Geico's Moat Isn't the Gecko. It's Float That Costs Less Than Nothing.

In 1951 a twenty-year-old took a Saturday train to Washington to see the offices of a small auto insurer, found the building locked, and was let in by a janitor who pointed him toward an executive working alone upstairs. The young man was Warren Buffett. The company was Geico. He would spend the next four decades circling it before finally buying the whole thing - the merger that made Geico a wholly-owned Berkshire subsidiary closed on January 2, 1996.¹ What Buffett saw that Saturday wasn't a brand. There was no gecko, no fifteen-minutes jingle, barely a sales force. There was a cost structure - and the cost structure was the company.

The official story is that Geico wins because it's cheap. That's true, but it's the symptom, not the moat. A competitor can always cut prices for a quarter. What it can't easily copy is the machine underneath that lets Geico cut prices forever and still make money on the policy.

The moat was poured in 1936, not the 1990s

A persistent brand legend says Geico's direct-to-consumer model was a clever 1990s pivot. It wasn't. Bypassing the agent was the founding premise. When Leo Goodwin and a banker named Cleaves Rhea chartered the company in San Antonio in 1936 - it moved to Washington a year later to sit near the country's densest pool of federal employees - the whole idea was to sell auto insurance straight to careful drivers by mail and phone, with no commissioned middleman in between.² That single decision is the headwater of the entire moat. Every traditional insurer pays an agent to acquire and renew a customer. Geico simply doesn't. The savings don't go into a marketing slide; they go into the price.

Strip the structure down and you can see why this compounds. An insurer makes money two ways: underwriting (premiums collected minus claims and expenses paid) and investing the cash it holds between the two. The no-agent model attacks the expense side. Berkshire ownership attacks the investing side. Together they form a flywheel, and the flywheel runs on its own thinness - the lower the cost to write a policy, the lower the price Geico can offer, the more policies it writes, the more float it generates to invest.

Float is borrowed money that pays you to hold it

Here is the part that makes the second engine special. When you pay Geico a premium, the claim might not come for months or years - or ever. In the meantime Geico holds your money and invests it. That held pool is called float, and across all of Berkshire's insurance units it stood at roughly $169 billion at the end of 2023.⁴ Geico is one major contributor to that pool, not the whole of it - and that distinction matters, because the moat isn't that Geico has float. Plenty of insurers have float. The moat is the cost of it. When an insurer underwrites at a profit, the float is better than free: someone is paying you to hold their money until you might have to give it back. Berkshire's permanent capital removes the one thing that usually forces insurers into bad bets - the quarterly pressure to grow premiums even when the price is wrong. Geico can walk away from underpriced business. Most can't afford to.

Then Progressive found the crack

A cost moat tells you how cheaply you can write a policy. It says nothing about whether you priced that policy correctly. And pricing - knowing which drivers will actually crash - is where Geico quietly fell behind. The tool is telematics: programs that watch how you really drive, then price the risk you really pose rather than the risk your demographics imply. Progressive's Snapshot has been collecting that data for over a decade, through both an app and a plug-in device, long enough to build an actuarial picture of behavior no feature can shortcut. Geico's equivalent, DriveEasy, didn't launch until 2019, and it's app-only.⁸ On paper Geico looks ahead - DriveEasy tracks eight behavioral factors to Snapshot's five.⁸ But the asset was never the number of factors. It's the years of accumulated data behind them.

The cost of that gap is brutally specific, and Berkshire said it out loud. At the 2022 annual meeting, Ajit Jain laid the two companies side by side: Geico holds about a 7-point expense-ratio advantage over Progressive - that's the no-agent moat, intact - but Progressive holds about a 12-point loss-ratio advantage over Geico.⁷ Read that again. Geico is cheaper to run, and still loses, because Progressive pays out so much less on claims relative to premiums. Better pricing beat a better cost structure. Jain noted Progressive had been on the telematics bandwagon for more than a decade while Geico had begun serious effort only in the prior two years.⁷

The scoreboard followed the logic. In 2022 Progressive wrote $38.93 billion in direct premiums to Geico's $38.12 billion, taking the No. 2 spot in U.S. auto insurance.³ Berkshire's letters and a generation of press still reflexively call Geico the runner-up. It isn't. By 2024-2025 Geico sits firmly at No. 3, around 11.5% to 12.6% share behind State Farm and Progressive, even while writing more than $42 billion in premiums and leading the market outright in nine states plus D.C.⁶

So is the moat actually broken?

The fair objection cuts the other way: a company that just earned an estimated $7.8 billion underwriting on 18 million policies clearly still has a powerful moat.⁵ Both things are true. The cost-structure flywheel - no agents, permanent capital, near-free float - is intact and remains genuinely hard to copy; that's why Geico still prints underwriting profit at a scale most insurers can only envy. What narrowed is not the moat's depth but its width. A cost advantage protects you against rivals who must charge more. It does not protect you against a rival who charges less for the same driver because it knows that driver is safer. Progressive didn't out-cheap Geico. It out-priced it. And the painful part is that closing the gap means waiting for time to pass: telematics data depth can't be bought, only accumulated, which is why Jain's 'catch up in a year or two' has not been publicly confirmed as achieved. The moat is real. It's just thinner than the gecko ever let on.

Buffett understood the cost structure in 1951 and bought the whole machine in 1996. The machine still works. What changed is that the game moved from who can write a policy most cheaply to who can price a driver most precisely - and for a decade Geico fielded a world-class answer to the wrong question. The gecko was never the moat. Neither, it turns out, was being cheap. The real moat was the cost structure that let Geico be cheap and profitable at once. It still holds. It just no longer wins on its own - because the most valuable thing in insurance was never the lowest price. It was knowing, before the crash, who was going to crash.