Chubb

I have written forty-seven of these letters and this is the first about an insurance company, which means I have spent a year and a half writing about business without explaining the idea that made Warren Buffett rich. Let me correct that first, because it takes two minutes and it is worth more than anything else on this page.

You pay your insurer a premium in January. If your house burns down in March, they pay you in March. If it burns in nine years, they pay you in nine years. If it never burns, they never pay you at all. In the meantime they are holding your money, and they are investing it, and every penny of the investment return belongs to them.

Multiply that across every policy in force and you have what is called float — a very large pool of other people's money, sitting on the insurer's balance sheet, working for the insurer's account. Chubb invests one hundred and seventy-five billion dollars. Its shareholders put in seventy-four billion and it borrowed seventeen. The remaining eighty-four billion or so is float.

Now the only question that matters: what does that eighty-four billion cost?

The answer is in a single number that every insurance investor should know and most people have never heard of. The combined ratio is claims plus expenses divided by premiums. At a hundred per cent the insurer breaks even on the underwriting — which means it has borrowed eighty-four billion dollars at nought per cent interest, and that alone would be a marvellous business. Above a hundred, the float costs something. Below a hundred, something rather wonderful happens.

Chubb's combined ratio last quarter was eighty-three point eight per cent. The average American insurer's was ninety-two point nine.

Chubb keeps sixteen cents of every premium dollar as pure underwriting profit — and it gets to invest the float on top. It is not borrowing at zero. It is being paid to hold other people's money, and it is being paid nine points better than its own industry for the privilege. Last quarter that float produced a record one point eight eight billion dollars of investment income. That is roughly seven and a half billion a year, on money Chubb was paid to take.

You can see the same thing in the cash flow statement, which is where I would send anyone who wants to check my arithmetic. Chubb earned ten point three billion dollars in 2025. Fourteen and a half billion actually came in the door. The four-billion gap is premiums collected against claims not yet paid — the float, growing.

Before I go on I must warn you about the screen, because Chubb's is a museum of nonsense. It will tell you the bankruptcy score is 1.08, which is nominally distress; that formula was built in 1968 on sixty-six manufacturers and it penalises exactly the balance sheet an insurer is supposed to have. It will tell you the return on invested capital is three and a half per cent, which is the worst of the four, because it treats the float as capital Chubb had to raise when in truth it is capital Chubb is paid to hold. The right figure is return on equity, and it is fifteen point two. And a discounted cash flow model in our own data gives seven hundred and twenty-four dollars against a share price of three hundred and forty. I have now watched that model produce plus two hundred and forty-five per cent for one property company, minus fifty-eight for another, minus sixty-nine for Deere and plus a hundred and thirteen here. When a method's answers span three hundred points, the method is the finding.

Two measures survive, and they are all you need: price to book, and the combined ratio.

Now let me tell you the thing I like most about this company, which most people report as bad news.

Its largest book of business is shrinking. North America Commercial premiums fell two point three per cent last quarter.

Here is why that is the best sentence in the results. Any insurer on earth can grow as fast as it likes. All it has to do is charge too little. The premium arrives this year; the claim arrives in five. A management team wanting a good half-decade can simply underprice, book the revenue, take the bonus and leave the reserve strengthening to a successor. The industry has done precisely this, in cycles, for two hundred years. Which means revenue growth in insurance is nearly worthless as a signal, and revenue decline is often excellent news.

Look at what Chubb did with the space. It declined large-account business at the prices on offer — and grew middle market and small commercial by eight point nine per cent, where pricing still works. Consolidated premiums grew a pedestrian three point six per cent. And underwriting income rose eighteen point eight per cent. Less business, considerably more money from it. That is a company pricing rather than chasing, and after twenty-two years of Evan Greenberg it is what I would expect.

The long record backs it up. Earnings per share have gone from eight dollars eighty-seven to twenty-five seventy-three in nine years, while the share count fell fifteen per cent. Book value per share rose twelve point three per cent in the last year alone. And note the two bad years in that table — seven seventy-nine in 2020, twelve thirty-nine in 2022 — because they are the most instructive part of it. Catastrophes arrive when they arrive, and anybody valuing an insurer on one year's earnings is making a mistake in one direction or the other.

So: what stops me?

Two things, and the first is not in any ratio. I looked for a chief executive succession plan and could not find one. The July announcements — Wixtead as executive chairman, O'Donnell as president — concern the reinsurance unit, not the parent. The entire case for Chubb rests on a cultural property: the willingness to write less business when the price is wrong. That is maintained by people, and one person in particular, who has been doing it since 2004. Twenty-two years in, with nothing published, that is a real risk carried by anyone who owns this. It is why I stopped short of scoring the management a ten, and it is the thing I would most like to see answered.

The second is simply the price, and I am afraid it is decisive.

The shares are one point seven four times book value, against a ten-year median nearer one point three six — a premium of roughly twenty-eight per cent to Chubb's own history. And because thirty-nine billion dollars of the balance sheet is goodwill and intangibles from the ACE combination and the build-out in China, on tangible book you are paying two point five eight times. Both numbers are true. I would not ignore the second.

The market has, in other words, worked out how good this is. The consensus target is five per cent away — almost exactly the rate at which the same analysts expect earnings to grow next year. Twenty-two of forty-three rate it a buy and nobody is arguing about the price.

Here is how I would put it to you.

This is one of the finest businesses I have examined this year. The float is free — better than free — and the man collecting it has spent two decades proving he will walk away from bad business, which in insurance is the only virtue that ultimately matters. Book value compounds at twelve per cent and the shares yield a token one point two, which is entirely correct: I would far rather Chubb kept the money.

But at one point seven four times book, my expected return is roughly the twelve or thirteen per cent that book value grows, minus whatever the multiple gives back — and a multiple twenty-eight per cent above its own decade-long average is more likely to give back than to expand. That is a perfectly respectable outcome and it is not a bargain, and I have learned not to pretend the two are the same thing.

So I would own this and I would be patient about the price. If you hold it already, I see nothing here that would make me sell — you own an exceptional underwriter compounding book value at twelve per cent, and selling that because the multiple is full is how people end up buying it back higher. If you do not, I would want it nearer three hundred dollars — about one and a half times book, roughly ten and a half times next year's earnings, and comfortably above the year's low. That does not require a disaster. It requires the shares to stand still for a while as book value keeps compounding underneath them, which is the most ordinary thing in the world and happens all the time.

Insurance rewards patience more reliably than almost any business I know, because the mistakes take five years to surface and the discipline takes five years to pay. Chubb has spent twenty-two years being paid to hold other people's money. It can spend another six months being fairly priced while I wait.