Berkshire Hathaway BRK-B

EARNINGS IN FOUR DAYS — 2026-08-08. Berkshire Hathaway reports second-quarter 2026 results on Saturday 2026-08-08, four days after this dive. Vendor consensus is operating EPS $5.04 on revenue $96.523B. The last two prints split: 2026-05-02 delivered $5.25 against $5.08 (a beat); 2026-03-02 delivered $4.73 against $5.51 (a 14% miss). This dive is written into a pre-print window at the top of the 52-week range, and the timing is an explicit input to the verdict, not a footnote. House rule: a name reporting on or before 2026-08-09 cannot carry a Buy — Core. It also means the Q2'26 10-Q — which contains the second-quarter share-repurchase table, the single most decision-relevant number on this name — is not yet available. Everything balance-sheet in this dive is as at 2026-03-31, and we say so wherever it matters. What we would do with the four days: nothing. Wait for the print, and read the repurchase table before the earnings headline.

Reference table

1. What the businesses are

The profile describes Berkshire as a "vast global conglomerate," headquartered in Omaha, listed on the NYSE, with CIK 0001067983. The 10-K's reportable segments are insurance underwriting (GEICO, Berkshire Hathaway Primary Group, Berkshire Hathaway Reinsurance Group), insurance investment income, BNSF, Berkshire Hathaway Energy, manufacturing, service and retailing, McLane, and Pilot.

Segment revenue, from the vendor's seg_prod — and we checked it against the 10-K:

Reported FY2025 revenue was $371.444B, so roughly $3.07B sits outside the disclosed segments.

We tested seg_prod against the 10-K and it holds up — which is unusual and worth stating. The 10-K's manufacturing narrative gives industrial products up $1.5B (+4.1%), building products up $239M (+0.9%) with Clayton Homes at $12.9B and other building products at $13.8B, and consumer products at $14.4B (−3.0%). Summed, that reconstructs to roughly $79.2B against the file's $78.487B — a rounding-level gap. Named sub-businesses check too: Precision Castparts $10.8B (+4.6%, aerospace +7.5%, pre-tax earnings +34.2%), Marmon $12.8B (+4.7%), Lubrizol $6.2B (−3.0%, pre-tax earnings −20.6%). seg_prod is used as reported. Two caveats: the "Insurance Group" line conflates premiums with investment income and is not the filing's own segmentation, and seg_geo is entirely empty — stated as a gap, not estimated around.

Where segment earnings actually moved (10-K and Q1'26 10-Q narrative, after tax):

Three things this table says. First, the insurance engine is decelerating: underwriting profit fell 19% in 2025 and the 10-K warns earnings "may decline in the future from the ongoing impacts of competition within the industry and rising claim cost trends." Second, GEICO's underwriting margin is compressing measurably — the loss ratio was 72.3% in 2025 against 71.8% in 2024, the expense ratio 12.4% against 9.7%, so the combined ratio moved from 81.5% to 84.7%; in Q1'26 it deteriorated further to 87.3% (loss ratio 73.9%, up 4.9 points; expense ratio 13.4%, up 2.6 points), driven by bodily-injury severity up 12-14% and frequency up 5-7%. Third, BNSF is the bright spot, with after-tax earnings up 8.8% in 2025 and 13.4% in Q1'26 on improved operating efficiency, and the operating ratio down 2.3 points to 65.6%.

Two structural items from the filings. The reinsurance group's property premiums declined in Q1'26 on "increased competition and lower rates" — a soft market — but on 2026-03-23 NICO entered a whole-account quota share with Tokio Marine subsidiaries on risks attaching over a ten-year term from 2026-04-01, which the 10-Q says "is expected to generate meaningful premium volumes over its term." And on 2026-01-02 Berkshire completed the acquisition of OxyChem from Occidental for approximately $9.5B, with preliminary assets of $10.8B — the first material deployment of the cash pile in this file, and equal to 2.5% of the cash and Treasury Bill balance.

2. The GAAP earnings illusion — the single most important thing to understand, and it runs the opposite way from what screens suggest

Berkshire's reported GAAP earnings are dominated by unrealised marks on a $305.7B securities portfolio. Management says so itself, in writing, twice.

From the Q1'26 10-Q: "We believe that investment gains and losses, whether realized from dispositions or unrealized from changes in market prices and exchange rates, are generally meaningless in understanding our reported periodic results or evaluating our periodic economic performance." And: "the investment gains and losses recorded in earnings in any given period has little analytical or predictive value."

The quarterly evidence, on a Class B per-share basis:

In four quarters the reported GAAP result moved from $2.13 to $14.28 to $8.90 to $4.68 per share — a 6.7-fold range — while operating earnings sat between $4.47 and $5.25. The business did not change by 6.7 times. The stock market did.

The derivation, done from the filings rather than asserted. The 10-K states: "Net earnings attributable to Berkshire shareholders were $67.0 billion for 2025 and included after-tax investment gains of approximately $30.7 billion and after-tax impairment losses of $8.3 billion on our equity method investments."

• FY2025 GAAP net earnings $67.0B = $31.04 per B share (2,157.3M weighted shares)
• Less after-tax investment gains $30.7B, add back the $8.3B equity-method impairment → $44.6B = $20.68 per B share
• Vendor consensus FY2025 EPS was $21.48 on four analysts — confirming the estimates are struck on an operating basis and that our derivation lands within 4%

And it reproduces exactly at the quarterly level. The Q1'26 10-Q: "Net earnings attributable to Berkshire shareholders were $10.1 billion for the first quarter of 2026 and included after-tax investment losses of approximately $1.2 billion." That gives operating earnings of $11.3B, or $5.24 per B share. The reported operating EPS for that quarter was $5.25. A one-cent difference. The method is verified, not assumed.

What it means for the multiple, stated precisely:

• Trailing GAAP diluted EPS = $33.60. The vendor's priceToEarningsRatioTTM of 15.40x is computed on this. It is inflated by investment marks and is REJECTED as a lens.
• Trailing OPERATING EPS ≈ $21.46 (FY2025 operating $20.68, less Q1'25 $4.47, plus Q1'26 $5.25). The honest trailing multiple is $517.22 ÷ $21.46 = 24.1x.
• Independent confirmation: the vendor's operatingCashFlowPerShareTTM is $21.09 — within 1.7% of the derived operating EPS, while incomeQualityTTM (operating cash flow ÷ net income) is 0.627, exactly what you would expect when 37% of reported net income is non-cash marks. Cash flow tracks operating earnings, not GAAP earnings.
• A 24.1x trailing operating multiple is a completely different investment case from a 15.4x GAAP one, and every screen in the market shows the latter.

The same distortion runs through the annual history. FY2022 reported a full-year net LOSS of $22.759B (−$10.36 per B share) in a year when operating cash flow was a positive $37.224B. FY2023 reported $96.223B ($44.27). FY2024 reported $88.995B ($41.27). FY2025 reported $66.968B ($31.04). Consensus operating EPS across the same window was $17.27, $19.84 and $21.48 — roughly half the GAAP figure in the good years.

Therefore: this dive values Berkshire on book value per share and on operating earnings. GAAP earnings are reported for completeness and used for nothing. The vendor's priceToEarningsRatioTTM (15.40x), priceToEarningsDilutedRatioTTM (15.40x), earningsYieldTTM (6.50%), priceToEarningsGrowthRatioTTM (−1.47) and grahamNumberTTM ($504.79) are all rejected as lenses. Recomputed on operating earnings, the Graham number is $403 against a $517.22 price.

3. Book value per share — and the share-class check the renderer cannot do

First, the share-class verification, because getting this wrong invalidates every number on the page. The 10-K states that each Class B share possesses economic rights of one-fifteen-hundredth (1/1,500) and voting rights of one-ten-thousandth (1/10,000) of a Class A share. The data contract flags share-class errors as a failure mode the renderer cannot detect (FWONK, RPRX). We tested it four ways:

1. Market cap ÷ price = $1,115,567,971,055 ÷ $517.22 = 2,156.9M shares against the Q1'26 weighted-average diluted count of 2,157.186M — a 0.015% difference. Reconciles.

2. Q1'26 net earnings ÷ share count = $10.106B ÷ 2,157.186M = $4.68, exactly the reported epsDiluted. B-equivalent.

3. Shareholders' equity ÷ share count = $727.2B ÷ 2,157.186M = $337.11, against the vendor's shareholdersEquityPerShareTTM of $337.097. Reconciles.

4. The 10-Q's own footnote: "Net earnings per average equivalent Class B share outstanding is equal to one-fifteen-hundredth of the equivalent Class A amount."

Verdict: the payload's per-share metrics ARE on a B-equivalent basis and the market capitalisation reconciles to price × shares. There is no 1,500x error in this file. One minor defect does exist and is flagged in Section 8: bookValuePerShareTTM of $338.15 includes the $2.3B of noncontrolling interests, overstating the Berkshire-shareholder figure by $1.05 (0.31%). We use $337.11.

Book value per share, derived from reported equity and share counts:

The 10-K corroborates the FY2025 figure directly: "Berkshire's shareholders' equity at December 31, 2025 was $717.4 billion, an increase of $68.1 billion since December 31, 2024" — a 10.5% increase. But this series is itself contaminated by equity-market marks, which is why the knowledge base's estimate of intrinsic value growth in 2025 is ~9.3%, not 10.4%, and why we do not treat 16.6% (FY2024) as a repeatable rate.

At $517.22 against $337.11, the price-to-book multiple is 1.534x — which the vendor's priceToBookRatioTTM of 1.5343 independently confirms. Tangible book value per share is $283.70, giving a price-to-tangible-book of 1.823x; the $53 gap is $83.074B of goodwill and $34.842B of intangibles from decades of acquisitions.

The multiple has expanded, and that matters. At the 52-week intraday low of $460.38 against FY2025 book value per share of $332.55, the stock traded at 1.384x. Today it is at 1.534x — a 10.8% re-rating while book grew 1.4% in the intervening quarter. The recent move is predominantly multiple, not compounding, and that is the opposite of the setup we want.

On the residual-income frame, honestly: it does not work here and we will not pretend otherwise. The MKL dive used (ROE − g) ÷ (cost of equity − g) because Markel's inputs were stable. Berkshire's are not. Trailing GAAP return on equity is 10.31% but is inflated by investment gains; operating return on equity is $21.46 ÷ $337.11 = 6.37%, which understates economic return because a third of the equity sits in Treasury Bills and the securities portfolio's returns land mostly in unrealised marks that never touch operating earnings. Worse, with a beta of 0.615 the CAPM cost of equity falls close to the assumed growth rate, and the formula explodes: at a 9.5% cost of equity and 5% growth it yields 1.18x; at 8.0% it yields 1.77x; at 7.1% it yields 2.55x. A frame whose output swings from below spot to 66% above spot on a 2.4-point input change is not evidence, and no conclusion in this dive rests on it.

4. Balance sheet — the cash pile, and why enterprise value is the wrong lens even after we fix it

From the 10-Q, at 2026-03-31 — the most recent balance sheet in existence for this company, because the Q2'26 10-Q arrives with the 2026-08-08 print:

• Cash, cash equivalents and U.S. Treasury Bills held by insurance and other businesses (net of payables for unsettled purchases): $373.5B — equal to $173.14 per B share, or 33.5% of the share price
• Equity and fixed maturity securities, excluding equity method investments: $305.7B — $141.71 per share
• Float: $176.9B, up ~$500M from year-end — $82.00 per share, at a negative average cost in Q1'26
• Shareholders' equity: $727.2B, up $9.8B in the quarter
• Cash equivalents include $19.7B of Treasury Bills with maturities of three months or less; the balance includes $17.2B of unsettled Treasury Bill purchases also carried in liabilities and paid shortly after quarter-end

From the FY2025 balance sheet (2025-12-31), which is the last full statement the vendor supplies:

• Cash $51.877B · short-term investments $321.434B · total cash and short-term investments $373.311B
• Long-term investments $336.405B · total investments $657.839B
• Total debt $138.954B (short-term $13.267B, long-term $120.834B, finance leases $4.853B)
• Total assets $1,222.176B · total liabilities $502.473B · equity $717.419B (noncontrolling $2.284B)
• Interest coverage 14.43x · debt-to-equity 0.201 · financial leverage 1.72x · effective tax rate 18.81%

The debt build from the 10-K narrative, which the vendor does not give you: Berkshire parent $22.7B; Berkshire Hathaway Finance Corporation $18.3B; BNSF $24.1B; BHE $59.3B — and, critically, "Berkshire does not guarantee the repayment of any borrowings of BNSF, BHE or their subsidiaries." $83.4B of the $138.954B of consolidated debt — 60% of it — is non-recourse to the parent.

The vendor's enterprise value is wrong by $322 billion, and it is the textbook version of the defect. km_ttm.enterpriseValueTTM reports $1,203.561B, built on a reported netDebt of $87.077B — which is total debt ($138.954B) less cash and cash equivalents ($51.877B), ignoring the entire $321.434B of short-term investments, which for this company are U.S. Treasury Bills.

> Corrected net position: $373.311B − $138.954B = +$234.357B NET CASH.

> Corrected enterprise value = $1,115.568B + $138.954B − $373.311B = $881.211B.

> The vendor overstates enterprise value by $322.35B, or 36.6% relative to the corrected figure.

But the more important point is that enterprise value is the wrong lens for Berkshire entirely, and we reject it on conceptual grounds regardless of the arithmetic. Three reasons, each independently sufficient:

1. The portfolio backs policyholder liabilities. Consolidated claim liabilities were approximately $152B at 2025-12-31 and float was $176.9B at 2026-03-31. You cannot net $373.5B of Treasury Bills against the equity value as though it were surplus cash when $329B of obligations sit against it.

2. 60% of the debt is non-recourse. Adding BNSF's $24.1B and BHE's $59.3B to the parent's equity value produces a number that corresponds to no claim any Berkshire shareholder actually bears.

3. EBITDA is meaningless for an entity whose earnings are half insurance underwriting and investment income. The vendor's ebitdaMarginTTM of 20.09% is an artefact.

Accordingly this dive REJECTS evToSalesTTM (3.206x), evToEBITDATTM (15.956x), enterpriseValueMultipleTTM (15.956x), netDebtToEBITDATTM (1.167x), evToOperatingCashFlowTTM (26.45x) and evToFreeCashFlowTTM (50.42x) — on both correctness and conceptual grounds. The corrected figure is reported only to document the defect. currentRatioTTM (4.274) and quickRatioTTM (4.058) are likewise rejected: Berkshire does not present a classified balance sheet in any economically meaningful sense, and the vendor's $70.536B of "total current liabilities" against $476.289B of "total current assets" excludes the $431.937B of non-current liabilities that are mostly insurance obligations.

Is the cash a fortress or a drag? Both, and the file lets us say which is which. It is a fortress in that management holds a self-imposed $30B floor below which repurchases stop, and the 10-K's language is "Financial strength and redundant liquidity will always be of paramount importance at Berkshire." It is a drag in that $373.5B earning Treasury-Bill yields is the reason insurance investment income fell 8.5% in 2025 and 7.4% in Q1'26 as rates declined. The knowledge base reads it the second way: "record $325B cash (~25% of assets, highest since 2005) — an implicit read that opportunities are scarce and valuations rich" (2024-12-19), and "its swelling cash pile shrinks its investable universe and limits future rates" (2026-01-08).

5. Cash flow — and why free cash flow is not the metric here

The 10-K confirms: "In 2025, our diverse group of businesses generated net operating cash flows of $46.0 billion. Our consolidated capital expenditures... were $20.9 billion in 2025, which included capital expenditures of $14.4 billion by BNSF and BHE. BHE and BNSF forecast capital expenditures in 2026 of approximately $15 billion."

Operating cash flow has averaged $40.7B across four years including FY2022, when GAAP net income was NEGATIVE $22.8B. That is the strongest available evidence that the underlying business is far steadier than reported earnings suggest, and it is a direct corroboration of Section 2.

But two-thirds of the free cash flow is not free. For an insurance holding company, operating cash flow includes net float accumulation — premiums collected now against claims payable later. The 10-K forecasts "claim payments in 2026 are currently expected to exceed $30 billion with respect to claims occurring prior to 2026." The vendor's freeCashFlowYieldTTM of 2.14% and priceToFreeCashFlowRatioTTM of 46.73x are therefore both technically computed and economically uninformative, and neither appears in the valuation.

There is no dividend — lastDividend 0, dividendYieldTTM 0, dividendPayoutRatioTTM 0, and $0 of common dividends paid in every year of the cash-flow statement. The entire shareholder distribution is the buyback, and in FY2025 that was zero.

6. Valuation — priced in or room?

At $517.22 (market cap $1.1156T, 2,157.2M B-equivalent shares, book value per share $337.11 at 2026-03-31):

Estimate coverage is thin for a $1.1 trillion company and must be flagged. Three analysts on FY2026 and FY2027 EPS, two on revenue, and one on FY2028. The FY2028 row shows revenue of $367.146B — 8.3% BELOW FY2027 — which is a single-analyst artefact, not a forecast of decline. No conclusion here rests on the FY2028 row; FY2027 is the primary forward anchor. The est EBITDA and EBIT rows are separately corrupt: FY2026E EBITDA of $89.864B implies a 24% step-up from FY2025's actual $72.356B while the same row's net income of $45.004B is 33% BELOW FY2025's actual $66.968B. Both cannot be true. The EBITDA/EBIT block is rejected and forward valuation runs on epsAvg, which is internally consistent (FY2026E EPS $20.861 × 2,157.2M shares = $45.0B, matching netIncomeAvg exactly).

Peer context is unusable and we say so. The vendor peer set is AIG ($42.1B), Bank of America ($446.4B), Brookfield Wealth Solutions ($15.1B), JPMorgan ($958.0B), Old Republic ($10.6B), Principal Financial ($24.7B), Sun Life ($45.6B) and Visa ($690.0B). Two are money-centre banks, one is a payments network, and none is a diversified conglomerate with a railroad and a utility. No peer multiples are supplied and no peer comparison is drawn.

The vendor's own quality model rates Berkshire B+ with an overall score of 3, and notably scores priceToEarnings 2 and priceToBook 2 out of 5 — the two lowest components. And that P/E score is computed on the flattering 15.40x GAAP figure; on the honest 24.1x it would be worse.

6a. What today's price assumes (the inversion)

At $517.22 — 1.534x book and 24.0x FY2027 consensus operating earnings — the price embeds roughly the following falsifiable claims:

• Book value per share compounds at ≥9-10% a year and the market keeps paying ~1.53x. (Our number, from the equity series in Section 3: $301.11 FY2024 → $332.55 FY2025 → $337.11 at 2026-03-31.) This is doing almost all the work. At the 52-week low the market paid 1.384x on the same business; a return to that multiple is a −10% price outcome even with book flat.
• Operating earnings grow only about 3% a year, and that is acceptable at 24x. (Consensus-derived: $20.86 FY2026E → $21.53 FY2027E = +3.2%, on three analysts.) A 24x multiple on 3% earnings growth requires the multiple to be justified by balance-sheet quality rather than by growth — which is a defensible position, but it is a position, not a fact.
• The $373.5B cash and Treasury Bill pile — $173.14 per share, 33.5% of the price — gets deployed at returns above the bill yield it currently earns. (Filing-derived, 10-Q at 2026-03-31.) The market is paying for optionality it has not been shown. The only deployment in the file is OxyChem at $9.5B, or 2.5% of the pile. Meanwhile the pile is a measurable drag: insurance investment income fell 8.5% in 2025 and 7.4% in Q1'26 as rates declined.
• No succession de-rating. (Our reading.) Greg Abel became chief executive on 2026-01-01 with responsibility for major capital allocation. The market is currently paying a HIGHER multiple of book than it paid before the transition completed — the opposite of a discount.
• The five-name equity concentration does not suffer a drawdown. (Filing-derived: 61% of equity securities fair value in American Express, Apple, Bank of America, Coca-Cola and Chevron at 2026-03-31, down from 65% at year-end.) This is the most fragile assumption in the price. A 20% decline in those five would take roughly $37B pre-tax — about 4% of book value — straight out of the metric the entire valuation rests on, and it would do it in a single quarter's mark.

6b. The return bridge (why the multiple moves)

Expected return over 12-24 months decomposes as: book-value-per-share growth (+9-10%) + multiple drift (we assume roughly FLAT) + shareholder yield (~0%).

Our base case assumes essentially NO multiple expansion — and unusually for this format, that is the argument against the name rather than for it. Our $540 base implies 1.602x book, a 4.4% re-rating from 1.534x. But the shareholder yield leg is genuinely zero: there is no dividend, and Berkshire repurchased nothing in the whole of 2025 and only "relatively minor amounts" in Q1'26. So the return is book compounding, plus or minus the multiple, and nothing else.

Run it forward honestly. If book value per share compounds 10% over the next year to roughly $371 and the multiple simply holds at 1.534x, the price is $569 — a 10.0% return, which is precisely what the knowledge base predicts ("Expect ~10% normalized forward returns — similar to S&P 500 but less volatile"). If instead the multiple reverts to the 1.384x the market paid at the 52-week low, the same $371 of book gives $513 — a 0.8% loss. The entire twelve-month outcome sits between −1% and +10% and turns on a multiple that has already expanded 11% in the last few months.

Contrast with MKL, deliberately. Markel's base case worked with no re-rating at all because it started at 1.111x book against a residual-income-justified 1.64x. Berkshire's starts at 1.534x with no reliable justified multiple to anchor against. That is the difference between the two names and it is the whole verdict: same model, same accounting distortion, opposite starting price.

The bull case at $605 (1.795x book) requires either a very large acquisition that visibly converts the cash pile into earning power, or a sustained buyback at scale. Neither has happened, and that leg is fragile.

6c. Variant perception (where we differ, what would surprise)

• We differ most sharply from the screens, and in an unusual direction — we are LESS bullish than the data appears. Every stock screener in the market shows Berkshire at roughly 15.4x trailing earnings, which is cheap for a business of this quality. The honest figure is 24.1x. The gap is not subtle: $12.14 per share, or 36% of the reported EPS, is investment marks. We have verified this three ways (the 10-K's own $30.7B disclosure, the exact quarterly reproduction of Q1'26's $5.25, and operating cash flow per share of $21.09 against derived operating EPS of $21.46). Watchable number: the operating EPS line in the 2026-08-08 release against consensus $5.04 — and note it will be reported four dollars away from whatever the GAAP headline says.
• We differ from the knowledge base on timing, not direction. The lane's most specific and highest-conviction valuation claim (2026-04-25, conviction 80) puts B-share intrinsic value at $570-585 — but it was struck against a price of ~$480, describing the stock as "about $0.85 on the dollar." At $517.22 the discount has compressed from roughly 19% to 11-13%. Two other valuation claims in the lane are now AT or BELOW spot: the Feb-2026 sum-of-parts at $550 (conviction 65) is 6.3% above, and the 2025-05-01 estimate of ~$520 (conviction 45) is 0.5% above. The panel is bullish and the price has run through most of what the panel said it was worth. Watchable number: whether any refreshed intrinsic-value estimate appears above $585.
• We differ from the street's level but not its direction. Consensus target $465.50 is 10.0% BELOW spot and targetHigh of $481 is also below spot — a configuration we see rarely, and one that is not a single-source artefact (high $481, low $450, median $465.50 show genuine dispersion, and the 4 buy / 6 hold / 0 sell distribution is separately sourced). We think $465.50 is too pessimistic — it implies 1.38x book, exactly the 52-week-low multiple — but the direction of the street's disagreement with the market is the same as ours.
• Positive surprise that would force a re-rate: the Q2'26 10-Q on 2026-08-08 disclosing a material share repurchase above $500. Berkshire buys only below management's conservatively determined intrinsic value. Zero in 2025, restart on 2026-03-04 near $480, "relatively minor" in Q1'26. A large Q2 number would be Greg Abel telling us, with money, that $500+ is below intrinsic value — the single most informative data point available on this name. A very large acquisition deploying the cash would do the same.
• Negative surprise that would break the thesis: the same 10-Q disclosing ZERO repurchases in Q2'26, i.e. management stopped once the stock crossed roughly $500 — which would put the intrinsic-value threshold below today's price, from the best-informed source there is. Secondarily: GEICO's combined ratio deteriorating past the 87.3% printed in Q1'26; a catastrophe quarter (Q1'26 had none, Q1'25 had $860M after tax); or a drawdown in the five concentrated equity holdings.

Synthos fair values

All three anchors are multiples of the book value per share of $337.11 implied by the $727.2B of shareholders' equity in the 2026-03-31 10-Q, cross-checked against FY2027 consensus operating EPS of $21.53.

• Bear ~$440 — 1.305x book. Cross-check: 20.4x FY27E operating EPS, and roughly $267 of book above the $173.14 per share of cash and Treasury Bills. The scenario: a market drawdown hits the concentrated equity portfolio and takes book down 5-7%; GEICO's combined ratio keeps deteriorating past 87.3%; a catastrophe quarter arrives; and the multiple compresses below the 1.384x seen at the 52-week low as the market re-prices Abel's first capital-allocation cycle. −14.9%, and below the 52-week low of $460.38.
• Base ~$540 — 1.602x book. Cross-check: 25.1x FY27E operating EPS; sits between the knowledge base's Feb-2026 sum-of-parts of $550 and the 2025-05-01 estimate of ~$520, and below the April-2026 range of $570-585. The scenario: the business does what it has done — book compounds ~10%, underwriting stabilises, BNSF keeps improving, and the multiple holds. +4.4%.
• Bull ~$605 — 1.795x book. Cross-check: 28.1x FY27E operating EPS. The scenario: Abel deploys a material share of the $373.5B pile into a large acquisition at attractive terms, or the buyback runs at scale above $500 and confirms management's intrinsic-value view, and the market pays a defensive premium for a beta-0.615 asset with a 4.2% maximum drawdown. +17.0%.

Base is 4.4% above spot; asymmetry is roughly 1.14:1 (14.9% down, 17.0% up) — essentially symmetric, which is the arithmetic definition of no edge. Combined with a print in four days, that is a Watch, not a Buy.

7. Knowledge base — 62 claims, one source at 61%, and a lane the price has run through

Tracked knowledge-base claims naming Berkshire: 62. Raw hits: 62. Discarded: 0. Quarantined (quarantine_misattributed): 0.

The search covered the aliases BRK-B, BRK.B, BRK, Berkshire and Berkshire Hathaway across all 51,928 records, exact-matching entity names case-insensitively. Stance split: 42 bullish, 18 neutral, 2 bearish.

Breadth is the weakness and we state it before the quotes. Twelve distinct sources — but we_study_billionaires accounts for 38 of 62 claims, 61.3% of the lane. The remainder: invest_like_the_best 8, business_breakdowns 4, compound_and_friends 3, lyn_alden 2, and one each from Josh Brown, anthony_pompliano, cognitive_revolution, luke_gromen, arthur_hayes, odd_lots and Clay Finck. A single independent source supplying six claims in ten is a concentration problem even when that source is credible and non-management. We do not treat 62 claims as 62 independent observations.

Weighting applied. Zero management voices — no _mgmt slug appears, and no claim originates with Berkshire's own executives — so the independent-voice count is the full 62, which is a genuine strength relative to names whose lanes are dominated by their own chief executive. One claim (2025-11-18, anthony_pompliano, conviction 45, "Holds Berkshire Hathaway (~5%) as a core long-term position") is in the host's own voice and is promoter-class: LOW weight, and it carries no analytical content. One record has a BLANK date (Clay Finck, neutral, conviction 40) — flagged and excluded from any recency reading.

The two "bearish" claims are not bearish on the stock, and this matters. They are: "Buffett's 'never lose money' rule is the wrong message; taking risk and tolerating losers is how you win" (2025-09-18, conviction 85) and "Buffett's Dexter Shoe purchase failed by overestimating the moat" (2023-09-30, conviction 80). Both are philosophical or historical critiques, not investment cases against the equity. The honest reading is that this lane contains ZERO genuine counter-argument — which is a reason to trust it less, not more.

The valuation lane — the claims that actually name a price:

> 2026-04-25 · we_study_billionaires · bullish · conviction 80 · entities: BRK

> "Berkshire trades at a discount to fair value — about $0.85 on the dollar, B-share intrinsic ~570-585 vs price ~480; he's been buying B shares (460 in August, ~480 recently)."

> 2026-02-28 · we_study_billionaires · bullish · conviction 65 · entities: Berkshire Hathaway, BRK

> "Sum-of-parts (operating earnings at 17x plus equities and net cash) values BRK A-shares near $550 vs ~$497 trading — modestly undervalued."

> 2025-05-01 · we_study_billionaires · neutral · conviction 45 · entities: BRK

> "Berkshire's intrinsic value grew ~10% over the past 2 years (Bloomstrand est. ~$520/share); if that continues, the stock should compound at about that rate long term, though price now exceeds that IV estimate."

> 2026-04-25 · we_study_billionaires · bullish · conviction 68 · entities: BRK

> "Berkshire's intrinsic value grew ~9.3% in 2025; its key operating parts can grow earning power 10-12% annually on average over the next 10-15 years."

A share-class error INSIDE the knowledge base, flagged. The 2026-02-28 claim says "values BRK A-shares near $550 vs ~$497 trading." Class A shares trade at 1,500 times the B price; $550 and $497 are B-share figures. The claim's label is wrong; its numbers are right. We use it as a B-share estimate and note the defect — the same error class the data contract warns about on the vendor side, appearing on the knowledge-base side.

Read together, the valuation lane says the stock was worth $520-585 per B share at various points across fifteen months, against a price that has now reached $517.22. The freshest and highest-conviction estimate ($570-585, conviction 80) was struck when the price was ~$480. The panel is bullish. The price has consumed most of what the panel said it was worth.

The structural lane — why the business is durable:

> 2026-01-10 · bullish · conviction 85 — "Insurance float is a powerful engine: run underwriting, invest the float for extra returns, keep cash for claims — the Buffett model Fairfax copied."

> 2025-10-30 · bullish · conviction 82 — "BNSF is an irreplaceable moat — impossible to rebuild the network today, cheapest way to move goods, and positioned for the Asian-century Pacific trade shift; Buffett recovered his capital and now earns ~12-13%."

> 2026-02-28 · bullish · conviction 75 — "Long Berkshire; its moat is a durable culture of prudent, decentralized, ethical capital allocation that persists through the Buffett-to-Abel transition."

> 2026-05-02 · bullish · conviction 62 — "Under Greg Abel, Berkshire will run tighter operations — he'll act on underperforming managers rather than leave them alone, a middle ground between Buffett's hands-off delegation."

The lane's own limiting case — and it is the more useful half:

> 2026-02-28 · neutral · conviction 70 — "Expect ~10% normalized forward returns — similar to S&P 500 but less volatile; growth is slowing and its trillion-dollar size caps upside."

> 2026-02-28 · neutral · conviction 55 — "Buffett's alpha largely disappeared after Reg FD removed information asymmetry; at scale even Berkshire returns fell in line with the S&P 500."

> 2026-03-25 · invest_like_the_best · bullish · conviction 75 — "Use Berkshire B shares as your default index substitute; ~10%/yr is reasonable, doubling every 7 years via rule of 72."

> 2024-12-19 · neutral · conviction 55 — "Berkshire holds record $325B cash (~25% of assets, highest since 2005) — an implicit read that opportunities are scarce and valuations rich."

> 2026-01-08 · bullish · conviction 60 — "Berkshire is a core holding expected to compound intrinsic value ~10%/yr, though its swelling cash pile shrinks its investable universe and limits future rates."

The most recent claim in the lane is the most momentum-flavoured and the least analytical: "Been pounding the table on insurance (Travelers, All State) making record highs; Berkshire is firing on all cylinders — Coca-Cola, Amex, transports, utilities — it has everything investors want now" (2026-07-28, Josh Brown, bullish, conviction 70). "Everything investors want now" is a description of positioning, not of value, and it arrives with the stock at its 52-week high. We read it as sentiment confirmation rather than as a valuation input.

Why Moderate rather than Moderate-High. Sixty-two claims, twelve sources, zero quarantined and zero management voices is a real lane. But 61% comes from one source; both bearish tags are philosophical rather than investment cases; a large share of the claims are Buffett-the-investor principles that happen to tag the entity rather than views on the equity at a price; the panel's own forward-return estimate is ~10%, i.e. market-like; and the specific valuation claims have been overtaken by the price. A lane that says "worth $520-585" against a $517 price is confirming, not differentiating.

8. Data integrity — what we rejected from the vendor file and why

Nine material findings. Listed rather than silently corrected. Where a filing and the vendor disagree, the filing wins and we say so.

1. km_ttm.enterpriseValueTTM ($1,203.561B) ignores $321.434B of short-term investments — a $322B error — and enterprise value is REJECTED as a lens regardless. The reported netDebt of $87.077B nets total debt ($138.954B) against cash and cash equivalents only ($51.877B), omitting the entire short-term investment balance, which for this company is U.S. Treasury Bills. Corrected: net CASH of $234.357B; corrected enterprise value $881.211B. The vendor overstates by 36.6%. But EV is conceptually invalid here: $152B of claim liabilities and $176.9B of float sit against the portfolio, and "Berkshire does not guarantee the repayment of any borrowings of BNSF, BHE or their subsidiaries" (10-K) — so $83.4B of the consolidated debt, 60% of it, is non-recourse to the parent. evToSalesTTM (3.206x), evToEBITDATTM (15.956x), enterpriseValueMultipleTTM (15.956x), netDebtToEBITDATTM (1.167x), evToOperatingCashFlowTTM (26.45x) and evToFreeCashFlowTTM (50.42x) are all rejected.

2. Trailing GAAP earnings multiples are economically misleading — REJECTED as a lens. priceToEarningsRatioTTM and priceToEarningsDilutedRatioTTM of 15.40x are arithmetically correct on trailing GAAP diluted EPS of $33.60, but the 10-K states FY2025's $67.0B of net earnings "included after-tax investment gains of approximately $30.7 billion and after-tax impairment losses of $8.3 billion." This dive uses derived trailing operating EPS of $21.46, giving 24.1x, corroborated by operatingCashFlowPerShareTTM of $21.09 and incomeQualityTTM of 0.627. Also rejected on the same basis: earningsYieldTTM (6.50%), priceToEarningsGrowthRatioTTM (−1.47, negative because GAAP EPS fell), forwardPriceToEarningsGrowthRatioTTM (4.81), grahamNumberTTM ($504.79 — recomputed on operating earnings it is $403), and returnOnEquityTTM (10.31% — the operating figure is 6.37%).

3. earn_cal mixes operating and GAAP EPS across rows — INTERNALLY INCONSISTENT, and only three of five rows are usable. Cross-checking each epsActual against the corresponding quarter's GAAP epsDiluted in inc_q: 2025-05-02 reports $4.47 against GAAP $2.13 (operating); 2025-08-01 reports $5.73 against GAAP $5.73 (GAAP — contaminated); 2025-11-03 reports $14.27 against GAAP $14.28 (GAAP — contaminated, and note it is scored as a "beat" against a $5.39 estimate, which is meaningless); 2026-03-02 reports $4.73 against GAAP $8.90 (operating); 2026-05-02 reports $5.25 against GAAP $4.68 (operating). Two of five rows are the wrong basis. We therefore derive trailing operating EPS from the 10-K's own disclosure rather than from earn_cal, and we discard the 2025-08-01 and 2025-11-03 rows entirely.

4. earn_cal revenue actuals contradict inc_q — the income statement wins. The 2026-03-02 row reports Q4'25 revenue of $100.667B; inc_q for 2025-12-31 reports $94.232B — a $6.4B (6.8%) difference. The 2025-08-01 row reports $98.879B against inc_q's Q2'25 figure of $92.515B — $6.4B again. We use inc_q.

5. est EBITDA and EBIT rows are inconsistent with the same rows' net income — REJECTED, and EBITDA rejected conceptually throughout. FY2026E ebitdaAvg of $89.864B is a 24% step-up from FY2025's actual $72.356B, and FY2026E ebitAvg of $77.058B is a 31% step-up from FY2025's actual operating income of $58.880B — while the same row's netIncomeAvg of $45.004B is 33% BELOW FY2025's actual net income of $66.968B. Both cannot be true. Per the data contract, forward valuation runs on epsAvg, which is internally consistent (FY2026E $20.861 × 2,157.2M shares = $45.0B, matching netIncomeAvg). EBITDA is in any case meaningless for an entity half of whose earnings are underwriting and investment income.

6. The FY2028 estimate row is a single-analyst artefact — REJECTED. numAnalystsEps and numAnalystsRevenue are both 1, and the row shows revenue of $367.146B, 8.3% BELOW FY2027E's $400.258B. No conclusion rests on it; FY2027 (3 analysts on EPS, 2 on revenue) is the forward anchor.

7. bookValuePerShareTTM ($338.15) includes noncontrolling interests — a 0.31% overstatement, corrected. $338.15 × 2,157.186M = $729.5B, which is total equity including the $2.284B of minority interest. The Q1'26 10-Q states Berkshire shareholders' equity of $727.2B, giving $337.11 — which the vendor's own shareholdersEquityPerShareTTM of $337.097 confirms. We use $337.11. Small, but it is exactly the class of silent per-share error the data contract flags, and it moves the price-to-book from 1.529x to 1.534x.

8. Balance-sheet sub-categories are not comparable across FY2023 and FY2024/FY2025 — sub-lines REJECTED, totals used. shortTermInvestments moves $129.619B (FY2023) → $286.472B (FY2024) → $321.434B (FY2025) while longTermInvestments moves $407.621B → $318.934B → $336.405B — an $89B swing between two lines in one year that reflects a vendor re-mapping alongside genuine Treasury Bill accumulation. taxAssets reports $93.009B in FY2023 and $0 in FY2024 and FY2025, with otherNonCurrentAssets simultaneously reporting −$42.029B in FY2023 — a negative asset balance. Only totals are used, and currentRatioTTM (4.274) and quickRatioTTM (4.058) are rejected as meaningless for an entity that does not present a classified balance sheet.

9. Reported revenue is identical to five significant figures across FY2024 and FY2025 while the segments moved — FLAGGED, not corrected. inc_a reports $371.433B (FY2024) and $371.444B (FY2025) — an $11 million difference on $371 billion, or 0.003%. Meanwhile seg_prod shows the disclosed segments falling from $370.892B to $368.372B (−$2.52B), driven by Pilot Travel Centers −$4.693B. We cannot reconcile a flat consolidated line with a $2.5B segment decline from this file, and we flag it rather than assert either figure is wrong. The Q1'26 comparison is cleaner and is used where growth is discussed: $93.675B against $89.725B, +4.4%.

10. seg_geo is entirely EMPTY, and insider data is stale by more than two years. The most recent insider transaction in the file is 2024-01-12; the rest are 2023. For a name where the 8-K of 2026-03-05 and the Q1'26 10-Q both carry live capital-allocation signals, the insider block contributes nothing. Both are stated as gaps.

Filing-versus-vendor agreements worth recording, since the standing rule cuts both ways: the 10-K's "net operating cash flows of $46.0 billion" matches cf_a at $45.969B; "capital expenditures... $20.9 billion" matches $20.927B; "shareholders' equity at December 31, 2025 was $717.4 billion" matches bal_a at $717.419B; "There were no share repurchases in 2025" matches commonStockRepurchased at $0. seg_prod reconciles to the 10-K's manufacturing narrative within rounding. Where this file is right, it is right — which is what makes the ten defects above worth naming individually.

Non-equity tripwire: NOT triggered. BRK-B is common equity — $517.22 price, 4.25M daily volume against a 4.08M average, a 52-week band 12.6% wide, no dividend, and a market capitalisation of $1.1156T. The beta of 0.615 is genuinely low but is a property of the business, not a par-instrument artefact.

9. Technicals — at the high, overbought, into a print

• Price $517.22, up 0.80% today. This is the 52-week CLOSING high (tech.hi52 $517.22, pct_from_hi 0.0%) and 0.1% below the 52-week intraday high of $517.75. It is +12.7% above the 52-week closing low of $459.11.
• ABOVE both moving averages and stretched: +5.0% over a 50-DMA of $492.77 and +5.4% over a 200-DMA of $490.58, both rising. RSI 83.2 — deeply overbought; readings above 80 are rare on a beta-0.615 mega-cap. MACD +5.64.
• Maximum drawdown from peak over the trailing year: −4.2%. That is the single most impressive technical statistic here and it cuts both ways: it is genuine evidence of the defensive character, and it means there has been no pullback to buy in twelve months.
• Relative performance is split, and the split is unfavourable for entry: 3-month +10.4% vs SPY +7.4% and QQQ +7.6% (outperforming); 6-month +6.1% vs SPY +10.9% and QQQ +15.6% (−4.8 to −9.5 points); 12-month +9.4% vs SPY +24.1% and QQQ +30.7% (−14.7 to −21.3 points). The name lagged badly for a year and has outperformed for a quarter — the crowd has just arrived, which is corroborated by the most recent knowledge-base claim describing it as having "everything investors want now."
• Sentiment corroboration runs the other way from the chart: 4 buys, 6 holds, 0 sells, with a consensus target of $465.50 — 10.0% BELOW spot and a targetHigh of $481 also below spot. Dispersion is genuine ($450 low, $481 high), so this is a real consensus, not the single-source artefact seen elsewhere in this batch.
• Practical read: the chart is strong, the valuation is full, the analysts are underwater, and the print is in four days. There is no technical entry here — the stock would have to give back 5% to reach its 50-DMA and 11% to reach the 1.384x book multiple it traded at in the last twelve months. The 200-DMA of $490.58 corresponds to roughly 1.455x book and is the first level at which the arithmetic becomes interesting.

10. Insiders — no usable signal, and the corporate buyback is the substitute

The insider block in this file is unusable and we will not pretend otherwise. The eight transactions supplied run from 2023-10-27 to 2024-01-12 — more than two and a half years stale — and every one is either a G-Gift disposition (Ajit Jain, 28 shares; Charlotte Guyman, several small charitable transfers) or a C-Conversion of Class A into Class B at a price of $0. There is not one open-market purchase or sale, and nothing at all since January 2024. No inference is drawn.

The substitute signal is far better, and it is the most valuable disclosure on this name. Berkshire's repurchase policy is an explicit intrinsic-value statement from the best-informed buyer in the market. From the 10-K: "Berkshire's common stock repurchase program permits Berkshire to repurchase its Class A and Class B shares at prices below Berkshire's intrinsic value, as conservatively determined by Berkshire's Chief Executive Officer after consultation with the Chairman of the Board... We will not repurchase our stock if it reduces our consolidated cash, cash equivalents and U.S. Treasury Bills holdings to below $30 billion."

The record:

The reading, stated carefully. Berkshire declined to buy a single share throughout 2025, a year in which the shares traded well below today's price. It began buying on 2026-03-04, at a price the knowledge base independently places near $480 (2026-04-25: "he's been buying B shares (460 in August, ~480 recently)" — note that claim describes the speaker's own purchases, but it fixes the market price of the period). Management's conservatively determined intrinsic value is therefore somewhere above roughly $480 and we have no evidence about $517. The Q1'26 amount was "relatively minor," which is consistent with either conviction constrained by the 27-day window or with tentative buying near the threshold.

Note also the governance change. The 10-K records that "In 2025, Berkshire's common stock repurchase program was amended to permit Berkshire to repurchase... at any time that Berkshire's Chief Executive Officer, after consultation with the Chairman of the Board, believes that the repurchase price is below Berkshire's intrinsic value... Prior to the amendment, the program permitted Warren Buffett... to repurchase." The buyback decision has passed from Buffett to Abel. The March 2026 restart is Abel's first public intrinsic-value statement, and Q2'26 will be his second.

Management change, for completeness: Greg Abel became chief executive on 2026-01-01 (appointed May 2025); Buffett remains Chairman; Ajit Jain is Vice Chairman of insurance; and per the 8-K of 2026-05-07, Charles C. Chang succeeded Marc D. Hamburg as Chief Financial Officer on 2026-06-01 at an $8,000,000 cash salary. The 10-K's key-personnel risk factor names all of them and states that "Major capital allocation and investment decisions are the responsibility of Mr. Abel." The Q1'26 10-Q reports no material change to any risk factor relative to the 10-K.

Earnings-call transcripts are not available on this plan. Management Q&A is absent from this analysis, and the 8-K exhibits carry the earnings release rather than the call. This is a genuine gap on a name where the annual meeting and the chief executive's commentary carry unusual weight, and it is stated rather than filled with invented quotes.

11. Verdict, kill-criteria and flip conditions

Watch.

This is a very good business at a price that has caught up with it, four days before a print, at the top of its range.

Quality: $373.5B of cash and Treasury Bills at 2026-03-31 (33.5% of the share price), $305.7B of securities, $176.9B of float at negative cost in every year disclosed, $727.2B of equity, interest coverage of 14.4x, 60% of consolidated debt non-recourse to the parent, a beta of 0.615, a maximum trailing-year drawdown of 4.2%, favourable prior-year reserve development everywhere it is disclosed, and BNSF after-tax earnings up 13.4% in Q1'26. Nothing in this dive disputes the quality.

Price: 1.534x book and 24.1x trailing operating earnings — not the 15.40x GAAP figure every screen shows. The multiple has expanded 10.8% from the 1.384x paid at the 52-week low while book grew 1.4%. Revenue rose $11 million on $371 billion between 2024 and 2025. Insurance underwriting after-tax earnings fell 19% in 2025 and GEICO's combined ratio deteriorated 7.5 points year-over-year in Q1'26 to 87.3%. Consensus operating EPS growth from FY2026E to FY2027E is 3.2%. Base fair value $540, +4.4%; bear $440, −14.9%; bull $605, +17.0% — asymmetry of 1.14:1, which is no edge.

Sentiment and timing: at the 52-week closing high, RSI 83.2, +5.4% over the 200-DMA, street consensus 10.0% below spot with targetHigh also below spot, and a print on 2026-08-08.

Conviction: 62 claims, zero quarantined, zero management voices — but 61% from one source, no genuine bear case in the lane, a panel forward-return estimate of ~10%, i.e. market-like, and specific intrinsic-value estimates of $520-585 that the price at $517.22 has largely consumed.

What we would do with the four days: nothing. Wait for the print, and read the Q2'26 10-Q's Item 2 repurchase table before the earnings headline. Berkshire bought nothing in all of 2025 and restarted on 2026-03-04 near $480. Whether Greg Abel kept buying above $500 is the single most informative fact available on this name, it is unknowable today, and it becomes knowable on Saturday. That is a rare, clean, dated resolution of the exact question the verdict turns on — and it is why this is a Watch rather than a Hold.

Pre-registered KILL criteria — what would keep this a Watch or push it to Avoid:

• Zero share repurchases disclosed in the Q2'26 10-Q. Management stopping once the stock crossed roughly $500 would place their conservatively determined intrinsic value below today's price, from the best-informed source available.
• GEICO's combined ratio deteriorating past ~90% from the 87.3% printed in Q1'26 (loss ratio 73.9% + expense ratio 13.4%), on top of the move from 81.5% to 84.7% across FY2024 to FY2025.
• Insurance underwriting after-tax earnings falling below ~$6B annualised, from $9.0B (FY2024) to $7.3B (FY2025) — the trajectory the 10-K itself warns about.
• Book value per share failing to compound at ≥8% over four quarters, which would remove the only leg of the return bridge that is currently working.
• A further re-rating above ~1.75x book (~$590) without a corresponding deployment of the cash pile — paying a bull-case multiple for an option that has not been exercised.
• A material adverse move in the five concentrated equity holdings (61% of the portfolio at 2026-03-31), which flows into book value within one quarter's mark.

Pre-registered FLIP TO BUY (initiate, staged):

• A material Q2'26 repurchase disclosed on 2026-08-08 at prices above $500. This is the primary trigger. It converts the buyback from an unknown into management's own confirmation that today's price is below intrinsic value, and it would justify a staged entry immediately.
• A large acquisition — meaningfully bigger than OxyChem's $9.5B, which was 2.5% of the pile — deploying the cash into operating earning power at an attractive multiple. This is the one thing that would change the exponential score.
• A price below ~$490 (roughly the 200-DMA, ~1.455x book) on market beta rather than company-specific news, which would restore a mid-single-digit-plus margin of safety to the $540 base.
• A price below ~$460 (~1.365x book, the 52-week-low multiple), at which point the base case offers ~17% and the name becomes a straightforward accumulation.
• Operating EPS above ~$5.50 in the Q2 print against consensus $5.04, combined with an improving GEICO combined ratio — evidence that underwriting has stabilised rather than continued to slide.

Where BRK-B would fit in the Synthos Framework Portfolio. The low-beta defensive / ballast sleeve, at a 2-3% weight — but not yet, and not this week. The correct action today is to place it on the watchlist with the 2026-08-08 repurchase disclosure as the named trigger and ~$490 as the named price trigger, whichever arrives first. On overlap within this batch: Berkshire and Markel occupy the same conceptual slot — insurance float compounding into a diversified holding company — and the two dives reach opposite conclusions for the same reason. Markel is 1.111x book on a 12.4% return on equity with zero street buy ratings; Berkshire is 1.534x book on a 10.3% GAAP return on equity with a street target 10% below spot. Same model, same accounting distortion, opposite starting price. If only one is held today, it should be Markel. Logged as a tracked Synthos call (Watch) as of 2026-08-04 at $517.22.

Single biggest risk: equity-portfolio concentration. At 2026-03-31, 61% of the fair value of Berkshire's equity securities sat in five companies — American Express, Apple, Bank of America, Coca-Cola and Chevron — down from 65% at year-end. The valuation rests on book value per share, and book value absorbs those marks in the quarter they occur. A 20% decline in those five would take roughly $37B pre-tax, about 4% of shareholders' equity, out of the denominator of every multiple in this dive, and it would do so with no change whatsoever in the operating businesses. This is also the mechanism that makes the GAAP earnings unusable in the first place — the same $305.7B of securities that distorts the income statement is the concentrated exposure that sits under the balance sheet.

Most fragile assumption in the price: that the market keeps paying 1.53x book for a business compounding book at ~10% with 3% operating-earnings growth and a third of its market capitalisation sitting in Treasury Bills. It paid 1.384x within the last twelve months. Nothing in the FY2025 or Q1'26 filings — flat revenue, underwriting profit down 19%, investment income down 8.5%, a combined ratio deteriorating 7.5 points at GEICO — justifies the multiple being higher today than it was then. If the multiple simply reverts to where it traded at the 52-week low, a full year of book compounding buys you nothing.

Provenance & disclosures

• Traceability: 62 tagged knowledge-base claims name Berkshire (breadth 12, net conviction positive-moderate; 42 bullish, 18 neutral, 2 bearish). Searches covered BRK-B, BRK.B, BRK, Berkshire and Berkshire Hathaway across 51,928 records, exact-matching entity names case-insensitively. 62 raw hits, 0 discarded, 0 quarantine_misattributed. Source distribution: we_study_billionaires 38 (61.3% — a breadth weakness, stated in Section 7), invest_like_the_best 8, business_breakdowns 4, compound_and_friends 3, lyn_alden 2, and one each from Josh Brown, anthony_pompliano, cognitive_revolution, luke_gromen, arthur_hayes, odd_lots, Clay Finck. Weighting applied: ZERO management (_mgmt) voices, so the independent-voice count is the full 62 — a genuine strength relative to lanes dominated by a company's own executives. One promoter-class claim (2025-11-18, anthony_pompliano, conviction 45, in his own voice) carries LOW weight and contributes no analytical content. One record carries a BLANK date (Clay Finck, neutral, conviction 40) and is excluded from any recency reading — flagged. Both claims tagged bearish are philosophical or historical critiques of Buffett rather than investment cases against the equity, so the lane contains no genuine counter-argument; this is treated as a reason for caution, not confidence. One knowledge-base claim (2026-02-28) contains its own share-class error, labelling B-share figures ($550 intrinsic vs ~$497 trading) as A-shares; corrected and used as B-share estimates. All quotes verbatim from stored claim text.
• Data as-of: fundamentals — balance sheet and float 2026-03-31 (Q1'26 10-Q, filed 2026-05-04); income statement 2026-03-31 (Q1'26); annual statements 2025-12-31 (FY2025 10-K, filed 2026-03-02) · estimates 2026-08-04 · prices 2026-08-04 ($517.22 close, +0.80%; 50-DMA $492.77; 200-DMA $490.58; RSI 83.2; volume 4.25M) · knowledge-base claims 2026-08-04. The Q2'26 10-Q DOES NOT YET EXIST — it arrives with the 2026-08-08 print. Every balance-sheet figure in this dive is as at 2026-03-31 and no figure is projected past it. All numbers from the Synthos vendor payload for BRK-B or from the SEC filings on disk (10-K 2026-03-02, 10-Q 2026-05-04, 10-Q 2025-11-03, 8-K 2026-03-05, 8-K 2026-04-16, 8-K 2026-05-07); no figure comes from memory, recall or external retrieval.
• Data rejections (detailed in Section 8): enterpriseValueTTM corrected from $1,203.561B to $881.211B (omits $321.434B of short-term Treasury Bills; net debt of $87.077B is actually net CASH of $234.357B) — and enterprise value rejected as a lens entirely, with all six EV-derived multiples discarded; trailing GAAP price-to-earnings (15.40x) rejected in favour of derived trailing operating earnings (24.1x), along with earningsYield, both PEG variants, grahamNumber and returnOnEquityTTM; earn_cal rows for 2025-08-01 and 2025-11-03 discarded as GAAP-basis rather than operating-basis, and earn_cal revenue actuals rejected in favour of inc_q; est EBITDA/EBIT rows rejected as inconsistent with the same rows' net income, and EBITDA rejected conceptually; the FY2028 estimate row rejected as a single-analyst artefact showing revenue 8.3% below FY2027; bookValuePerShareTTM ($338.15) corrected to $337.11 for the inclusion of noncontrolling interests; FY2023 balance-sheet sub-categories rejected as non-comparable (taxAssets $93.009B → $0, otherNonCurrentAssets −$42.029B), with currentRatio and quickRatio discarded; FY2024-to-FY2025 revenue identical to five significant figures while segments fell $2.52B — flagged, not corrected; seg_geo entirely empty and insider data stale to January 2024 — both stated as gaps.
• Share-class verification (the error class the renderer cannot detect): the 10-K confirms each Class B share carries one-fifteen-hundredth (1/1,500) of the economic rights and one-ten-thousandth (1/10,000) of the voting rights of a Class A share. Market capitalisation ÷ price gives 2,156.9M shares against the Q1'26 weighted-average diluted count of 2,157.186M (0.015% difference); Q1'26 net earnings of $10.106B ÷ 2,157.186M = $4.68, matching reported epsDiluted; $727.2B of shareholders' equity ÷ 2,157.186M = $337.11, matching shareholdersEquityPerShareTTM. The payload's per-share metrics are confirmed to be on a B-equivalent basis; there is no 1,500x error in this file. The only per-share defect found is the noncontrolling-interest inclusion in bookValuePerShareTTM, corrected above.
• Basis note — the most important one in this dive: Berkshire's GAAP earnings are dominated by unrealised marks on a $305.7B securities portfolio, and management says so in the filings ("generally meaningless in understanding our reported periodic results"; "little analytical or predictive value"). FY2025 GAAP net earnings of $67.0B included after-tax investment gains of ~$30.7B and after-tax impairment losses of $8.3B, giving operating earnings of ~$44.6B, or $20.68 per B share against reported GAAP EPS of $31.04. Q1'26 GAAP net earnings of $10.1B included after-tax investment losses of ~$1.2B, giving derived operating EPS of $5.24 against the $5.25 actually reported — a one-cent verification of the method. Trailing operating EPS is therefore ~$21.46, confirmed independently by operatingCashFlowPerShareTTM of $21.09 (within 1.7%) and by incomeQualityTTM of 0.627. This dive values on price-to-book (1.534x) and on operating earnings (24.1x trailing, 24.0x FY27E); GAAP earnings are reported for completeness and used for nothing.
• Cash-flow note: for an insurance holding company, operating cash flow includes net float accumulation and is not distributable — the 10-K forecasts claim payments "expected to exceed $30 billion" in 2026 on pre-2026 occurrences. The vendor's 2.14% free-cash-flow yield and 46.73x price-to-free-cash-flow are both computed correctly and used for nothing. There is no common dividend; the entire shareholder distribution is the buyback, which was $0 in FY2025.
• Estimate coverage: thin for a $1.116 trillion company — 3 analysts on FY2026 and FY2027 EPS, 2 on revenue, and 1 on FY2028. The FY2028 row is rejected. FY2027 is the primary forward anchor. epsAvg is used exclusively; the EBITDA and EBIT blocks are rejected per the data contract's known failure mode.
• Peer note: the vendor peer set (AIG, Bank of America, Brookfield Wealth Solutions, JPMorgan, Old Republic, Principal Financial, Sun Life, Visa) contains two money-centre banks and a payments network and no diversified conglomerate. No peer multiples are supplied and no peer comparison is drawn. The most useful comparison in-house is Markel, covered the same day, and it is drawn explicitly in Section 11.
• Fair-value caveat: the $440 / $540 / $605 anchors are multiples of the book value per share of $337.11 implied by the $727.2B of shareholders' equity in the 2026-03-31 10-Q (1.305x / 1.602x / 1.795x), cross-checked against FY2027 consensus operating EPS of $21.53 (20.4x / 25.1x / 28.1x). Stated arithmetic, not a discounted cash flow. A residual-income frame was attempted and rejected as unusable: with a beta of 0.615 the cost of equity falls close to the assumed growth rate and the output swings from 1.18x to 2.55x on a 2.4-point input change. No conclusion rests on it.
• Timing: Berkshire reports Q2'26 on 2026-08-08, FOUR DAYS after this dive (consensus operating EPS $5.04, revenue $96.523B). Per house rules this caps the verdict below Buy — Core, and the verdict is Watch. The Q2'26 10-Q filed with that release contains the second-quarter share-repurchase table, which is the named trigger for a flip to Buy. The last two prints split: 2026-05-02 $5.25 against $5.08 (beat); 2026-03-02 $4.73 against $5.51 (14% miss). Earnings-call transcripts are not available on this plan; management Q&A is absent from this analysis and no quotes are invented.
• Not investment advice. Independent research, educational and informational only, never personalised. No recommendation to buy, sell or hold any security is made to any person.
• Version: 2026-08-04-full.