Berkshire Hathaway Inc.Figures

Berkshire is first of all an insurer. An insurer collects premiums today and pays claims later; in between it holds money that is not quite its own yet, and it invests that money. The money has a name, float, and it stood at USD 177.5bn on 30 June 2026, against USD 88bn at the end of 2015. Over the two half-years compared in the latest quarterly filing its cost is negative: the group is paid to hold it. That float funds three things. Seventy-three named operating companies in the annual report, from BNSF with its 34,880 employees to Central States Indemnity with nineteen. A portfolio of listed securities, USD 263,095.7m filed with the US regulator at 31 March 2026 across twenty-six issuers. And USD 359.2bn of cash and US Treasury bills at 30 June 2026, below which the group forbids itself to fall through buybacks, with a floor set at USD 30bn.

At Berkshire the doctrine is written down. Thirteen principles set out in 1983, republished in 1999 in a document called An Owner's Manual, never withdrawn since. The most important one for understanding the house is stated negatively, and therefore testably: regardless of price, there is no interest at all in selling a good business already owned. The text even offers the opposite image, that of the card player discarding his least promising card each turn, to say that this is not the house's way. One consequence follows directly: no average holding period is published, and none can be, since an average presupposes exits. The only quantified marker is an accounting one: the consumer products pocket carries USD 100,486m of value for USD 8,650m of cost, or 11.6 times, recalculated. Debt follows the same rule of thrift. Excluding borrowings of BHE and BNSF, two regulated entities that finance their own networks, group borrowings amount to USD 43.3bn, covered 8.3 times by cash, recalculated. No dividend has been declared since 1967, but that is not an absolute rule: it is a test, reviewed each year by the board, which holds as long as one dollar retained is reasonably likely to create more than one dollar of value for shareholders.

Berkshire is the only house in this universe that refuses to publish a reference value. The phrase "book value" returns zero occurrences in the regulatory annual filing for fiscal 2025, while the word "Berkshire" appears there 882 times: the measurement works, the absence is real. The refusal is old and argued: two people looking at the same facts will almost inevitably reach slightly different estimates of intrinsic value, and the house prefers to supply the facts rather than its own figure. The consequence for a reader is absolute: at Berkshire, neither a discount nor a premium can be computed. What is published instead is abundant and verifiable, line by line.

That leaves decentralisation, and one figure makes the claim testable: twenty-eight people at head office for 387,815 employees across the group, or 0.0072% of the workforce, recalculated. Both numbers are published side by side. Since 1 January 2026 the house has been run by Gregory E. Abel, who has set out five capital allocation criteria: thoroughly understand the business, partner with high-integrity leaders who act like owners, avoid anything that could damage the group's reputation, act quickly and concentrate capital on few ideas, then let compounding do its work. The stated horizon is not ten years, it is perpetuity. One change has already shown: no share buybacks in 2025, USD 4.8bn in the first half of 2026, most of it in the second quarter.