I've been hosting the InsurTech Geek Podcast for seven years. In that time I've talked to reinsurers, cat modelers, satellite imaging companies, parametric pioneers, and state regulators. Climate came up in almost every conversation. And the thing I noticed, recording episode after episode, is that the conversation kept changing.
Not because the science changed. Because the money started talking.
The Number That Changed Everything
Back in November 2021, I had Nat Manning from Kettle on the show. Kettle is an AI-powered reinsurance company, and Nat dropped a stat I haven't forgotten: a 3x increase in billion-dollar catastrophes in the U.S. over the prior ten years. And that surge had driven a 60% drop in return on equity for the reinsurance industry.
Read that again. Sixty percent.
"There's been a 3x increase in billion-dollar catastrophes over the last ten years in The U.S. That's led to this 60% drop on return on equity for the reinsurance industry." — Nat Manning, Kettle, InsurTech Geek Podcast Episode 68
That's not an environmental policy debate. That's a balance sheet problem. And it was already baked in by 2021.
When the Model Stops Working
The way property insurance pricing works — or worked — is you take 100 years of weather data and you build actuarial tables. Frequency and severity, region by region. You price to those tables, you buy reinsurance to cover the tail, and you run the business.
The problem is the present is now outside the distribution those models were trained on. When your historical data says a certain level of flooding happens once every 50 years and it's now happening every seven, your price is wrong. Your reserves are wrong. Your reinsurance treaty may not cover what you thought it would.
The early podcast episodes, call it 2019 through 2021, were heavy on the physical science. Guests talked about why the models were breaking. What was shifting in the underlying climate data. It was a lot of "here's what the data shows."
By 2022 and 2023 the conversation had moved. Guests stopped explaining the problem and started talking about how to operate inside it. How do you actually underwrite coastal property right now? How do you build a portfolio that doesn't blow up in a bad hurricane season? That's when the tech companies started mattering in a real way — Kettle doing ML-based cat modeling, Cape Analytics pulling aerial imagery to score property risk at the parcel level, Betterview scoring roof condition from satellites so you know what you're writing before you write it.
Better data, better models. Not a solution to the underlying physics, but a way to price more accurately against them.
Then It Got Political
By 2024 and into 2025 the conversations on my podcast got harder. Not harder scientifically. Harder politically.
Because what happens when the math says rates need to go up 40% in coastal Florida, but the state won't let you charge it? Or California won't let you use forward-looking climate models in your rate filings — only historical data — even though historical data is precisely what's broken? You get State Farm, Allstate, and AIG pulling back from those markets. Not out of ideology. Out of math.
People treated those retreats like a scandal. They weren't. They were a rational response to a pricing problem regulators created by mandating rate suppression while the underlying risk kept escalating. Florida, California, Louisiana — all three became case studies in what happens when political pressure and actuarial reality stop overlapping.
The households in those markets didn't get protected by rate suppression. They got abandoned by carriers who couldn't make the numbers work. That's the outcome. That's what happened.
This Is Not an ESG Story
I want to be clear about something because I see this framing constantly and it's wrong. Climate risk in insurance is not an ESG story. It's not about sustainability reporting or carbon pledges or what a carrier posts on LinkedIn during Earth Month.
It's an actuarial story. It's a solvency story.
The carriers that figure out how to price climate risk correctly — using satellite data, ML-based cat models, parametric structures, cat bonds — will survive and probably thrive. There's real money in being right about risk when your competitors are wrong.
The carriers that can't price it, or that operate in markets where regulators won't let them price it, will either retreat or fail. We've already seen the retreats. The failures are a lagging indicator.
The silver lining — and there is one — is that the tools are genuinely better now. Satellite imagery has gotten cheap and fast. IoT sensors on rooftops and inside buildings can surface real-time condition data. Parametric structures let you pay out on an index trigger instead of waiting for adjusters to touch every property. Cat bonds have brought non-insurance capital into the risk stack. These aren't theoretical innovations anymore. They're being deployed.
The regulatory environment just hasn't caught up. States are still fighting about whether carriers can use forward-looking climate models in filings. Meanwhile the climate doesn't care what your rate filing says.
The next ten years in insurance will be won or lost on who prices this right. The data exists. The models exist. The capital structures exist. The only question is whether the industry — and the regulators — can move fast enough to use them before the next category five makes the decision for everyone.