Updated Q3 2026 Balance-sheet marks and share count from the Q3 2026 Form 10-Q (September 30, 2026); earnings multiples from the 2025 Form 10-K. Base case $524 per B share. Earlier estimates.
Book value per share rose +7.6% in nine months to $357.73 — an annualised +10.2%, and on track for the 10% a year this valuation assumes.
The cash pile shrank because everything else grew. Net of the $20.5 B owed for Treasury bills already bought, cash and Treasuries fell to $352.3 B from $369.0 B. The $372.8 B headline is a record only because it ignores that payable.
Operating earnings rose 8.0% to $37.0 B — but strip the currency gain on Berkshire's yen and euro borrowings and the rise is 3.5%, and the third quarter goes from -5.8% to -0.8%. Insurance did the damage: underwriting profit fell 16% and investment income 10%, while BNSF, Berkshire Hathaway Energy and the manufacturing and retail businesses each grew by double digits.
Most significantly, Berkshire bought $14.4 B of its own stock at an average of $497.12, retiring 0.9% of the shares in the quarter and 1.3% year to date, after buying none in 2025. It closed a $6.8 B acquisition of Taylor Morrison in July, and turned from a net seller of stocks into a net buyer.
3Q26 Berkshire facts that matter
| 30 Sep 2026 | 31 Dec 2025 | Change | |
|---|---|---|---|
| Cash & Treasuries, net of the T-bill payable | $352.3 B | $369.0 B | -4.5% |
| Quoted equities, at market | $335.5 B | $297.8 B | +12.7% |
| Equity method investments | $20.7 B | $20.0 B | +3.4% |
| Fixed maturity securities | $17.1 B | $17.8 B | -4.1% |
| Insurance float | $178.5 B | $176.4 B | +1.2% |
| Parent company debt | $20.4 B | $22.6 B | -9.7% |
| Total assets | $1,303.9 B | $1,222.2 B | +6.7% |
| Shareholders' equity, excluding minorities | $761.4 B | $717.4 B | +6.1% |
| Equivalent B shares, billions | 2.128 | 2.157 | −1.3% |
| Book value per B share | $357.73 | $332.55 | +7.6% |
Book value: where the +7.6% came from
Equity rose $43.9 B: $57.9 B of net earnings in, less $14.4 B spent on buybacks. The composition matters more than the total — $37.0 B was operating profit, but $20.9 B was after-tax gains on the quoted portfolio, a mark rather than a cash flow, which reverses when the market does. Roughly a third of this year's book value growth is the stock market having a good year, which is why the twenty-six-year record of 10.5% is the number worth trusting.
The buybacks work against the measure, by design. Repurchasing at $497.12 — about 1.4× book — lowers book value per share by roughly $1.87; without them book would have grown 8.1%. That is the trade working as intended: an accounting measure swapped for intrinsic value per share.
Operating earnings: the headline is a currency swing
The third quarter alone fell 5.8%; the nine months rose 8.0%. Neither number is what it looks like, and the reason is one line in the "Other" segment.
Berkshire's non-dollar senior notes are revalued into dollars every quarter and the swing runs through earnings. That line earned $0.2 B in the nine months against a loss of $1.3 B a year earlier — a $1.5 B reversal, non-cash, with no operating content at all. It is 54% of the entire nine-month increase in operating earnings. In the third quarter it ran the other way, which is most of why the quarter printed a decline.
| After-tax operating earnings | 3Q 2026 | 9M 2026 |
|---|---|---|
| As reported | -5.8% | 8.0% |
| Excluding the currency revaluation only | -0.8% | 3.5% |
| Excluding the whole "Other" segment | -5.2% | 1.3% |
The honest reading is the middle row: roughly flat in the quarter and up 3.5% in nine months. The bottom row strips "Other" entirely — currency, equity-method earnings from Occidental and Kraft Heinz, and holding-company interest — and leaves nine-month growth of 1.3%. The operating businesses Berkshire actually runs grew about one per cent.
All three underwriting units turned
| Pre-tax underwriting profit | 3Q 2026 | 3Q 2025 | 9M 2026 | 9M 2025 |
|---|---|---|---|---|
| GEICO | $982 M | $1,773 M | $3,392 M | $5,767 M |
| Berkshire Hathaway Primary | $119 M | $506 M | $868 M | $425 M |
| Berkshire Hathaway Reinsurance | $582 M | $884 M | $1,868 M | $1,227 M |
| Total | $1,683 M | $3,163 M | $6,128 M | $7,419 M |
Every one of the three fell in the quarter — the first clean sweep down in years, and the 10-Q says so in its own words. The nine-month picture looks kinder only because of two things that will not repeat: Berkshire had no significant catastrophe losses at all in the first nine months of 2026, against $1.1 B pre-tax of Southern California wildfire losses in the first quarter of 2025, and prior-year reserve releases across the group were larger this year than last. Take out the reserve releases and the absent catastrophes and the underwriting year is roughly $4.5 B rather than the $6.1 B reported. A cat-free year that still loses a fifth of its underwriting profit is the finding, not the headline.
Insurance — underwriting profit plus investment income — earned $13.3 B after tax in the nine months to September 30, 2026, against $15.1 B in the same nine months of 2025. Everything else — the railroad, the utilities, the manufacturing and retail businesses and the holding company — earned $23.8 B against $19.1 B. In one year insurance has fallen from 44% of operating earnings to 36% — the biggest change in the shape of Berkshire's earnings in this filing, and almost entirely GEICO.
Net earnings were $57.9 B against $47.8 B, the difference being $20.9 B of after-tax investment gains against $13.5 B.
What got better, and what got worse
Better, in order of what it is worth:
- BNSF's productivity is real and repeatable. Volumes up 4.2% against compensation up 0.6%. A sustained wedge between volume and labour cost is worth more in a business this capital-intensive than any pricing cycle.
- Precision Castparts has re-rated operationally — pre-tax earnings up 32% on revenue up 13.9%, driven by aerospace and by industrial gas turbines for power generation. It is capacity-constrained rather than demand-constrained, which is the better problem.
- Berkshire Hathaway Energy is growing on load, not rates. Retail electricity volumes 5.5% higher.
- The buyback restarted and then accelerated — nothing in 2025, then $14.4 B, two thirds of it in the third quarter.
- Underwriting discipline at the primary insurers. Premiums written fell 4.0%; walking away from business priced below the risk is the behaviour that makes the franchise worth a premium, even when it makes this year's profit worse.
Worse, in the same order:
- GEICO's cycle has turned, and it is still early. Detailed below; it is the largest open question in the valuation.
- The nine-month improvement at reinsurance and the primary group is reserve releases and the absence of catastrophes, not underwriting. On an accident-year basis both books deteriorated.
- Investment income is in structural decline. Interest income fell 10.3% after tax on a larger cash balance. The fixed-maturity book is a parking lot with almost no duration, so every fall in short rates lands immediately and in full. On $352.3 B of cash and Treasuries, Berkshire takes the whole rate move with nothing offsetting it — the price of insisting on safety over yield, and a cost that compounds into 2027.
- Reported growth is substantially manufactured. Nine-month operating earnings 8.0% becomes 1.3% without "Other"; segment revenue growth is fuel prices and two acquisitions.
- Low-quality items are propping up several businesses — refunds of trade tariffs paid in 2025, gains on business and property disposals, and a prior-year tax-credit catch-up at Duracell that makes the consumer comparison meaningless. Strip them and building products, consumer products and retailing are flat to down.
- Housing and the consumer are weak wherever they appear: factory-built home volumes down at Clayton on lower margins, other building products' profit down in the quarter, the real-estate brokerage at close to nothing, vehicle gross margins down at the dealerships, and McLane's retail sales down 9%.
- Float has stopped growing. Up $2.1 B in nine months, because $1.1 B of retroactive-reinsurance commutation and ordinary run-off offset the new business. A flat float with a rising cost of float is the quiet erosion of Berkshire's cheapest funding.
The through-line. The two best businesses and the best capital allocation got better, while the two biggest profit pools — underwriting and the income on the float — deteriorated at the same time: one cyclically and visibly at GEICO, one quietly across the rest of insurance under cover of reserve releases, and one mechanically through short rates. The consolidated result was then smoothed by a currency swing on yen debt. Underlying operating earnings are roughly flat; the offset is a railroad and an aerospace business that are genuinely compounding, and, for the first time since 2024, a repurchase programme being used at scale.
Capital allocation: $33.9 B in, $57.3 B out
| Nine months to 30 Sep 2026 | $ billions |
|---|---|
| Cash flow from operations | 33.9 |
| Acquisitions of businesses, net of cash acquired | (16.2) |
| Capital expenditure | (16.1) |
| Treasury stock purchased | (13.8) |
| Net equity purchases | (11.2) |
| Dividends paid | 0.0 |
| Total deployed | (57.3) |
Berkshire spent about $23.4 B more than it earned. That is where the cash went, and it is what the end of a cash build looks like.
Acquisitions, $16.2 B. OxyChem closed in January for $9.4 B and has earned $0.2 B pre-tax on $3.9 B of revenue — thin, and a reminder that chlor-alkali is bought for a cycle. Taylor Morrison closed in July for $6.8 B, putting Berkshire into site-built homebuilding at scale. The two brought in $16.2 B of net assets with only $1.8 B of goodwill and intangibles: Berkshire paid close to tangible value, which is unusual at this size.
Capital expenditure, $16.1 B, about $10.2 B of it the railroad and the utilities, with $4.7 B more guided for the fourth quarter.
Buybacks, $14.4 B, accelerating all year — $0.2 B in 1Q 2026, then $4.5 B in 2Q 2026, then $9.6 B in 3Q 2026. September was both the heaviest month and the dearest, at $506.64 a B share against $496.24 in July: a company buying with more conviction as the price rose.
That average price is the most useful number in the filing. This site's independently derived base case is $524 per B share. Berkshire's nine-month average of $497.12 is 5.1% below it, and its third-quarter average 4.1% below. Two quarters running, on two different balance sheets, the people who know most about what Berkshire is worth have been buying just under where this analysis puts the value.
Net equity purchases, $11.2 B, against $10.6 B of net selling a year earlier: $41.8 B bought against $30.6 B sold, realising $11.9 B of taxable gains. Alphabet is now among the five largest holdings and is very probably the largest single purchase of the quarter, though position-level confirmation waits on the 13F due mid-November. No dividend, as ever.
What nine months says about the assumptions
The valuation applies a multiple to a normalised profit for each segment, and those figures move slowly. A segment running persistently away from its normalised figure is evidence.
GEICO needs attention. It is the largest open question in the valuation, and underwriting is the assumption a reader can change directly on the valuation page.
| GEICO | 3Q 2026 | 3Q 2025 | 9M 2026 | 9M 2025 |
|---|---|---|---|---|
| Premiums earned | $11.4 B | $11.3 B | $33.9 B | $33.1 B |
| Combined ratio | 91.4% | 84.3% | 90.0% | 82.6% |
- The underlying deterioration is worse than the reported ratio. Strip the prior-year reserve releases out of both years and the nine-month combined ratio is 91.8% against 84.1% — 7.7 points of accident-year decay, not the 7.4 the headline shows. And the releases are growing, $628 M against $495 M, which cushions the print and defers the recognition.
- Severity is the problem, and it is in the worst coverage. Bodily injury frequency up 4–6% and severity up 9–11% — long-tail, high-severity claims, and the coverage where reserve adequacy is hardest to judge inside a quarter.
- Price is going the wrong way into that. Average premium per policy has fallen. Policies in force are growing because GEICO is buying the growth.
- The expense ratio is a choice, not a cost shock — 2.5 points higher on commissions and advertising. Spending up on growth at the moment unit economics deteriorate is the decision worth questioning.
- There is no earned-premium tailwind left. Premiums written turned negative in the quarter, down 0.3%, while earned premium still rose on the earn-out lag. Written below earned means the lag stops helping in the fourth quarter and works against GEICO through 2027, with the loss ratio already at 75.7%.
Reinsurance looks better than it is. Pre-tax earnings of $1.9 B against $1.2 B, and a property/casualty combined ratio of 82.9% against 86.0% — both flattered, and in three distinct ways.
- All of the premium growth is one contract. Written premium rose 2.6%, but a single whole-account quota share with Tokio Marine, which began on 1 April and takes a slice of its non-life book on risks attaching over ten years, contributed about $700 M. Without it, written premium shrank 1.7% and earned premium fell 3.6%. The 10-Q is explicit about why: price competition in US property. This is a shrinking book with one large new treaty laid on top, which is a different thing from a growing one.
- Reserve releases did the rest. $1.2 B of prior-year releases against $898 M, again from lower-than-expected property losses, again partly offset by increases in casualty estimates. Property cushion is being harvested while casualty quietly absorbs it — a finite trade.
- No catastrophes, against $760 M of wildfire losses last year. Normalise the releases and the catastrophes and the accident-year combined ratio is 90.9% against 86.9% — about 4 points worse, on less premium.
BNSF is ahead of its normalised figure. Revenue of $19.4 B in the nine months, up 11.5%, and $6.9 B in the quarter, up 14.7%; an operating ratio of 64.9% against 65.6%, and 63.9% in the quarter; pre-tax profit of $6.1 B against $5.3 B, up 15.5%. Volumes 4.2% higher with coal down 6.3%.
Two things about those numbers. The operating ratio understates the progress: fuel is largely passed through in surcharges, so a quarter of rising fuel prices inflates revenue and cost together and compresses the ratio's improvement to almost nothing. The incremental margin is the better measure, and it was 41% over the nine months. And the source of it is visible in one comparison: volumes up 4.2% while compensation rose 0.6%, wage inflation substantially offset by productivity. Carloads per dollar of pay are up 3.6%. That is an operating improvement, not a freight cycle, and it is the reason this is the one normalised profit raised this quarter — $5.5 billion to $6.0 B — deliberately short of the $6.2 B the nine months imply, because a railroad's operating ratio is cyclical as well as structural. At 20× that is $10 billion of value, about $4.70 a share.
Berkshire Hathaway Energy grew on data-centre load. Retail electricity volumes were 5.5% higher, the most under-discussed growth driver in the company. Wildfire liabilities have shrunk to $0.6 B unpaid from about $1.2 B. Electric utility margin was $7.1 B, up 8.2%, on volume growth across all three utilities — the load is arriving faster than the rate cases that will eventually earn on it. The normalised profit is unchanged at $4.0 B: load growth this fast is only a year old, regulated returns on it lag by design, and a through-cycle figure should wait for a rate-case cycle to confirm it rather than capitalise the first good year.
Manufacturing, Service and Retail earned $11.7 B against $10.3 B. Revenue growth is not the thing to look at here — OxyChem was not owned a year ago, Taylor Morrison was not owned a quarter ago, and Pilot's revenue line is a fuel price. Margin is the thing to look at, and it divides the segment cleanly.
| Nine months to September 30, 2026 | Revenue | Pre-tax | Margin | Prior margin |
|---|---|---|---|---|
| Industrial products | $35.5 B | $6.8 B | 19.3% | 18.7% |
| Building products | $21.3 B | $3.0 B | 14.2% | 15.3% |
| Consumer products | $10.3 B | $1.2 B | 11.2% | 11.4% |
| Service | $20.4 B | $2.5 B | 12.3% | 12.1% |
| Pilot | $42.3 B | $0.4 B | 1.0% | 0.9% |
| McLane | $36.4 B | $0.5 B | 1.2% | 1.4% |
| Retailing | $14.5 B | $1.0 B | 6.8% | 6.7% |
Industrial products is carrying the segment — pre-tax earnings up 30.1% with margin up 0.6 points. Within it, Precision Castparts is the real asset: revenue up 13.9% and pre-tax earnings up 32.2% on aerospace demand and on industrial gas turbines for power generation — the electricity build-out arriving in Berkshire's income statement, in a franchise the market still models as an aero-cycle play.
Two of the best-looking numbers are peaks. IMC's pre-tax earnings rose 47.1%, but on raw-material cost pass-through and customers pulling purchases forward, and management states plainly that rising raw-material prices will reduce IMC's earnings over the rest of 2026 and into 2027. That is a rare explicit guide-down and it should be taken at face value. Marmon's 6.8% is mostly gains on business and property disposals, an internal transfer of Acme Brick into the group, and tariff refunds — almost none of it operating.
Margin is falling where the consumer is. Building products' margin fell 1.1 points despite the revenue added by Taylor Morrison, because Clayton's homebuilding margins dropped and factory-built volumes fell; the rest of the group is only positive on refunds of 2025 trade tariffs. Consumer products' apparent collapse in the quarter is an artefact — Duracell booked three years of manufacturing tax credits in a single quarter of 2025 — and underneath it the group's earnings grew modestly. McLane's margin, already 1.2%, fell again on thinner distribution margins and retail customer losses: structurally eroding, not cyclically weak.
Pilot's 57% needs two adjustments in opposite directions. It carries a $116 M charge for a disposal that closed in November, so the quarter understates the business; and its nine-month margin was depressed by derivative losses on rising fuel prices while the matching gain on physical inventory is deferred until the fuel is sold. If prices hold, that gain lands in a later quarter. The segment's normalised profit is raised by exactly the two acquisitions and nothing else — $0.7 billion for OxyChem, $0.5 billion for Taylor Morrison, taking it from $13.6 billion to $14.8 B. Neither is annualised from a stub period, and neither is double counted: both were paid for in cash that has already left the balance sheet this valuation marks to.
Valuation
| Year-end 2025 | 30 Sep 2026 | |
|---|---|---|
| Downside | $458 | $486 |
| Base | $495 | $524 |
| Upside | $525 | $554 |
| Book value per B share | $333 | $358 |
| Base case, × book | 1.49× | 1.46× |
The sum of the parts is now a quarterly exercise with one rule: the balance-sheet marks and the share count roll every quarter because they are facts; the normalised profits and the multiples stay anchored to the 10-K because they are estimates of through-cycle earnings power. The old estimate is kept in full as published, with every vintage side by side in the valuation archive.
At $571 the price sits 8.9% above the base case and 3.0% above the upside case, on 1.60× book value against a twenty-six-year median of 1.50×. Whether that matters is the question this whole site exists to answer.