Chubb invests $175.4 billion. Its shareholders put in $73.8 billion and it borrowed $17.6 billion. The other $84 billion belongs to somebody else.
That $84 billion is float — and the company is paid to hold it. Not lent it cheaply. Paid.
This is the single idea Berkshire Hathaway was built on, it is explained badly almost everywhere, and once you see it you cannot look at an insurer the same way again.
Money you owe, that never leaves
An insurer collects a premium today and pays the claim years later. In between it holds the money and invests it. Every individual dollar in that pool is owed to someone — but as old claims are paid, new premiums arrive, and the pool itself never empties.
So the insurer has permanent capital that does not appear as equity and does not appear as debt. It sits in the liabilities as reserves, and it behaves like money the company owns.
The only question that matters is what it costs.
| Funding source | What it costs | Who decides |
|---|---|---|
| Equity | The return shareholders demand | The market |
| Debt | A stated interest rate | The lender |
| Float at a 100 combined ratio | Nothing — free money | The underwriter |
| Float at Chubb's 83.8 | Minus 16.2% — you are paid to hold it | The underwriter |
| Float at the US industry's 92.9 | Minus 7.1% | The underwriter |
Everyone else in finance pays for their money. A good underwriter is paid to take it, and then keeps what it earns on it.
What that actually does over time
Chubb's earnings per share went from $8.87 to $25.73 in nine years while the share count fell 15%. That is 2.9 times, or about 12.6% a year compounded — from an insurance company, in an industry generally described as mature.
The engine is not clever investing. The portfolio is mostly bonds. The engine is that roughly half the portfolio is money the company was paid to accept, and any return at all on borrowed-at-negative-cost money compounds into the shareholders' half.
Book value per share compounded 12.3% in the last year alone.
Float outside insurance, where nobody calls it that
Paychex runs payroll. Between collecting payroll taxes from its clients and remitting them to the government, it holds about $5.8 billion a day — and keeps the interest.
Look at what that is. It is float, with one improvement: the clients pay a fee for the service on top. In insurance the underwriter has to earn the negative cost by underwriting well. Here the depositor is charged for the privilege of depositing.
And the market currently hates it. Our report found $100 invested five years ago worth $111, against $194 in the index, with the peer group down 38% in a single year — the whole "boring compounder with a compliance moat" category repriced for artificial intelligence.
Client retention identical to last year. Revenue per client rising. Price realisation named by management as a growth driver. Margins expanding.
The disruption is visible entirely in the multiple and not at all in the results. That is not proof it will not arrive — it is a statement about what has happened so far, which is the only evidence anyone actually has.
Neither of these was cheap when we looked. Chubb traded at 1.74 times book against a ten-year median nearer 1.36 — the market has worked out how good it is. Paychex had rallied about 40% off its low, above every analyst price target on the street.
A wonderful funding structure is a reason to want a business. It is never on its own a reason to pay any price for it.
How to find float, and what breaks it
Any business paid before it delivers holds somebody else's money in the meantime — insurers, payroll processors, subscription businesses with annual billing, deposit-taking institutions. The balance sheet shows it as a liability. The cash is real and it is invested.
For an insurer that is the combined ratio: below 100 the float is free or better, above 100 it is expensive. For everyone else, ask what the company gives up to get the money — a discount for annual billing is a cost of float, stated as a discount.
Float is permanent only while the business is growing or steady. A shrinking insurer is one paying out more than it collects, which means selling investments into whatever market exists that year. The pool that never empties does empty, if the underwriting stops.
An underwriter paid 16 cents a year to hold $84 billion, and a payroll processor whose depositors pay it a fee.
Combined ratios, the float arithmetic and the valuation, worked through.
Frequently asked
What is insurance float?
Money an insurer has collected in premiums but not yet paid out in claims. Because claims can arrive years after the premium, the insurer holds a large pool in the meantime and invests it. Chubb invests $175.4 billion of which roughly $84 billion is neither shareholders' money nor borrowed — it is float, and it is there permanently even though every individual dollar of it is owed to someone.
Why is float better than borrowing money?
Because of its cost. Debt has an interest rate you pay. Float has a cost set by underwriting: if an insurer pays out 83.8 cents for every premium dollar collected, as Chubb did, it is not borrowing at zero — it is being paid about 16.2 cents a year for every dollar it holds. Negative-cost funding of that size is the rarest advantage in finance.
What is a combined ratio?
The share of each premium dollar that goes out again in claims and expenses. Below 100 means underwriting itself is profitable and the float is free or better; above 100 means the insurer is paying for the privilege of holding the money. Chubb's 83.8 against a US industry average of 92.9 is the entire reason the business compounds.
Do non-insurance companies have float?
Some do, and they are rarely recognised for it. Paychex holds about $5.8 billion a day of client payroll taxes between collecting them and remitting them, keeps the interest, and charges the client a fee on top — float where the depositor pays you to hold it. Any business paid before it delivers has a version of the same thing.