Berkshire's insurance operations earned $1.72 billion of underwriting profit in Q1 2026, up 29%. Almost all of that increase is the absence of last year's California wildfire charges in reinsurance. GEICO's own underwriting profit fell 34.8%.

GEICO wrote 1.5% more premium in the quarter while Progressive grew policies in force 9%. Greg Abel described the trade at the Omaha meeting: "It's not going to be easy to just restart the growth engine."

Key Takeaways

  • GEICO's underwriting profit fell 34.8% to $1.416 billion from $2.173 billion, while the group total rose 29% because P&C reinsurance swung from $68 million to $637 million against a wildfire-charged base quarter.
  • The combined ratio moved 79.8% to 87.3%, a 7.5-point deterioration split between 4.9 points of loss ratio and 2.6 points of expense ratio.
  • Underwriting expenses rose $338 million against $168 million of additional written premium, which is the acquisition cost of defending share rather than a cost overrun.
  • GEICO grew policies in force 2% against Progressive's 9%, at combined ratios of 87.3% and 86.4% respectively.
  • The rebuild consolidates more than 600 legacy systems into 15 to 16 platforms, against a headcount already down from 42,156 in 2020 to 29,541 at the end of 2025.

The Group Number and the GEICO Number Are Different Stories

Berkshire's total operating earnings reached $11.35 billion, up 18% from $9.64 billion, with insurance contributing $1.72 billion of underwriting profit and $2.68 billion of investment income on the float.

The 29% increase in underwriting profit is mostly a base effect. Q1 2025 carried roughly $860 million of California wildfire catastrophe charges in reinsurance, so with no comparable event the P&C reinsurance group moved from $68 million to $637 million. BH Primary swung from a $144 million loss to a $476 million gain on a $660 million decline in losses and LAE, partly from favourable prior-year development.

GEICO went the other way. Underwriting profit of $1.416 billion is down 34.8% from $2.173 billion, on three visible movements: losses and loss adjustment expenses up $853 million on higher frequency and severity across private passenger auto property damage, collision and bodily injury; underwriting expenses up $338 million or 29.3%, which the 10-Q attributes to a steep increase in policy acquisition expense tied to advertising; and written premiums up only $168 million, or 1.5%.

The favourable frequency trend that ran through 2024 reversed in early 2026. Both halves of the loss ratio are now moving against the book at once.

The Expense Line Is the Filing Signal

An 87.3% combined ratio remains exceptional against an AM Best 2026 P&C industry composite of 96.9% and a personal auto average that typically runs between 98% and 102%. The direction and its composition are what carry information.

ComponentQ1 2026Q1 2025 (Approx.)Walk
Loss ratio73.9%~69.0%+4.9 points
Expense ratio13.4%~10.8%+2.6 points
Combined ratio87.3%~79.8%+7.5 points

The loss ratio movement is a market event: frequency and severity reverting after an unusually favourable run. The expense ratio movement is a decision. GEICO's expense ratio has historically been among the lowest in personal auto, the structural payoff of a direct model that carries no agent commission. A 2.6-point increase says that advantage is being spent.

The arithmetic is the point. Underwriting expense rose $338 million while written premium rose $168 million, so the additional acquisition spend bought roughly half its own cost in new premium during the quarter. That is defensible if the policies acquired retain and earn out, and it is exactly the question retention makes uncertain, because the shopping activity Abel described as unprecedented across the auto space cuts both ways for a carrier competing on brand and price rather than on individual risk-level segmentation.

Progressive's quarter is the comparison that makes it concrete: 9% policies-in-force growth against GEICO's 2%, at a 86.4% combined ratio against 87.3%, on written premium of $23.6 billion against $11.7 billion. Progressive is not paying more per policy to grow faster. The difference is in rating precision built on telematics accumulated since 2010, which shows up in filings as segmented rates that allow aggressive growth in preferred segments while holding margin.

The Repair Is Slower Than the Funding Source

The rebuild Abel is financing is real and it is infrastructure, not a model deployment.

GEICO historically ran more than 600 legacy systems that, in Ajit Jain's words, "don't really talk to each other." The programme consolidates those into 15 to 16 integrated platforms, roughly a 97% reduction in application count, alongside a shift from buying technology to building it, a Palo Alto engineering hub opened in October 2025, and cloud repatriation after finding that cloud bills rose 2.5 times over a decade while reliability declined. Headcount has already fallen from a 2020 peak of 42,156 to 29,541 at the end of 2025.

Abel's stated AI posture is deliberately narrow: human involvement in every recommendation, a repeatability safeguard ("if we ask it to ignore today's information and just focus on yesterday, and we get the same answer, then we've safeguarded the system"), and no deployment without demonstrated value. Jain drew the boundary explicitly, describing AI as a productivity tool for routine work and saying he does not think it will reach a point where you can make a tradeoff on pricing or settling a claim for many years.

That is a coherent governance stance. It is also a statement that the tools being deployed do not address the thing the growth gap is made of, which is accumulated behavioural data rather than model capability.

Meanwhile the funding source is contracting. GEICO's full-year 2025 underwriting profit was $6.824 billion on $45.2 billion of premium written; the Q1 2026 run rate of roughly $1.4 billion a quarter annualizes near $5.6 billion. That is still ample to fund technology investment in the hundreds of millions a year out of underwriting cash flow without competing for capital at the Berkshire level.

But the 79.8% that produced the 2025 figure was not a repeatable operating point, and an 87% to 90% range is closer to the long-run one. The rebuild has to close a data gap on a timeline of three to five years, using a budget drawn from a margin that is normalising over the same period, against a competitor whose advantage compounds with every mile its policyholders drive.

Further Reading

Sources