This is investment research, not personal financial advice.

BERKSHIRE HATHAWAY INC.'s Class B shares were trading at about US$533.535 at 10:50 a.m. ET on Monday, 10 August, up 2.25% from Friday's US$521.80 close after Berkshire filed its second-quarter report on Saturday (Yahoo Finance 2026; Berkshire 2026 Q2). That is an intraday observation, not a completed reaction-day close. The early move was favourable, but it does not prove that the filing caused the whole move or settle what the capital was worth.

Berkshire's latest uses of capital matter only if they can improve per-share economics at a company with US$359.2 billion of June liquidity. This audit separates gross activity from net deployment, investment income from operating earning power, and insurance float from equity. It then asks what assumptions the roughly US$533.54 quote embeds.

What Monday’s move was reacting to, and what remains hidden

The eye-catching Q2 figures are real. Berkshire bought US$23.467 billion of equity securities during the quarter and sold US$3.693 billion, producing US$19.774 billion of Q2-only net equity purchases. That is an author calculation: US$23.467 billion minus US$3.693 billion. First-quarter activity had gone the other way, which is why the half-year totals are less dramatic: US$39.405 billion bought, US$27.780 billion sold and US$11.625 billion net bought (Berkshire 2026 Q1; Berkshire 2026 Q2).

Berkshire also repurchased about US$4.5 billion of its own shares in Q2. The cash-flow statement records US$4.444 billion of H1 repurchase payments, while the treasury-stock acquisition disclosure rounds the H1 cost to about US$4.8 billion, mostly incurred in Q2. The distinction is timing and presentation rather than an economic gulf. May and June Class B repurchases occurred around US$476–US$488, including a June average of US$487.98, below Monday morning’s quote (Berkshire 2026 Q2).

What the filing did not disclose is equally important. Berkshire’s Q2 Form 13F was not available at the market snapshot, so the US$23.467 billion gross purchases cannot yet be mapped confidently to named US-listed holdings. A 13F is also a quarter-end positions report, not a complete transaction ledger: it omits some foreign securities, cash, private businesses and certain instruments. Until that filing appears, assertions about which company Berkshire “bought” are inference rather than evidence. Monday’s reaction was thus to aggregate allocation data, not a fully identified portfolio.

Berkshire’s business model is an internal capital market

Berkshire is often described as an insurer or conglomerate, but its economic design is an internal capital market with four main allocation channels. First, GEICO, General Re, Berkshire Hathaway Reinsurance Group and primary insurers collect premiums before claims are paid. That creates float. Second, the group owns cash-generating operating businesses: BNSF’s rail network, Berkshire Hathaway Energy’s regulated and contracted assets, and a broad manufacturing, service and retailing collection. Third, it holds a concentrated public-equity portfolio. Fourth, the parent can move retained cash among securities, whole-company acquisitions, subsidiary reinvestment and Berkshire share repurchases without routinely paying a dividend (Berkshire 2025).

The model’s advantage is breadth. A conventional insurer can invest its portfolio; an industrial group can reinvest or acquire; Berkshire can do both while retaining permanent corporate capital. BNSF’s more than 32,500 route miles, energy networks and regulated assets are difficult to reproduce. Decentralised management keeps headquarters small and gives subsidiary managers autonomy, while the parent decides where incremental capital goes (Berkshire 2025).

Scale also creates the central problem. At an equity value of roughly US$1.14 trillion, a successful US$5 billion acquisition barely changes group value. The liquidity reserve can absorb a major catastrophe or market seizure, but every undeployed dollar earns a resetting short-term rate rather than a business return. OxyChem and Taylor Morrison widen the operating set into chemicals and housing, yet they also add cyclical sensitivity. The architecture is unusually resilient; it is not automatically high-return at any purchase price or at every scale.

Five reporting periods show cash strength and noisy accounting profit

The history below puts the current deployment burst in context. All amounts are US$ billions. Author calculations are explicitly identified; H1 2026 is a six-month period and is not directly comparable with full years.

Period Revenue NPAT Operating cash flow Capex OCF less capex* Equity purchases Equity sales Net equity purchases* Repurchases Float
2022 302.020 (22.759) 37.350 15.464 21.886 67.930 33.664 34.266 7.854 164.0
2023 364.482 96.223 49.196 19.409 29.787 16.462 40.631 (24.169) 9.171 169.0
2024 371.433 88.995 30.592 18.976 11.616 9.237 143.359 (134.122) 2.918 171.0
2025 371.444 66.968 45.969 20.927 25.042 16.923 30.686 (13.763) 0 176.0
H1 2026 195.483 35.773 21.653 10.631 11.022 39.405 27.780 11.625 4.444 177.5

* Author calculations: OCF less capex equals operating cash flow minus capital expenditure; net equity purchases equals purchases minus sales. Historical figures are drawn from Berkshire’s annual and interim reports (Berkshire 2022; Berkshire 2023; Berkshire 2024; Berkshire 2025; Berkshire 2026 Q2).

Revenue grew at an author-calculated 7.1% compound annual rate from 2022 to 2025, but virtually none of that growth occurred in 2025. NPAT moved from a US$22.8 billion loss in 2022 to US$96.2 billion in 2023, then US$89.0 billion and US$67.0 billion. That is not an operating collapse. Accounting rules route unrealised public-equity gains and losses through earnings, making consolidated NPAT a noisy mixture of business performance and market marks. H1 2026 NPAT of US$35.773 billion included US$11.444 billion of after-tax investment gains (Berkshire 2026 Q2).

Cash tells a steadier but still lumpy story. OCF less capex ranged from US$11.6 billion to US$29.8 billion in the four full years. The low 2024 outcome coincided with enormous net equity sales, not weak liquidity. Float rose each period, from about US$164 billion in 2022 to US$177.5 billion in June 2026. The ledger shows why a single earnings multiple or one quarter’s purchases cannot describe Berkshire’s allocation cycle.

The deployment audit: US$32.625 billion is acceleration, not transformation

Two whole-company transactions frame 2026. Berkshire closed its acquisition of OxyChem on 2 January for about US$9.4 billion in cash, with Occidental retaining legacy environmental liabilities. It then closed Taylor Morrison on 24 July for about US$6.8 billion. Taylor Morrison is a post-quarter event, so June cash and June operating results do not include it (Berkshire 2026 Q2; Berkshire 2026 results).

A consistent through-24-July deployment definition is:

  • H1 net public-equity buying: US$11.625 billion;
  • OxyChem: US$9.4 billion;
  • Taylor Morrison: US$6.8 billion; and
  • H1 treasury stock acquired: about US$4.8 billion.

That totals US$32.625 billion. The author calculation is US$11.625 billion + US$9.4 billion + US$6.8 billion + US$4.8 billion. It equals an author-calculated 9.1% of US$359.2 billion June liquidity. The acquisitions alone used US$16.2 billion, or 4.5% of that pool. This definition uses half-year net listed-equity activity, rather than adding Q2’s US$19.774 billion net amount to the acquisitions, because Q1 sales partly funded Q2 buying. Using gross purchases or mixing quarterly and half-year windows would overstate the balance-sheet change.

Repurchases deserve separate treatment. They reduce the share denominator only when price is below management’s estimate of intrinsic value and consolidated cash and Treasury bills remain above US$30 billion. Buying around US$488 offers evidence about management’s assessment at those transaction prices, but it does not create a mechanical floor at US$533.54. Whole-company acquisitions likewise convert cash into operating assets on day one; they create value only if future after-tax cash returns exceed the foregone cash yield and risk-adjusted hurdle. No acquisition synergy is included in the valuation below.

Float is Berkshire’s funding advantage, not free equity

Insurance float reached approximately US$177.5 billion at June 30, up US$1.1 billion from year-end. Berkshire reported that float’s average cost was negative in both H1 2025 and H1 2026 because combined insurance operations produced underwriting profits (Berkshire 2026 Q2). Economically, being paid to hold long-duration funds is a rare financing advantage.

But float is still generated by claim obligations. Its amount, duration and cost depend on pricing, catastrophe experience and reserve adequacy. It has neither ordinary debt’s fixed maturity schedule nor common equity’s permanent freedom from repayment. Calling all US$177.5 billion “free capital” ignores claims; deducting every dollar at par ignores the value of durable negative-cost funding.

That ambiguity is why the scenario analysis applies an explicit economic float charge from 0% to 100%. Every 25 percentage-point change moves equity value by about US$20.7 per B share. The author calculation is US$177.5 billion multiplied by 25%, divided by 2.1475 billion June B-equivalent shares. Float treatment is therefore not a footnote. It is one of the largest judgments in Berkshire valuation, and it depends on underwriting evidence.

GEICO is the clearest warning inside an otherwise profitable half

H1 insurance underwriting earnings rose from US$3.328 billion to US$3.448 billion, but the aggregate conceals sharp deterioration at GEICO. GEICO underwriting earnings fell 39.7%, from US$3.994 billion to US$2.410 billion, while its combined ratio worsened from 81.7% to 89.3%. Claims frequency and bodily-injury severity rose, and the underwriting expense ratio increased by 2.7 percentage points (Berkshire 2026 Q2).

The absence of a significant 2026 catastrophe makes that signal cleaner, not easier to dismiss. Group underwriting benefited from favourable reserve development, including roughly US$1.5 billion of reductions in prior-year ultimate claim estimates. Those releases may not recur. GEICO’s deterioration instead concerns current personal-auto economics and operating expense. A ratio below 100 still means an underwriting profit, and 89.3% is not a crisis ratio, but the speed of the change challenges the proposition that scale and data produce continuously superior margins.

Peer context prevents an overly Berkshire-centric reading. Chubb’s large global property-and-casualty franchise traded around 1.83 times recent book in the packet’s same-day comparison, versus about 1.53 times for Berkshire, although Chubb lacks Berkshire’s railroad, utility and immense net cash block (Chubb 2026). Berkshire’s cheaper book multiple is not proof of cheaper operating earnings; its asset mix is fundamentally different. The operational test is whether GEICO restores pricing, frequency and expense discipline while total float remains negative-cost.

Owner earnings and an operating ROIC proxy require wide error bars

Berkshire’s H1 after-tax segment earnings were US$3.448 billion from underwriting, US$2.935 billion from BNSF, US$2.005 billion from BHE and US$7.669 billion from manufacturing, service and retailing. Doubling those four blocks gives an author-calculated US$32.114 billion annualised core-earnings anchor. It excludes insurance investment income, mark-to-market gains, equity-method income and foreign-exchange effects because the related financial assets are separately valued below (Berkshire 2026 Q2).

The cash cross-check is lower. H1 OCF of US$21.653 billion less US$10.631 billion of capex leaves an author-calculated US$11.022 billion, or US$22.044 billion annualised. That subtraction is deliberately blunt. BNSF and BHE spent US$6.689 billion of H1 capex against US$3.525 billion of depreciation and amortisation, and forecast another US$8.6 billion in H2. Some excess is growth capital with future value; some is maintenance, safety and reliability spending required to preserve current earnings. Neither treating all capex as maintenance nor treating depreciation as the full economic charge is credible.

An author-estimated operating ROIC proxy illustrates the uncertainty. Start with June equity of US$747.910 billion, remove after-tax public equities of US$278.155 billion, excess cash/T-bills above a US$30 billion reserve of US$329.229 billion, fixed maturities of US$17.034 billion and equity-method stakes of US$19.948 billion, then add US$43.302 billion of Insurance and Other debt. The residual operating equity is US$146.846 billion. Adding approximately US$85.3 billion of debt housed in BNSF and BHE gives US$232.146 billion of indicative operating capital. The US$32.114 billion earnings anchor divided by that amount is 13.8%.

This is not standard reported ROIC. It uses point-in-time rather than average capital, an after-tax estimate for listed equities, after-interest segment earnings rather than NOPAT, and a residual balance-sheet method that cannot cleanly allocate insurance liabilities, goodwill or working capital. Excluding BNSF/BHE debt would raise the same proxy to 21.9%, demonstrating denominator sensitivity. The defensible conclusion is a range of owner-earnings outcomes, not false precision: the scenarios use US$22–US$38 billion, with US$32.1 billion as an accounting anchor.

The balance sheet preserves optionality but rates set its opportunity cost

At June 30, Insurance and Other held US$35.096 billion of cash and equivalents and US$324.905 billion of short-term US Treasury bills, less US$0.771 billion of unsettled-purchase payables. That is an author-calculated US$359.229 billion of net liquidity. Public equities were worth US$323.779 billion against US$106.521 billion of GAAP cost; fixed maturities were US$17.034 billion; equity-method holdings were US$19.948 billion; Insurance and Other borrowings were US$43.302 billion; and shareholders’ equity was US$747.910 billion (Berkshire 2026 Q2).

This is crisis capacity, but not costless idleness. The three-month Treasury yielded 3.90% on 6 August, the 10-year nominal Treasury 4.69%, and the 10-year real Treasury 2.43% (FRED 2026a; FRED 2026b; FRED 2026c). At 3.90%, US$359.229 billion would produce an estimated US$14.0 billion of annual pre-tax carry, before changes in mix and balance. A 100-basis-point rate move changes pre-tax annual income by about US$3.59 billion. H1 insurance investment income already fell 8.3% to US$5.738 billion as rates declined, showing that the cash engine reprices quickly (Berkshire 2026 Q2).

The real hurdle matters for OxyChem, Taylor Morrison and future deals. Returns must compensate for cyclicality and execution risk above a 2.43% real sovereign yield, while long-duration rail and utility values face a 4.69% nominal discount-rate reference. Conversely, falling rates reduce cash income but may lift asset values and ease housing conditions. Liquidity is both an earnings asset and an option; macro changes pull those values in opposite directions.

Float-adjusted sum-of-the-parts spans US$230 to US$685

The primary framework values financial assets separately from operating franchises so investment income is not counted twice. All figures and scenario outputs in this section are author estimates or author calculations, not forecasts. The financial block equals net cash/T-bills less a US$30 billion required reserve, plus after-tax public equities, fixed maturities and equity-method stakes, less Insurance and Other debt and an explicit float charge. At the June mark, estimated after-tax public equities are US$278.155 billion. The author estimate subtracts 21% of the US$217.258 billion difference between fair value and GAAP cost from US$323.779 billion. Actual tax basis, timing and deductions differ.

Operating value equals normalised core owner earnings multiplied by a scenario multiple. Division uses 2.1475395 billion June B-equivalent shares so the denominator matches the June balance sheet.

Scenario Key author-estimated assumptions Per-B scenario output*
Severe Stocks down 35%–45%; 100% float charge; US$22–US$25bn owner earnings; 9–11x US$230–US$285
Bear Stocks down 15%–25%; 50%–75% float charge; US$27–US$30bn; 12–14x US$335–US$420
Base Stocks down 5% to up 5%; 0%–35% float charge; US$31–US$34bn; 16–18x US$475–US$575
Bull Stocks up 15%–25%; 0% float charge; US$35–US$38bn; 19–21x US$605–US$685

*Rounded author-calculated outputs based on Berkshire’s June balance sheet. These are assumption-defined scenario ranges, not probabilities.

The severe range combines a deep equity decline, par treatment of float and owner earnings near annualised OCF less capex. The bull range requires more than portfolio appreciation: float must behave as durable negative-cost capital, core earnings must rise and operating franchises must retain premium multiples. Sensitivities are large. A 10% public-equity move changes estimated after-tax value by roughly US$11.9 per B share; one multiple turn on US$32.1 billion changes value by about US$15 per share. The wide spread is information about the model’s dependence on float, earnings quality and discount rates, not statistical confidence.

Reverse valuation: what the intraday quote already assumes

Berkshire’s market capitalisation was an author-calculated US$1,142,143.7 million, or about US$1.142 trillion. The calculation uses the filed B-equivalent share count and the intraday quote. June book value was US$348.26 per B-equivalent share, making the US$533.535 quote approximately 1.53 times book. GAAP P/E is less useful because trailing earnings include changing investment marks.

Reverse the sum-of-the-parts instead. With no economic float charge, the current-mark financial block is about US$601.1 billion, leaving about US$541.1 billion for operations at the intraday equity value. Against US$32.114 billion of annualised core earnings, that is an estimated 16.8 times. Charging 35% of float reduces the financial block by US$62.1 billion and raises the residual operating multiple to roughly 18.8 times, or requires around US$35.5 billion of core owner earnings at 17 times. Treating float fully at par raises the implied operating multiple to roughly 22.4 times, or requires about US$42.3 billion at 17 times.

The quote therefore sits in the upper half of the base scenario. It can be reconciled with the US$32.1 billion anchor if float receives almost full economic credit and operations receive a high-teens multiple. A more debt-like treatment of float requires either stronger sustainable owner earnings or a richer operating multiple. The reverse exercise identifies the embedded bargain between assumptions; it does not establish which treatment is correct.

Moat evidence has unusually specific counter-evidence

The moat case starts with negative-cost float, extreme liquidity and deployment flexibility. It includes irreplaceable rail rights-of-way and energy networks, a decentralised culture attractive to owner-managers, tax-efficient retention and a demonstrated willingness to wait. Berkshire’s 2025 property-and-casualty combined ratio was 87.1%, better than its five-, ten- and twenty-year averages, and 2025 OCF was US$46.0 billion (Berkshire 2025).

Against that stand several facts. GEICO’s current deterioration is operational, not catastrophe-driven. Alphabet, American Express, Apple, Bank of America and Coca-Cola represented 66% of the US$323.8 billion public-equity portfolio, making concentration a source of both strength and downside. PacifiCorp had recorded US$2.85 billion of cumulative probable wildfire losses, with US$572 million unpaid at June 30; Berkshire said material additional loss was reasonably possible but could not estimate a range. BNSF and BHE require heavy continuing capital. OxyChem and Taylor Morrison increase exposure to chemical and housing cycles. Greg Abel's tenure as CEO is also a live test of whether culture and judgment are institutional rather than personal (Berkshire 2026 Q2; AP 2026).

These are not generic risk factors. Persistent positive-cost float would directly weaken the funding moat. Core owner earnings staying near the US$22 billion cash cross-check would undermine a premium operations value. Repeated acquisitions earning below the real Treasury hurdle would turn optionality into drag. Conversely, a restored GEICO cost ratio, core earnings above US$35 billion excluding investment income and repeated attractive deployment above US$30 billion annually would contradict the stronger bear assumptions.

The crux timeline and thresholds that resolve the audit

The next evidence arrives in stages. The Q2 13F should identify many quarter-end US-listed positions, while still not reconstructing every trade. The Q3 report should show whether equity deployment continued after June, whether the July Taylor Morrison cash outflow changed liquidity materially and whether GEICO’s deterioration persisted. The 2026 annual report should provide a fuller year of OxyChem contribution and capital allocation under Abel. Taylor Morrison and OxyChem returns, impairments and cash conversion require several reporting periods rather than a closing-day inference.

The monitoring thresholds below are author-defined analytical thresholds, not company guidance:

  • GEICO: a combined ratio returning below 85% would indicate material restoration; 90% or worse for two further quarters would confirm deterioration beyond one half.
  • Float: roughly US$175 billion or more with negative cost preserves the recent funding case; a decline greater than 5% or persistent underwriting losses would increase the economic liability charge.
  • Core earnings: above US$35 billion excluding investment income would support the upper operating assumptions; below US$27 billion would align more closely with bear inputs.
  • Cash conversion: annualised OCF less capex above US$30 billion would narrow the owner-earnings gap; around US$22 billion despite heavy investment would keep the low-end caveat active.
  • Deployment: more than US$30 billion a year of defined net deployment becomes material only if subsequent cash returns exceed cash and real-rate hurdles.
  • Reserves: an additional after-tax reserve, wildfire or litigation charge above US$5 billion would be economically visible, though still absorbable; the model sensitivity is roughly US$2.3 per B share before tax effects for each US$5 billion charge.
  • Liquidity: a falling balance is not inherently negative; the distinction is whether it reflects value-accretive assets, repurchases and claims, or operating cash weakness.

Source notes and an observational verdict on an unfinished reaction

Monday morning’s 2.25% rise was a favourable but provisional reaction to faster capital deployment and resilient aggregate operations. The filing supports the narrow claim that Berkshire became a net equity purchaser in H1 and active in Q2. It does not support the larger claim that US$359.2 billion of liquidity has been transformed: defined net deployment through 24 July was about US$32.625 billion, or 9.1% of that pool.

At about US$533.54, the market was not pricing a severe dislocation. The quote sat in the upper half of the author-estimated base range and required generous economic treatment of float, high-teens operating valuation, or owner earnings above the US$32.1 billion anchor. GEICO, cash-rate sensitivity, capital intensity and acquisition returns remain the principal tests.

Confidence is highest in reported balance-sheet totals and transaction arithmetic, moderate in the core-earnings separation, and lowest in residual operating capital, float treatment and scenario multiples. The unavailable Q2 13F prevents identification of the equity purchases; Monday’s session was incomplete at the observation time; Taylor Morrison was absent from the June balance sheet; acquisition synergies are unmodelled; and the ROIC proxy cannot fully allocate insurance liabilities or maintenance capital. The reaction verdict is therefore observational: the early tape endorsed the direction of allocation, while the filing left the return on that allocation unresolved.

References