Berkshire Hathaway didn't purchase shares in GEICO until the mid to late 70s (he was already worth hundreds of millions) and didn't buy it outright until 1995 (when he was a multibillionaire).

Before purchasing GEICO, Buffett averaged a 40% annualized return in his investment partnerships over 13 years, then averaged a 25% annualized return in Berkshire Hathaway for two and half decades. He was already the greatest investor in history before buying GEICO, and since purchasing it Berkshire Hathaway returns have dropped significantly (though not because of GEICO). GEICO has had only a small part to do with Berkshire's success, in fact it's a small portion of their insurance companies, which include General Re.

Float can be negative. In fact, the vast majority of insurers carry negative float, they lose money on the policies they write and squeeze out profits only with the interest they earn on float. Buffett's use of float is only successful because Berkshire is incredibly disciplined and doesn't own any insurance business without demanding it produce an underwriting profit.

Lastly, GEICO's float is probably the least useful float of all their insurance companies. Float's value is in proportion to the insurance contract duration, auto insurance is a short term contract and the float has to be returned quickly, greatly restricting the types of investments that can be made with it. Ajit Jain's group writes insurance contracts that have durations up to decades long, and that float can be invested much more aggressively.

Even if Berkshire didn't produce an underwriting profit, it'd still be very successful because of the returns they make on their float. There's a line obviously you can't cross but even if they had a combined ratio of like 1.05 they'd be fine. I like their discipline though.

Berkshire has always had a huge insurance component. They bought NICO in the late 60s and built the business to a large part around that.