Berkshire's Best Investor Is a Pile of Other People's Money It Will One Day Have to Repay

In March 1967, for $8.6 million, Warren Buffett bought a small, unglamorous insurer called National Indemnity Company and its sister, National Fire & Marine.⁵ The world remembers Berkshire Hathaway for what came after — the stock picks, the wholly owned railroads, the candy company held for half a century. But the thing Buffett actually acquired that day was not an insurance business. It was an account he could borrow from, indefinitely, on terms no bank would ever offer. At the time, that account held $19.4 million.⁶ By the end of 2024 it held about $171 billion.¹

The popular story is that Berkshire is a great investor that happens to own some insurance companies. The order is backwards. Berkshire is a machine for accumulating other people's money cheaply, and the investing is what it does with the money while it waits to give it back. Buffett has a name for that money. He calls it float.

Money they hold but do not own

An insurer collects premiums today and pays claims later — sometimes years later, sometimes decades, in the case of long-tail liabilities. In the gap between collecting and paying sits a pool of cash that legally belongs to policyholders but is physically in the insurer's hands. Buffett describes it plainly as 'money we hold but don't own.'⁴ Most insurers park that money in safe bonds and treat it as a regulatory necessity. Berkshire treats it as investable capital — the down payment on everything else it owns. That is the cross-subsidy hiding in plain sight: a low-margin, occasionally money-losing insurance operation manufactures the cheap leverage that funds a fabulously profitable investment operation. One arm of the company quietly bankrolls the other.

The mechanism is mundane and that is exactly why it is powerful. Borrow money from a bank and you pay interest and eventually repay the principal. Raise equity and you dilute the owners. Float does neither, as long as one condition holds: that the claims, over time, cost less than the premiums you took in to cover them. When that happens, the insurer earns an underwriting profit — and the float isn't borrowed at zero cost. It's borrowed at a negative cost. Berkshire gets paid to hold the very capital it then invests.

Why a tollbooth-sized insurer became a continent of capital

Here is the part that makes float compound rather than merely exist. A normal loan gets repaid and disappears. Float behaves like a permanent, growing balance, because as old claims are paid, new premiums flow in to replace them — and as the business grows, the pool grows with it. Float climbed from roughly $129 billion at the end of 2019 to about $171 billion by the end of 2024.⁷ In effect, the principal never comes due all at once. It rolls forward, decade after decade, available to be invested the entire time. That is the trick a bank loan can't match: leverage that doesn't mature, doesn't charge interest in a good year, and grows on its own as long as the underwriting stays disciplined. Buffett's 2017 letter notes the cost of that float has run below zero in most years since 1967.⁶

The 'free money' story is wrong, and Buffett is the one who says so

The honest objection to all of this is that 'free money' is a fairy tale, and the fair version of that objection is correct. Float is only cheap when the underwriting is good, and the underwriting is not always good. Berkshire's insurance book has posted losing years — the 2017 hurricane season alone, with Harvey, Irma and Maria, drove net underwriting losses across its reinsurance operations. Even the strong 2024 result, a $9.02 billion underwriting gain across all insurance operations, came after absorbing roughly $1.2 billion in estimated claims from Hurricanes Helene and Milton.⁷ In a year when the catastrophes are large enough, the float stops being negative-cost capital and starts being expensive borrowed money — which is exactly the 'albatross' Buffett warned about. The float is not a windfall. It is the residue of a competitive, cyclical, sometimes brutal business that has to be run well, every year, or the magic reverses.

And the engine matters less than it used to

There is a second, subtler problem with treating float as the master key to Berkshire. For years, analysts valued the company largely through its float, and that lens has dulled. Morningstar's analysts have flagged that float-based valuation of Berkshire 'has become less relevant as Berkshire's operations continue to diversify away from its core insurance operations,' and that such models are 'extremely sensitive to relatively minor changes to input variables.'⁸ As the railroad, the energy business, and the operating companies grow, insurance becomes one engine among several rather than the engine. The float story is true history and a real mechanism — but as a present-day valuation anchor, it explains a shrinking share of what Berkshire is. The thing that built the company is no longer the thing that fully describes it.

Strip the legend down and what remains is unglamorous and exact: a man bought a sleepy insurer for $8.6 million, recognized that its dullest feature — the gap between collecting premiums and paying claims — was a borrowing facility the rest of the world had failed to value, and spent six decades feeding it.⁵⁶ The genius was never just picking stocks. It was funding the picks with money he held but did not own, at a cost that, in the good years, fell below nothing. The catch he states out loud is that those years have to keep being good. Float is the best loan ever written — and the borrower has to keep earning the right to roll it over.