Over the past several years, climate innovation has largely focused on one question: how do we measure what’s coming? Hundreds of startups, billions of dollars, and a growing ecosystem of satellite providers, AI modelers, and analytics platforms have all pointed at the same problem. Better prediction.
That work matters. But our read on the full climate risk value chain led us to a different conclusion. The bottleneck isn’t prediction. It’s the absence of financial products that act on the prediction.
The system that absorbs climate risk is retreating
The global property and casualty insurance market collects roughly $1.6 trillion in premiums each year. Its job: take climate risk off the books of those who can’t carry it and hand it to capital that can. That function is failing. In 2023, climate events caused $357 billion in economic losses. Only $123 billion was insured. The $234 billion gap is not a rounding error, and it’s getting wider.
The reason is structural. Actuarial pricing assumes the past predicts the future, and that assumption is breaking. The U.S. experienced 28 billion-dollar weather disasters in 2023, compared to an average of 3.3 per year in the 1980s. The headline events aren’t the problem. Storms, hail, wildfire, flooding are smaller, more frequent, harder to model, and collectively more damaging than any single hurricane.
The industry’s response has been to withdraw. Major carriers are pulling back from California, Florida, and Louisiana. FAIR plans are stressed in many states. 35.6 million U.S. homes face coverage reductions. The carriers that built the system are leaving the markets where it’s needed most.
The retreat of traditional capital from climate risk is one of the largest openings in financial services today. We built our thesis around four patterns, mapped across the full value chain: risk modeling, geospatial intelligence, risk transfer, and underwriting.
1. Distribution Matters More Than Data
Climate risk analytics is a crowded market. Around $300 million went into parametric climate products between 2010 and 2022. Most failed, not because the products were bad, but because the distribution model proved difficult to scale. Standalone products that require new buying behavior don’t scale. The ones that do embed inside an existing transaction: a broker relationship, a lending workflow, or a software platform where the purchase happens naturally.
2. The Biggest Unmanaged Exposure Isn’t Catastrophic
It’s recurring. Insurance operators consistently point not to the hundred-year hurricane but to high-frequency, low-severity weather volatility: power outages, supply chain disruptions, crop losses, and revenue disruptions from events that occur every year but fall below the threshold for a traditional insurance claim. Most commercial business interruption policies require direct physical damage to trigger. An 18-hour grid outage that forces a restaurant to discard $15,000 of inventory is not a covered event. For millions of businesses, the most frequent climate loss they experience is one that their insurance policies were never designed to cover.
3. The Winners Compound Across the Value Chain
A company that only models risk sells a commoditized input. A company that combines its own intelligence layer with a financial product and a distribution channel builds feedback loops: proprietary data tightens pricing, real loss outcomes calibrate the model, and each layer reinforces the next. We look for companies building at more than one layer.
4. The Geography of the Opportunity
In the U.S., the protection gap is growing because carriers are retreating. In Latin America, many of these carriers never arrived. Government backstops comparable to FAIR plans or FEMA disaster recovery largely don’t exist. When climate events hit, losses fall directly on households, agricultural lenders, and local businesses with no mechanism to transfer them. Insurance penetration across the region remains a fraction of developed markets. The same dynamics that make LatAm harder for incumbents (limited data, thin distribution, regulatory fragmentation) are what make it more open for startups building climate-native products from scratch. As geospatial data gets cheaper and parametric structures gain regulatory acceptance, the surface area for new entrants keeps expanding.
Putting it together, we look for companies building a genuine financial product, not a pure analytics layer. The exposure they target is recurring and everyday, not the rare catastrophic event. They sell through channels where the purchase already happens. Their intelligence is proprietary and compounds over time.
We believe the next generation of climate companies won’t just measure risk more precisely. They’ll build the products that transfer it.