Business Overview

Berkshire Hathaway Inc.

NYSE: BRK.A / BRK.B

7 September 2026

Evidence base: Berkshire Hathaway SEC filings — FY2016, FY2020, FY2024 and FY2025 Forms 10-K, the Q2 FY2026 Form 10-Q (filed 10 August 2026), the 2026 proxy statement and Items 5.02/8.01 Forms 8-K — supplemented by independent industry sources (NAIC, Insurance Information Institute, BLS, Swiss Re sigma, Guy Carpenter, AAR, the Surface Transportation Board, EIA, EEI, CPUC) and the filings of named competitors.

This document is an analysis of the business. It is not a valuation and not an investment recommendation.

1. Executive Snapshot

Sources: Berkshire FY2025 Form 10-K; Q2 FY2026 Form 10-Q; NAIC 2025 P&C industry analysis and market share reports; BLS CPI motor vehicle insurance series.

2. What the Company Does

Two customer problems, one balance sheet

Berkshire solves two unrelated problems with the same asset. An insurance buyer facing a large or unusual risk needs a counterparty that will certainly still exist after the event occurs — a promise, not a product. The owner of a private business who wants to sell but does not want the company broken up, refinanced or resold in five years needs a permanent buyer. Both customers are paying for the same thing: a balance sheet large enough and patient enough that neither a catastrophe nor a credit market can force it to act.

That is why the two halves belong together. The insurance operation generates the money; the acquisition operation gives it somewhere permanent to go; and the size of the second is what makes the first credible.

The unit: one dollar of float

Berkshire defines float as “the approximate net policyholder funds generated through underwriting activities that are held for investment” — unpaid losses, benefit reserves and unearned premiums, less receivables and deferred acquisition costs (FY2025 10-K). Trace one dollar through it. A premium dollar arrives when the policy is written. It sits on the balance sheet for years, because a bodily-injury claim, a workers’ compensation award or a retroactive reinsurance contract pays out over a long tail. Berkshire invests it in the meantime and keeps every cent of what it earns. Eventually the claim is paid. If premiums exceeded claims and expenses, the dollar cost less than nothing to borrow.

Float was $176bn at 31 December 2025** and **$177.5bn at 30 June 2026, up from roughly $91bn at the end of FY2016. Berkshire states that its combined insurance operations produced a pre-tax underwriting gain in each of the three years to 2025 and that the average cost of float was negative in each. It does not publish a numeric cost of float; the FY2025 underwriting gain of $9,460m against average float of about $173.5bn implies a negative cost of roughly 5.4% (derived, not stated in the filing).

A second, quieter liability works the same way. Deferred and other income taxes payable stood at $87.0bn at FY2025, most of it tax owed on unrealised gains that will not be paid until the securities are sold. Float and deferred taxes together are about $263bn of funding against $717bn of equity — leverage that carries no interest, no covenants and no maturity date.

Where the earnings actually come from

Berkshire does not report an operating-earnings subtotal. It disaggregates after-tax earnings attributable to shareholders by source, which is the honest way to read it:

Source: Berkshire FY2025 Form 10-K, MD&A earnings-by-source table. The operating subtotal is derived by summing the operating lines; FY2025 investment gains of $30,737m are shown net of $8,255m of other-than-temporary impairments on Kraft Heinz and Occidental.

Two things follow. First, insurance supplies 44% of operating earnings but nearly all of the capital that produces the rest. Second, manufacturing, service and retailing is the largest single line — and the least interesting economically. It turned $214bn of revenue into $17.5bn of pre-tax profit in FY2025, an 8.2% margin. That collection of businesses matters because it absorbs capital and produces cash, not because it earns unusual returns.

What has been entering and leaving

• Pilot Travel Centers went from a 38.6% stake to control in 2023 ($8.2bn for 41.4$2.6bn). Its pre-tax margin has since fallen from 1.9% to 0.5%, and pre-tax earnings from $968m to $190m. Fuel retailing adds enormous revenue and almost no profit — the clearest illustration in the group that Berkshire’s revenue line is not a measure of its economics.
• OxyChem was agreed on 1 October 2025 and closed on 2 January 2026 for approximately $9.4bn. It contributed $2.6bn of revenue and $121m of pre-tax earnings in 1H26 — a 4.7% margin in its first half under Berkshire ownership.
• Taylor Morrison Home Corporation was agreed on 31 May 2026 at $72.50 per share, about $6.8bn, and closed on 24 July 2026 — after the balance sheet date. Purchase accounting is deferred to the Q3 2026 filing. It will be reported inside building products.
• Reporting structure has been simplified twice. General Re ceased to be a separate segment after FY2016 and now sits inside the reinsurance group; the Finance and Financial Products segment was dissolved entirely. Berkshire also bought in the last of the minority stake in BHE during 2024 and now owns it wholly.

3. Industry, Competitive Position & Moat

What the insurance industry actually earns

U.S. property and casualty insurers sell a promise to pay, and they make money two ways: an underwriting margin, and the investment return on reserves. The first is close to worthless in aggregate. Across the twelve years for which comparable data could be sourced, the industry’s combined ratio averaged about 99 and it underwrote at a profit in eight of twelve years (Insurance Information Institute for 2013–2019, NAIC for 2020–2025). Over a full cycle, underwriting is roughly a breakeven activity for the industry as a whole. The economics live in the float — which is exactly the asset Berkshire is organised around.

The industry is fragmented. The ten largest groups write 47.6% of all-lines premium and no one exceeds 11% (NAIC 2025 market share report). Personal auto is more concentrated, with the top four at 56% (NAIC/III 2023 data). Barriers that genuinely bind are statutory and rating-agency capital, and the loss-history data that makes segmentation possible. Barriers that sound impressive but do not bind include absolute scale and distribution breadth — Progressive added 3.7m policies in a single year straight through them.

Berkshire’s position: capital, not underwriting skill

The measurable moat is a capital ratio, and it is extreme. Berkshire’s U.S. insurers held 333bn of combined statutory surplus** at FY2025 against a total U.S. P&C industry policyholder surplus of **1,266bn — roughly 26% of the industry’s entire capital base standing behind 5.8% of its premium (Berkshire FY2025 10-K; NAIC 2025 year-end snapshot). The industry runs about 0.77x premium to surplus. Berkshire runs a small fraction of that.

The mechanism that converts this into earnings is investment freedom. The typical U.S. P&C insurer holds about 49% bonds and roughly 18% marketable equities, because NAIC risk-based capital and rating-agency models charge equities heavily and because reserves must be matched to payout patterns. Berkshire’s insurance subsidiaries held $294bn of equities against $529bn of investments at FY2025 — 56% in equities and only 3% in fixed maturities, with the remaining 40% in Treasury bills. The consequence is that Berkshire earns an equity return on liabilities that its competitors must fund with bonds. That single asymmetry, sustained across decades, is most of the compounding.

A second consequence is underwriting capacity. Berkshire states it will assume more single-event risk than any other insurer knowingly assumes, while avoiding aggregations above roughly $15bn of pre-tax loss from one catastrophe. Independent evidence supports that this capacity is scarce: Guy Carpenter records that only 14% of 2024’s global catastrophe losses were reinsured, against a historical average near 20%, as reinsurers raised attachment points and handed frequency risk back to primary carriers.

GEICO: a real cost advantage that is currently being spent

GEICO’s underwriting expense ratio was 12.4% in FY2025 against a U.S. industry aggregate of 25.8%, Progressive’s personal lines at 21.6% and Allstate’s property-liability at 21.4% (Berkshire FY2025 10-K; NAIC; company filings). Direct distribution removes the agency commission, and scale spreads fixed claims-handling cost. Roughly thirteen points of premium that competitors must spend on getting the policy, GEICO can keep or give back in price.

The outside evidence contradicts the comfortable reading of that fact. GEICO’s share of U.S. private passenger auto has fallen from about 13.8% (2019 data) to 12.3% (2023) to 11.6% (2024 data, per A.M. Best as cited in the FY2025 10-K), moving it from second to third in the market. Progressive gained 1.9 points of personal auto share in the first nine months of 2025 while running an 87.4% combined ratio and growing policies in force 10%. The cost advantage is real and unmatched by anyone; it has not, on its own, been enough to hold share against a competitor spending more to win it.

Rail and utilities: two regulated positions with different problems

BNSF is one of two western trunk systems in a duopoly with Union Pacific. The right-of-way was assembled in the nineteenth century and cannot be recreated; the industry has reinvested roughly $840bn of private capital since 1980 (AAR). The regulatory threat of the last cycle receded when the Seventh Circuit vacated the Surface Transportation Board’s reciprocal switching rule in August 2025, leaving the pricing structure intact.

The uncomfortable comparison is operational. BNSF’s operating ratio was 65.5% in FY2025 and 65.5% in 1H26; Union Pacific reported 59.7% in Q2 2026. Six points of operating ratio on $23bn of revenue is roughly $1.4bn of annual pre-tax earnings that BNSF does not earn. Berkshire’s filings do not make this comparison.

The live question is structural. Union Pacific’s proposed merger with Norfolk Southern sits at the Surface Transportation Board as Docket FD 36873. The application was rejected as incomplete in January 2026, refiled and accepted in April, held in abeyance in May, and revived in August 2026; comments are due 18 November 2026 and responses 16 February 2027, so no decision is possible before late 2027. BNSF has formally opposed it, arguing the combined system would control 45% of freight and eliminate around 300 intermodal lanes. Approval would leave BNSF the smaller western railroad facing a single-line transcontinental competitor. This is the largest unresolved question about the durability of a business contributing 12% of operating earnings.

Berkshire Hathaway Energy earns a legislated spread: rate base multiplied by an allowed return on equity of roughly 9.6% at about a 51% equity ratio (S&P Global RRA). Demand has turned after two decades of flat U.S. electricity consumption, and the industry is in a capital supercycle — EEI puts 2026 utility capex at $238.8bn, up 17%, and $1.4tn cumulatively through 2030. Berkshire has $10.6bn of BHE capex in FY2025 and guides to roughly $15bn across BNSF and BHE in 2026, so it is participating fully.

The offsetting exposure is wildfire liability, and it is asymmetric. PacifiCorp has accrued approximately $2.85bn of cumulative probable losses, paid about $2.3bn, and carried only $572m of unpaid liability at 30 June 2026. Against that, an Oregon jury awarded $305m to sixteen plaintiffs in February 2026 — roughly $19m each against a prior average near $5m — and S&P placed PacifiCorp’s BBB− rating on watch in March 2026, citing potential payouts approaching $50bn and warning of a fall to sub-investment grade. The Oregon Court of Appeals decertified the class in April 2026 and the Oregon Supreme Court hears argument on 3 November 2026. Berkshire does not guarantee BHE’s $61.8bn of borrowings, which limits the contagion but not the equity loss.

Testing what the company says about itself

What kind of company wins in P&C insurance? On the evidence, three kinds: the low-cost operator, the disciplined underwriter that beats the composite in every cycle position, and the firm whose capital lets it retain risk others must cede. Berkshire is decisively the third and structurally the first. It is not clearly the second — its FY2025 insurance results sit level with Progressive and behind Chubb and Allstate on combined ratio. The moat is the balance sheet, not the underwriting desk.

4. Growth Engine

Berkshire’s revenue grew 7.3% in the first half of 2026, to $195.5bn from $182.2bn. Almost none of it is what an investor would want it to be.

Source: Q2 FY2026 Form 10-Q. Pilot revenue rose on higher fuel prices with slightly lower fuel volumes, per management’s own explanation; Pilot’s pre-tax margin in 1H26 was 0.9%. Sixty-three percent of reported revenue growth came from one acquisition and a fuel-price pass-through that carries almost no margin.

The same decomposition holds inside the segments. Manufacturing revenue rose 11.6% as reported but 4.9% excluding OxyChem. Service and retailing rose 10.5% reported but 2.9% excluding Pilot. Earnings tell a better story than revenue: manufacturing pre-tax earnings rose 20.3% on organic strength at Precision Castparts (+33.6%), IMC (+56.6%) and Lubrizol (+16.5%), with OxyChem contributing only $121m of the $1,213m increase.

The growth levers, ranked

• 1. Compounding the float and redeploying retained earnings. Structural. Float grew from roughly $91bn (FY2016) to $177.5bn (Jun 2026) while Berkshire paid no dividend and retained everything. Each dollar added is interest-free funding with no maturity date. This is the only lever that worked in every year of the period.
• 2. Acquisition of whole businesses. Management-driven. $9.4bn (OxyChem) and $6.8bn (Taylor Morrison) closed in the first seven months of 2026, against $1.1bn of acquisitions in all of FY2025. This is the lever that has changed most under new leadership.
• 3. Insurance pricing and volume. Cyclical, and currently reversing. GEICO written premium grew 5.3% in FY2025 on policy count, but only 1.3% in 1H26, while U.S. personal auto prices turned negative. Berkshire’s primary group is deliberately shrinking — written premium fell 1.7% in 1H26 with RSUI down 13.2% — because management instructs underwriters to decline business priced below the risk.
• 4. BNSF volume and yield. Cyclical. Volumes rose 4.3% in 1H26 on west-coast import intermodal, market share gains and tight truck capacity, with revenue per car up 5.3%. Note that fuel expense rose 32.2% over the same period — a meaningful part of the yield gain is fuel surcharge recovery, not core price.
• 5. Regulated rate base growth at BHE. Structural but capital-hungry. Retail volumes rose 3.1% in 1H26 and electric utility margin 5.4%, funded by $4.9bn of half-year capex and $2.5bn of new borrowing.
• 6. Fuel-price pass-through at Pilot and McLane. Temporary and economically empty — tens of billions of revenue at margins of 0.9% and 1.3%.

5. Margin, Cash & Capital Allocation

Berkshire’s operating earnings compounded from $27.6bn in 2021 to $47.4bn in 2024 — 103% over five years on the company’s own measure, disclosed in the 2026 proxy. In FY2025 they fell to $44.4bn, the first decline in the series. The causes are specific and worth separating from noise: GEICO’s underwriting earnings dropped $989m as the expense ratio rose 2.7 points on advertising; the reinsurance group fell $886m on weaker prior-year reserve releases; and insurance investment income fell $1,487m pre-tax as Treasury bill yields declined. Nothing structural broke. The cycle turned.

The financial spine

Comparability warnings. (1) FY2016 pre-dates ASU 2016-01, adopted 1 January 2018 without restatement, which routed unrealised equity gains through net earnings; management states the change “significantly increases the volatility of our periodic net earnings” and that the resulting gains have “little analytical or predictive value.” Reported net earnings before and after 2018 are not comparable. (2) FY2020 operating earnings are depressed by $11.0bn of goodwill and intangible impairments. (3) FY2016 included a Finance and Financial Products segment, since dissolved, and General Re as a separate segment, since folded into the reinsurance group. (4) Cash figures for FY2024 and FY2025 are stated net of payables for unsettled Treasury bill purchases; FY2016 and FY2020 are not.

The share count line is the quiet one. Berkshire retired 12.5% of its A-equivalent shares between FY2016 and FY2024 — 206,098 shares — without issuing any, so every per-share figure above understates the growth in ownership per share. Berkshire has not declared a dividend since 1967.

Cash mechanics and capital intensity

Operating cash flow was $45.97bn in FY2025 against $30.59bn in FY2024 and $49.20bn in FY2023 — volatile, largely because $28.5bn of income taxes were paid in 2024 on the year’s enormous equity disposals versus $14.0bn in 2025. Capital expenditure was $20.9bn in FY2025, of which $14.4bn was BNSF and BHE. That is the important structural fact about the non-insurance half: BNSF and BHE consume roughly 70% of group capex to produce 21% of operating earnings, and industry data indicates that around 78% of a Class I railroad’s capex is non-discretionary maintenance. These are not capital-light toll roads. They must be rebuilt every year before a dollar of owner earnings appears.

Where the cash went, and what changed in 2026

Sources: FY2024 and FY2025 Forms 10-K, Q2 FY2026 Form 10-Q cash flow statements. 1H26 acquisitions comprise OxyChem; Taylor Morrison ($6.8bn) closed 24 July 2026 and is therefore not in these figures. FY2023 equity purchase and sale proceeds were not among the figures extracted from the sources read.

Read down the last two columns and the behaviour changes completely. Through 2024 and 2025 Berkshire was a large net seller of equities — $143bn of disposals in 2024 alone, on which it paid $101bn of taxable gains — bought back progressively less stock, and finally repurchased nothing at all in 2025 while accumulating $369bn of Treasury bills. In the first half of 2026 it bought $39.4bn of equities against $27.8bn of sales, becoming a net buyer of $11.6bn; spent $9.4bn on OxyChem and committed $6.8bn to Taylor Morrison; and resumed repurchases, spending $4.8bn, most of it in the second quarter. The cash pile fell from $369.2bn to $359.2bn — its first decline in years.

The timing is not incidental. Greg Abel became chief executive on 1 January 2026 and, per the FY2025 10-K, major capital allocation and investment decisions are his responsibility. The repurchase programme was amended in 2025 to give that authority to the chief executive after consultation with the chairman, replacing Warren Buffett by name. Berkshire filed an 8-K on 5 March 2026 stating that, “in the interest of transparency with our leadership transition,” it had commenced repurchasing shares the previous day. The hard constraint is unchanged: no repurchases that would take consolidated cash and Treasury bills below $30bn.

What the ranking reveals: over the last three years, capital expenditure has been the largest and steadiest use of cash, acquisitions the lumpiest, and buybacks entirely discretionary and price-dependent. The company will hold cash indefinitely rather than deploy it at a price it does not like — and then move $20bn in six months when it does.

6. Cyclicality, Constraints & What to Monitor

Where the business sits in its cycle

Berkshire’s largest profit engine is past its peak, and the evidence is unambiguous. U.S. personal auto insurance prices rose 20.3% year-on-year in December 2023, 11.3% in December 2024, 2.8% in December 2025 and fell 4.5% in the year to July 2026 (BLS CPI, motor vehicle insurance). That is a complete hard-to-soft transit in about thirty months, and personal auto is now in outright price deflation. The industry combined ratio tracked it: 102.5% in 2022, 101.7% in 2023, 96.9% in 2024 and 92.9% in 2025 — the best year in the series and, by construction, a peak (NAIC).

GEICO followed the same arc with a lag. Its combined ratio was 98.2% in FY2016, 90.7% in FY2023, 81.5% in FY2024, 84.7% in FY2025 and 89.3% in the first half of 2026. Underwriting earnings fell from $3,994m to $2,410m over the comparable half-years. Both blades of the scissors are moving. The loss ratio rose 4.9 points on higher claim frequency and severity, with bodily-injury severity running up 10–12%. The expense ratio rose 2.7 points to 14.0% as underwriting expenses climbed 28% on advertising and commissions — GEICO is now spending part of its structural cost advantage to defend the book rather than banking it as margin. Property-catastrophe reinsurance rates confirm the same phase, falling 12% at the January 2026 renewal after rising 27.5% in 2023 (Guy Carpenter rate-on-line index).

The second cyclical exposure is interest rates, and it is unusually direct. Berkshire holds about $360bn in Treasury bills. Insurance interest and other investment income fell 11.9% in FY2025 and 11.6% in the first half of 2026. Every fall in short rates is a mechanical reduction in an earnings line that carries no offsetting cost.

Catastrophe experience has been flattering. Significant catastrophe losses were $850m after tax in FY2025 against $1.2bn in FY2024, and there were no significant catastrophe losses at all in the first half of 2026. Independent data shows 2025 produced $107bn of global insured catastrophe losses including a record $40bn Los Angeles wildfire event, despite no U.S. hurricane landfall (Swiss Re sigma). Berkshire’s stated tolerance is to avoid aggregations from which a single event could produce more than roughly $15bn of pre-tax loss. A year in which that tolerance is tested has not occurred recently.

The downside case, as a mechanism

The mechanism that would actually hurt is a simultaneous one. Auto pricing continues to deflate while bodily-injury severity keeps running at 10–12%, pushing GEICO toward an underwriting loss of the kind it recorded in 2022; short rates fall, cutting the $10bn interest line; and a severe catastrophe year arrives. Berkshire has faced all three before — the reinsurance group lost $2.7bn pre-tax in 2020 and GEICO lost $1.9bn in 2022 — and the balance sheet absorbed it without incident. The point is not solvency, which is not in question. It is that operating earnings could fall materially from $44bn without anything structural having changed, and that a headline decline of that kind will be indistinguishable from a structural one in the reported figures.

Durable versus borrowed

Leading indicators, and where they are published

• GEICO’s loss, expense and combined ratios — Berkshire Form 10-Q, MD&A, quarterly. The expense ratio is the leading one: it moves before the loss ratio.
• CPI motor vehicle insurance, 12-month change — U.S. Bureau of Labor Statistics, monthly. The cleanest independent read on where personal auto pricing is heading.
• Cash and Treasury bills, and net equity purchases less sales — Berkshire balance sheet and cash flow statement, quarterly. This is the single best window into how the new chief executive allocates capital.
• Share repurchase dollars — Berkshire Form 10-Q, Part II Item 2, quarterly. Repurchases only occur below management’s own conservative estimate of intrinsic value, so the fact of a repurchase is itself information.
• Float, disclosed once per quarter in the MD&A liquidity discussion.
• Docket FD 36873, Union Pacific–Norfolk Southern — stb.gov. Comments due 18 November 2026; response comments 16 February 2027.
• AAR weekly rail traffic and Port of Los Angeles container statistics — the leading indicators for BNSF intermodal, which is where its current growth is.
• PacifiCorp unpaid wildfire liability — Berkshire Form 10-Q contingencies note; Oregon Supreme Court argument scheduled 3 November 2026.
• Guy Carpenter global property catastrophe rate-on-line index at the 1 January renewal — the reinsurance pricing cycle in one number.

7. Risks, Unknowns & Questions for Deeper Work

Cyclicality is covered in Section 6. What follows is what cyclicality does not capture, ordered by how much each compounds with the others.

• Equity concentration feeds back into insurance capacity. Five holdings — Alphabet, American Express, Apple, Bank of America and Coca-Cola — were 66% of a $324bn equity portfolio at 30 June 2026. Berkshire’s own risk factor states the mechanism plainly: because those holdings sit inside the insurance subsidiaries, a large decline reduces statutory surplus, which is what supports its claims-paying ratings and its ability to write new business. A severe equity drawdown therefore does not merely reduce book value — it constrains the underwriting franchise at exactly the moment a dislocated insurance market would offer the best pricing. This is the one risk that compounds with every other risk in this list.
• PacifiCorp’s wildfire tail is orders of magnitude larger than the accrual. $572m of unpaid liability was carried at 30 June 2026 against remaining unsettled Oregon and California claims that Berkshire itself totals at approximately $50bn before any doubling or trebling. Oregon law permits doubling of economic damages for gross negligence and trebling for vegetation trespass. S&P has PacifiCorp’s BBB− on watch. Berkshire does not guarantee BHE’s debt, so the loss is bounded by BHE’s equity, but that equity is the vehicle through which Berkshire participates in the utility capex supercycle — an impairment closes off a growth avenue as well as destroying capital.
• Reserve leverage on a $152.9bn liability. Berkshire notes that a small percentage increase in unpaid property and casualty losses materially reduces reported earnings; about 75% of the balance sits at GEICO and the reinsurance group. Social inflation — third-party litigation funding, expanding liability theories, larger verdicts — is the named driver, and bodily-injury severity running 10–12% is exactly what an emerging reserve problem looks like early. Note that 1H26 earnings were flattered by $869m of prior-year releases at reinsurance; releases and strengthening are the same lever pulled in opposite directions.
• The allocator has changed and the track record has not been established. Greg Abel became chief executive on 1 January 2026 with sole responsibility for major capital allocation. He has deployed roughly $20bn in his first seven months — OxyChem, Taylor Morrison, $4.8bn of buybacks and $11.6bn of net equity buying — after a year in which his predecessor deployed almost nothing. Both approaches can be right; they cannot both be right about the same opportunity set. Berkshire’s central mechanism is one person’s judgement applied to $360bn, and there is no historical record of this person applying it at this scale.
• An unresolved governance question sits behind the operating one. Warren Buffett, aged 95, holds 30.2% of voting power against 13.7% of economic interest and is deemed Berkshire’s controlling shareholder; a voting agreement caps him at 49.9%. The proxy states the board discusses succession at every meeting and that Buffett believes a family member should serve as non-executive chairman after his death, while noting the decision belongs to the then-board. The disposition of that voting block over time, and the governance structure that replaces it, is not described anywhere in the sources read.
• Scale itself constrains the return. At $717bn of equity with $360bn in Treasury bills, the number of acquisitions large enough to move the result is very small, and Berkshire must compete for them against private equity and strategic buyers. The evidence is in the figures: $2.9bn of buybacks in 2024 and none in 2025 while cash rose to $369bn is not a capital allocation strategy so much as an absence of qualifying opportunities.

What the sources could not answer

• GEICO’s absolute policies-in-force count is not disclosed in any filing read — only percentage changes. Without it, share loss cannot be separated from market contraction.
• Berkshire publishes no numeric cost of float, only the statement that it has been negative. The 5.4% figure used here is derived.
• Per-security cost basis and fair value for individual equity holdings ceased to be disclosed after FY2020; only three industry categories are given.
• BNSF’s revenue per car is given only as a percentage change, so the split between core price, mix and fuel surcharge cannot be computed from the filings.
• Todd Combs and Ted Weschler are not named anywhere in the 2026 proxy or the FY2025 10-K. Their current roles and the share of the portfolio they manage are unknown, which matters more now that the person who hired them is no longer chief executive.
• Charles Chang succeeds Marc Hamburg as chief financial officer on 1 June 2026; his compensation arrangements were expressly “not finalized” at the time of the 8-K.
• No Form 4 filings were available in the source archive, so insider transactions could not be examined directly. Buffett’s Class A holding fell 4.9% between the 2025 and 2026 proxies; the filings do not explain the mechanism.

Before forming a view, three things would need resolving: whether GEICO’s expense ratio increase is buying policy growth or merely defending a shrinking book; whether the Oregon Supreme Court restores the PacifiCorp class certification; and how Abel intends to run the repurchase programme, which is the one capital allocation decision an outside shareholder can observe in near-real time.

8. Investor Takeaways

• What this business really is. An insurance company that has been paid to borrow $177.5bn, wrapped around a capital allocator who invests that borrowing in equities and whole businesses, with a railroad, a utility and a large low-margin industrial group attached.
• The core economic engine. Negative-cost float invested at equity-like returns. Holding 26% of the U.S. industry’s statutory capital behind 5.8% of its premium is what permits a 56% equity allocation where competitors manage 18% — and that asymmetry, not underwriting skill, is the moat.
• The main growth lever. Deployment of retained capital. Everything else — insurance pricing, rail volumes, rate base — is cyclical or capital-hungry. The $20bn committed in Abel’s first seven months is the lever being pulled now.
• What could break the story. A severe equity drawdown that simultaneously reduces statutory surplus and constrains underwriting capacity, arriving alongside a soft insurance market and a bad catastrophe year. The three are correlated, which is the point.
• What to monitor. GEICO’s expense and combined ratios; the direction of the cash balance and net equity purchases; and Docket FD 36873 at the Surface Transportation Board.