to value Berkshire. WHY NOT A NORMAL EARNINGS DCF. $12,684M of the basis quarter's $25,667M of net earnings is after-tax investment gains, $10.9 billion of it unrealised movement on shares Berkshire did not sell, and the release prints its standing warning that the figure 'is usually meaningless'. Capitalising it would be indefensible and this model does not. What makes the alternative work is that Note 24 of every 10-Q publishes revenue AND earnings before income taxes for eight operating segments — including 'Insurance — investment income' as its own separate segment — plus each segment's interest expense, capital expenditure, depreciation and income tax expense. That is exactly the separation a sum-of-the-parts needs, and it is better disclosure than any bank in this repo gives. THE SEVEN VERTICALS ARE THE EIGHT OPERATING SEGMENTS MINUS 'Insurance — investment income'. Nothing is split below a reportable segment. They foot to reported consolidated revenue exactly: $97,313M of verticals + $3,701M of insurance investment income + $794M of corporate, eliminations and other = $101,808M, the reported total, difference zero. On pre-tax earnings: $10,691M of verticals + $3,685M investment income + $16,077M investment gains + $248M equity method + $1,362M corporate = $32,063M reported. TWO LINES ARE EXCLUDED ON PURPOSE. Insurance investment income is the interest and dividend income of the portfolio that netCash already carries at market — capitalising a 99.6%-margin income stream AND adding the assets that produce it would count the same portfolio twice. 'Corporate, eliminations and other' goes for the same reason (the 10-Q says its revenue 'includes corporate interest income and dividend income not allocated to operating businesses') and for a second one: its pre-tax earnings are dominated by foreign-currency gains and losses on non-dollar senior notes, +$1,362M this quarter against -$247M a year ago, a series that reads +851, +726, -1,177, +1,400, -158, -247, +1,412, +250, +1,381, +1,362 since 2024 Q1. Carrying that forward would be modelling a currency. 'EBITDA MARGIN' IS PRE-TAX EARNINGS MARGIN. Every margin on this page is segment earnings before income taxes over segment revenue, already after that segment's operating expenses, DEPRECIATION and its own INTEREST EXPENSE. It is not EBITDA, it is not an operating margin, and it is not pre-provision anything. NETCASH IS NOT ZERO — and that is the whole point, unlike the BAC and GS models where it is a deliberate zero. $671,241M = cash and equivalents (Insurance and Other) $35,096M + U.S. Treasury Bills $324,905M + fixed maturity securities $17,034M + equity securities $323,779M + equity method investments $19,948M, total $720,762M, LESS $49,521M of deferred tax on unrealised appreciation. That deferred-tax figure is DERIVED: Berkshire discloses the 'investments, including unrealized appreciation' component annually only, at $48,411M against $212,389M of net unrealised equity gains at 31 December 2025, scaled here to the $217,258M outstanding at 30 June 2026. The $5,513M of railroad and utility working cash is excluded, because those businesses' earnings are inside the verticals. NO DEBT IS NETTED: the seven segments' own interest expense sums to $1,437M against $1,335M of consolidated interest expense, so every dollar of Berkshire's interest cost is already charged inside a vertical margin; subtracting the $128,599M of borrowings as well would take it twice and would cut fair value by about $59.88 a share. NO FLOAT IS NETTED: the $177.5 billion of insurance liabilities is not asserted to be free, it is separately capitalised as the underwriting vertical and its cost is currently NEGATIVE. The reader should still know the size — float is $82.65 a share, 16.4% of the current price. The portfolio is ADDED AT MARKET and never discounted, which is correct rather than lazy: a marketable portfolio carried at fair value has an expected return equal to its own cost of capital, so its present value IS its market value. THE MOST IMPORTANT THING ON THIS PAGE: netCash is 60.0% of fair value. Every slider, scenario, growth rate, margin and discount rate moves only the other 40%. The whole 7.5%-to-11.0% cost-of-equity band spans $533.54 to $505.88, a 5.5% range; the whole 10x-to-14x exit-multiple grid at 9.0% spans $497.73 to $544.41, a 9.4% range. A reader who expects the toggles to move the answer the way they do on an operating company will be disappointed, and should be. TAX. corporate.taxRate is 18.267%, DERIVED as the basis quarter's own ex-investment-income segment effective rate: segment income tax expense $2,578M less investment income's $625M share (insurance's $1,074M less underwriting's separately disclosed $449M) = $1,953M on $10,691M of pre-tax. That is on exactly the same basis as the pre-tax figures the verticals carry. It is NOT the 19.6% consolidated rate, which is dominated by tax on investment gains and has nothing to do with the operating businesses. The first-half version of the same calculation gives 17.80%; substituting it raises fair value by $0.39 a share, from $521.07 to $521.46 — the research brief's estimate of $2.85 for this swap was checked against the built model and does not survive. OVERHEAD CARRIES NONCONTROLLING INTERESTS AND NOTHING ELSE, the way the GS model's overhead carries preferred dividends and nothing else. The seven verticals' after-tax earnings sum to $8,738M before minority interest; the release's after-tax figures for the same seven segments, which are attributable to Berkshire shareholders, sum to $8,650M. The $88M difference is minority interest, chiefly $46M at BHE and $42M in the manufacturing and service subsidiaries; grossed up at 18.267% that is $107.7M of pre-tax equivalent, 0.1106% of $97,313M of revenue. Segment operating expenses, depreciation and interest are already inside each margin, so a second overhead line would double-count them. THE CALIBRATION GATE, WHICH IS THE TEST OF ALL OF IT. With no growth, no seasonal factor and no margin glide the seven verticals produce $8,649.3M of free cash flow against $8,650M of reported after-tax operating earnings attributable to Berkshire shareholders from exactly those seven segments (underwriting $1,731M + BNSF $1,558M + BHE $891M + manufacturing, service and retailing $4,470M) — error -0.008%. BAC calibrated at -0.05% and GS at -0.43%. With the full driver running the FIRST PROJECTED QUARTER prints $8,729.7M, +0.92%, and the whole residual is BHE's 1.1193 third-quarter seasonal factor plus one quarter of growth. If a future revision breaks this, the fault is one of three things: the ex-investment-income tax rate has drifted from 18.267%, a margin has been taken from the wrong window, or somebody has put the investment-income or corporate line back into the verticals. The target is NOT the $12,983M operating-earnings headline, which includes $3,059M of insurance investment income and a $1,274M 'Other' line that is mostly a currency gain. CAPEX INTENSITY IS EXACTLY ZERO ON EVERY VERTICAL, and this is the model's one real conservatism failure, quantified rather than hidden. Depreciation is already inside every margin, so the alternative would be to charge (capex less depreciation) — but that would put BHE at 22.0% of revenue against an 8.97% margin, driving that vertical to about -13% free cash flow margin for twenty quarters, and it would ALSO under-tax the whole company, because the engine levies tax on EBITDA less capex while the real tax is levied on pre-tax earnings in which depreciation, not capex, is the deduction. THE COST OF THE CHOICE: Berkshire spent $5,645M of segment capital expenditure against $3,461M of segment depreciation in the basis quarter, $2,254M more (ex-insurance), and $3,951M more over the half — about 24.3% of first-half after-tax vertical earnings. Most of that excess is externally funded at the subsidiary: railroad and utility borrowings rose a net $2,776M in the half against $3,164M of excess BNSF and BHE capex, 88% of it. The genuinely unfunded residual is roughly $588M a quarter, about 6.8% of quarterly vertical earnings, and THAT is the number by which this model's free cash flow overstates cash actually available. BNSF and BHE forecast approximately $8.6 billion of capital expenditure over the remainder of 2026 against $6.7 billion in the first half. SHARE COUNT IS PERIOD-END BASIC: 1,431,693 Class A equivalent shares at 30 June 2026 x 1,500 = 2,147,539,500, the same convention as the GS and BAC models. Berkshire has no options and no convertibles, so 'diluted' here is only the Class A/Class B equivalence. THE ALTERNATIVES AND THEIR EFFECT: average equivalent Class B shares for the quarter were 2,154,664,073, 0.33% higher, which would LOWER fair value to about $519.35; the 29 July cover-page count of 2,140,710,161 is 0.32% lower and would RAISE it to about $522.73. Both are an order of magnitude smaller than the 2.2% average-versus-period-end swing that flipped Goldman's headline. THE COUNT IS HELD FLAT ACROSS ALL TWENTY QUARTERS, which is correct arithmetic and understates terminal EPS: the model prints $17.95 of terminal four-quarter earnings per share, while the basis quarter's 1.72%-a-year buyback pace would retire 8.3% of the count in five years and print $19.57 of EPS, +9.0%. On the trailing-twelve-month pace of 0.124% a year the gap is only 0.6%. Both ends are live, because Berkshire's programme is price-conditional and open-ended. CAPITAL AND PAYOUT — THE INVERSE OF THE BANK PAGES. Berkshire has no dividend (none since 1967), no holding-company regulatory capital ratio, and no dollar-capped buyback authorisation: the programme 'does not specify a maximum number of shares to be repurchased' and has no expiry. Its ONE disclosed constraint is that Berkshire will not repurchase if that takes consolidated cash, cash equivalents and U.S. Treasury Bills below $30 billion — and those holdings were $365,514M at 30 June 2026, 12.2x the floor. So unlike BAC (which paid roughly 100% of first-half earnings to common) and GS (99.5%, but only because a CET1 buffer was being spent down from 14.3% to 12.9%), Berkshire's earnings genuinely ARE fully distributable and are deliberately NOT distributed: $4,762M of buybacks against $24,329M of first-half operating earnings is a 19.6% payout, 13.3% against net earnings. The constraint is neither capital nor policy — it is PRICE. Abel bought nothing in April, then Class B at an average $476.01 in May and $487.98 in June, $4,527.9M in all, under a programme that permits repurchase only 'at prices below Berkshire's intrinsic value, as conservatively determined'. BOOK-VALUE ANCHORS, which matter more here than at an industrial because two-thirds of the balance sheet is marked to market every quarter. Berkshire shareholders' equity was $747,910M at 30 June 2026, $348.26 a share. The tape at $503.70 is 1.446x book; this model's $521.07 is 1.496x. Strip netCash out of both sides and the tape pays $410,474M for the seven operating businesses against $31,694M of trailing-twelve-month after-tax earnings from them — 12.95x, which is the model's most important valuation anchor and is derived entirely from filings and the tape with no third-party comparable involved. At this model's fair value the same businesses are worth $447,776M, 14.13x. SEASONALITY METHOD. Factors are the ratio of quarterly segment revenue to a CENTRED four-quarter moving average over 2022 Q1 to 2026 Q2 (2023 Q1 to 2026 Q2 for Pilot), averaged by calendar quarter and renormalised to mean 1.0 — never a ratio to the year mean. A vertical is left ASEASONAL wherever the worst window-to-window spread meets or exceeds the signal. Three verticals carry factors (BHE 5.7x clear, manufacturing 2.0x, service and retailing 1.6x on a 1.6% range) and four do not (BNSF 1.03x — a tie; McLane spread meets signal; Pilot spread 3.3x the signal; underwriting 1.21x on a sub-2% range, rejected on materiality). The basis quarter is deseasonalised by its own second-quarter factor before the driver runs. HISTORY BASIS. Segment REVENUE is on one basis for all eighteen quarters and was checked line by line across the break (FY2023 BNSF revenue is $23,876M in both the pre- and post-ASU filings). Segment MARGIN is not: Berkshire adopted ASU 2023-07 in its 2024 Form 10-K, and before that segment earnings were struck BEFORE interest expense, a difference of $4,630M of pre-tax earnings for FY2023 alone. The 2023 and 2024 Forms 10-Q publish no per-segment interest, so pre-2025 quarters cannot be converted without inventing an allocation. Consistent quarterly MARGIN history therefore begins at 2024 Q1 — ten quarters — and every margin quoted on this page comes from that window. Every Q4 point in the revenue history is DERIVED and flagged estimated: Berkshire files no fourth-quarter report, so Q4 is the annual figure less the nine-month figure. Pilot's 2022 quarters are explicit estimated zeros because Berkshire did not consolidate Pilot until 31 January 2023. NO CONSENSUS EXISTS. Berkshire holds no earnings call, publishes no outlook, and is not covered by the estimate feeds this repo uses. No beat or miss is asserted anywhere on this page, and the horizon is this model's own judgement — deliberately the least important choice here, since 60% of fair value is the investment column that no projected quarter touches.

Where each case comes from

From cash flow to fair value

The projected path

Every row is projected. A year-over-year change is shown only where the quarter it compares against exists — an em dash means there is no comparable quarter, not a flat year.

Model revisions

Assumptions are marked to reality as each quarter prints. Every change is appended here, with the fair value the model produced at the time, so the model's own history stays visible.