Florida homeowners insurer HCI Group: 18% true FCF yield, $4B reinsurance program, legally mandated customer pipeline from Citizens Insurance. I also pulled 15 years of data instead of just the good 3. The honest normalized yield is closer to 12%. Here's why I still own it.
The easy version of this story: HCI Group is a Tampa-based Florida homeowners insurer trading at 8x P/E, 18% true FCF yield, with a CEO who's been buying its newly IPO'd tech subsidiary Exzeo in the open market every two weeks through a real price decline. Zero insider sells in eighteen months at either company. $4.06 billion in reinsurance coverage with a $163 million maximum first-event retention against $1 billion-plus of tangible book value. Legally mandated customer pipeline from Citizens Insurance that handed HCI 60,820 policies and $216.7 million in annualized premium in 2025 alone.
The harder version: I went back to 2011 and ran the actual numbers. For ten straight years this was a 2-4% FCF yield company. The 18% yield is a three-year phenomenon driven by Florida's 2022-2023 tort reform and HCI's own underwriting technology maturing. Both are real and structural. But three good years is not fifteen good years and the honest thing is to say so.
There's also a reserve release embedded in 2025. The loss ratio fell from 55.3% to 29.4% partly because of favorable prior-year reserve development, not all structural improvement. Strip that out and the normalized yield is closer to 12%. Still compelling against a 4.3% treasury rate, but not the same story as 18%.
One more detail that surprised me: only $14 million actually flowed from HCI's insurance subsidiaries up to the parent holding company in 2025. Florida law caps subsidiary dividends. The consolidated $430M true FCF number is real but a meaningful chunk of it sits inside regulated subsidiaries that can't freely fund buybacks without clearing a statutory gate. Some of the current buyback is likely funded by the Exzeo IPO's $154.8 million in net proceeds, a one-time capital event.
The Citizens pipeline is the part that genuinely surprised me. Florida's "20% rule" means if a private insurer's quote comes within 20% of Citizens, the policyholder has to leave Citizens with no opt-out. This isn't politicians being smart. It's Tallahassee structurally admitting it can't out-underwrite a private company and building a rule that hands risk selection to whoever has better pricing software. HCI built that software specifically to cherry-pick the better risks before a competitor does. I don't love betting on politicians being smart. I'm comfortable betting on smart software beating policy makers.
The reinsurance structure is verified, not summarized. $4.06 billion total coverage up 16% year over year. Maximum first-event consolidated retention of $162.6 million. Every dollar beyond that goes to reinsurers and the Florida Hurricane Catastrophe Fund. Against $429.8 million true FCF and $1 billion-plus tangible book, a $163 million max loss is an earnings event not a solvency event. One honest asterisk: that retention is up about 4% from last year. HCI is taking slightly more first-event risk, not less.
What happened in the two actual storm years. Hurricane Irma 2017: net income minus $6.9 million. Hurricane Ian 2022: net income minus $55 million, true FCF minus $21.5 million. HCI recovered operationally within a quarter both times. But the one time this specific improved version gets tested by a real storm hasn't happened yet. That's the honest uncertainty.
I sized it at 350 shares because it's a genuine, deliberate bet on a real undiversifiable tail risk. Not a core holding. Not screaming. A well-verified position in something the market is correctly pricing with a discount for something hard to model.
Full piece with fifteen years of data, the Exzeo insider buying pattern, and the Citizens pipeline mechanics in comments. [https://cavemanscreener.substack.com/p/hci-rock-you-like-a-hurricane](https://cavemanscreener.substack.com/p/hci-rock-you-like-a-hurricane)
