Buffett's Real Edge Isn't Picking Stocks. It's Getting Paid to Borrow.

In March 1967, Warren Buffett spent about $8.6 million to buy a sleepy Omaha insurer called National Indemnity.⁴ On paper it looked like a modest deal for a small, profitable underwriter. But sitting on that company's balance sheet was something far more valuable than the company itself: more than $19 million of reserves — premiums already collected against claims not yet paid — against a net worth of just $6.7 million.⁵ Buffett didn't pay $8.6 million for an insurance business. He paid it to get the keys to $19 million of other people's money he could invest until the claims came due. The fuel was already in the tank. He just had to learn never to let it run dry.

The official story is that Berkshire Hathaway is the greatest stock-picking machine ever built, and Buffett the greatest stock-picker who ever lived. That's the part everyone copies and nobody can replicate. The real machine is hiding one layer down — in the boring, unglamorous discipline of insurance underwriting that turns the picking into something almost incidental.

The money he holds but doesn't own

Buffett named the thing precisely, in language plain enough to put on a refrigerator. An insurer takes your premium today and pays your claim later — sometimes years later. In between, it holds a pile of money that belongs, eventually, to policyholders, but is the insurer's to invest right now. That pile is called float.

Read that last sentence twice, because it is the whole strategy. Most companies pay for capital. They borrow at interest, or they issue equity and dilute owners. Berkshire built a way to get capital that, in good years, costs less than nothing — because if the premiums it collects exceed the claims and expenses it pays, the float is not a loan with interest. It's a loan that pays Berkshire to hold it. That is the difference between a fund manager who borrows to invest and one who gets paid to. Same leverage. Opposite sign.

Why the fuel only stays free if the underwriting stays disciplined

Here is where most explanations stop, and where the actual moat begins. Float is not magic. It is only cost-free when underwriting breaks even or better — when the insurer collects more in premiums than it ultimately pays out. Underwrite carelessly to chase volume, and the claims overwhelm the premiums, and suddenly that giant pile of investable money carries a real, grinding interest cost. The float doesn't fund you; it bleeds you. Buffett knew this from the start, which is why the unglamorous core of Berkshire was always a refusal to write bad policies just to grow the pile.

And the record shows the discipline held. Across the twenty years through 2023, Berkshire ran an underwriting profit in 18 of them, for a cumulative pre-tax underwriting gain of $29.2 billion.² That is not a footnote. It means Berkshire was paid $29.2 billion over two decades for the privilege of holding tens of billions of investable dollars — and then it earned investment returns on that money on top. The picking gets all the headlines. The being-paid-to-borrow is what made the picking matter at scale.

Compounding a liability into an asset

Trace the number and the strategy becomes a slope. Float grew from roughly $16 million in 1967 to $62 billion by the end of 2009, by Buffett's own count.⁷ It reached $88 billion in 2015 and $176 billion by the close of 2025.¹ Each leg up was bought with the same currency: more underwriting, kept disciplined, generating more premiums that arrive before claims are paid. GEICO is the cleanest illustration. Berkshire's subsidiaries had quietly built a 33.2% stake in GEICO for $45.7 million by 1980 — and then, holding tight, watched that stake creep up to over 50% without buying another share, purely because GEICO bought back its own stock. In January 1996, Berkshire paid $2.33 billion to mop up the rest.³ A low-cost auto insurer is, structurally, a float-manufacturing plant. Owning all of it meant owning all of its fuel.

The honest counter: isn't this just leverage with better PR?

The fair objection is that float is simply leverage, and leverage cuts both ways — so calling it Buffett's genius is dressing up a balance-sheet trick in folklore. There is real force in this, and Buffett does not dodge it. He has stated flatly that cost-free float is not the industry's normal condition: 'In most years, premiums have been inadequate to cover claims plus expenses.'⁷ For most insurers, the float carries a real cost, and the leverage works against them. And Berkshire is not immune — its own letters record underwriting losses in 2017, a $3.2 billion pre-tax hit, and again in 2022.² In those years the float was not free. The leverage bit back.

But that is exactly the point that survives the objection, not the one that kills it. If cost-free float were normal, it would not be a moat — every insurer would have one. It is precisely because the industry as a whole fails to achieve it that Berkshire's two-decade record of doing so is the differentiator. The leverage is generic. The discipline that makes the leverage cheaper than free is not. You cannot acquire Berkshire's underwriting culture by buying an insurance company, any more than Coca-Cola's brand came in the bottle. The hard part is the patience to walk away from premium volume that would dilute the quality of the float — for fifty years, through every soft market that tempted everyone else to write garbage.

The machine is bigger, and the engine is quieter

The 2025 numbers carry a warning inside the triumph. Float reached a record $176 billion, up another $5 billion.¹ But underwriting profit fell to $7.26 billion from $9.02 billion the year before, and investment income slipped to $12.51 billion from $13.67 billion — even in a higher-rate world that should have helped.⁸ The stock of float keeps inching up while the flow it throws off compresses. That is the natural tension of a maturing machine: the larger the pile, the harder it is to keep growing it without loosening the very discipline that makes it valuable. The genius was never the size. It was the refusal to let size corrupt the quality.

Strip away the legend of the stock-picker and what remains is stranger and harder to copy: a man who spent half a century building a machine that pays him to borrow, and then deploys what he borrows into the things only he gets to buy with it. The float is not a financing detail in the Berkshire story. It is the story. Anyone can learn to pick stocks. Almost no one can build the kind of fuel supply that lets bad years cost nothing and good years cost less than nothing — because that supply is made of discipline, and discipline is the one input that doesn't show up on the balance sheet you're trying to buy.