Fairfax Financial Holdings Limited FFH.TO

Key data

The business

Fairfax is two businesses bolted together. The first is property and casualty insurance sold worldwide through subsidiaries most people never hear of: Odyssey in reinsurance, Northbridge in Canada, Allied World and Crum & Forster in the United States, plus large operations in India and other emerging markets. They write coverage on everything from a factory's fire risk to a shipping fleet to a farmer's crop, collect the premium up front, and pay the claims years later. The second business is what Watsa does with the money that piles up in between: a large investment portfolio he runs with a fund manager's freedom, concentrated in a way no ordinary insurer would tolerate. The premiums float the investing, and the investing is where the outsized returns come from. The moat is not the insurance, which is a commodity, but the permanent capital and the culture that let Watsa hold unpopular positions for years and buy when others are forced to sell. He also owns whole companies outright, from a pet-insurance arm to Indian financial firms, sitting on the balance sheet next to the bonds.

The numbers

Fairfax reports in US dollars while the stock trades in Canadian dollars, so the multiples above convert the two; the tables below are the reported US-dollar figures.

The June quarter's revenue jumped 42% from a year earlier, but that top line is mostly investment gains flowing through, not more insurance sold: net income of $1.36B was actually down about 6%. The story is Q1 2026, the soft one at $0.70B, less than half the quarter before it, when the investment marks turned against them. Underwriting stayed profitable throughout; the volatility lives entirely in the investment column, which is the design, not a defect. The individual combined ratios by subsidiary were not in the pulled data.

2022's depressed per-share number is the accounting-standard changeover, not a bad year; ignore it as a base. What matters is the compounding underneath. Over the three years to 2025 revenue grew about 16% a year while the diluted share count fell from 25.3 million to 22.3 million, a 12% reduction. Return on equity ran about 17% over the trailing year, ahead of the roughly 15% growth in book value per share Watsa has aimed at for decades. At US$1,511 of book value per share and 1.3 times that in price, the market is finally paying up for a record it doubted for ten years. What this memo believes that the market may not: 1.3x book is fair if book keeps compounding in the mid-teens, but it is the first thing to fall if the investment gains that drove the re-rating revert; the print that settles it is a single quarter where underwriting slips to a loss and the investment book cannot cover it.

Management

Watsa controls Fairfax through multiple-voting shares; there were no open-market insider buys or sells in the past year, so the signal is in what the company did with its cash, not its officers with theirs. It repurchased $2.55B of stock in 2025 at an average of about $2,259, after $1.83B in 2024 near $1,596, the two largest buyback years in its history, at the two highest prices, near the current 1.3x book. That is a long way from the below-book repurchases of 2021 to 2023 that created so much value. Buying back stock as the multiple expands is far less accretive than buying it when the market hated the name, and it is the one place the capital-allocation record looks less disciplined than the reputation. Executive pay disclosure was not in the pulled data. The dividend is a token US$15 a share (CAD$21), under 1% of the price; nobody owns this for it.

How it fails or surprises you

Reserves set too light. Insurance profit is an estimate until claims actually arrive years later. If the reserves booked during the recent hard market prove too thin, the charge lands as adverse development three or four years out and hits book value directly. The line to watch is reserve development in the annual report; a swing from favorable to adverse is the honesty test of any insurer, and the first data past this run does not show it yet.

The investment book reverts. The re-rating to 1.3x book rests on a run of strong investment gains. Q1 2026's $0.70B quarter, less than half the one before, is the fact this bullish read explains least well: it shows how fast the marks turn. A year of flat or negative investment returns would stall book-value growth and drag the multiple back toward the 1x it carried for most of the last decade.

The rate and emerging-market tailwind (right tail). Fairfax's large Indian and emerging-market holdings, plus a bond portfolio now earning a real yield after the zero-rate years, could compound book value nearer 18–20% than 15% for several years. The market is paying for 15%. The prior watch-item, the fixed-income yield disclosure, is exactly this signal, and the Q3 quarter meant to test it is not yet reported; through June the multiple held at 1.3x and underwriting stayed in the black.

Closing thoughts

Watch two things and you will know whether the price is earned: whether the investment book keeps producing gains, and whether the insurance keeps paying out less than it takes in. Together they settle it. If underwriting holds but investments go flat, you are left holding a good insurer at a full price, and only the years of compounding decide the outcome. The fatter tail points up, because the bond portfolio now earns a yield that does not depend on Watsa picking the next winner. The real danger is the left tail: a reserve charge landing in the same year the investment book falls, which is what would take book value backward.

The bet is still that Watsa grows what each Fairfax share is worth faster than the stock has run, out of the premiums and the tens of billions he invests. It breaks if the insurance starts paying out more than it collects and the investment gains can no longer cover the gap. The one pair of numbers that tells you first: the combined ratio crossing 100 and a quarter of negative investment returns arriving together. Until those two move the wrong way at once, the compounding is intact.