When the Past Stops Predicting the Future

All insurance pricing rests on a single, quiet assumption: that the past is a usable guide to the future. Actuaries look backward at decades of losses to estimate the odds of losses ahead. This works beautifully as long as the underlying risk is stable. Climate change attacks the assumption at its root, and that attack — more than any single storm — is what makes climate a genuine frontier for the insurance industry and a first-order concern for anyone siting long-lived physical infrastructure.

The mechanism is worth stating precisely, because it is easy to blur into slogans. Insurers price catastrophe risk using models built on historical frequency and severity: how often a region floods, how strong its storms get, how far a wildfire tends to spread. When the climate is stationary, that history is a reliable sample of the future. When the climate is shifting, the historical record systematically understates tomorrow's risk — the hundred-year flood arrives every couple of decades, the wildfire season lengthens, the storms intensify. The data an actuary trusts most is precisely the data that is quietly going stale. This is not primarily about any one disaster; it is about the erosion of the statistical foundation the whole industry stands on. Uncertainty itself is the exposure.

Insurers respond to rising uncertainty in predictable, structural ways, and each one matters to an infrastructure operator. The first response is repricing: premiums rise in high-risk regions as models update to reflect worse odds. The second, and more consequential, is withdrawal. When a carrier decides it can no longer price a risk with confidence, it does not simply charge more — it stops writing the risk at all. Availability, not affordability, becomes the binding constraint. Whole categories of property in the most exposed regions can find that cover thins out or disappears, not because anyone can prove the next year will be catastrophic, but because the range of plausible outcomes has grown too wide to underwrite comfortably. The third response is tighter terms: lower sub-limits on climate perils, higher deductibles, more exclusions — the same defensive moves that showed up in cyber, deployed here against weather.

This is where climate risk collides directly with the central thesis of the subject. A data center is a multi-decade asset. It is sited today but must remain insurable for its entire operating life, and its financing assumes it will be. If a region's climate risk is rising, the facility may be insurable now and progressively harder to insure later — a slow-motion constraint that a snapshot of today's premiums completely misses. Worse, the perils that climate intensifies are exactly the correlated, regional ones that insurers most fear aggregating: floods, wildfires, and windstorms hit whole areas at once, so the very clustering that data-center economics encourages compounds the very risk that climate is worsening. Cheap-power siting and climate exposure are frequently the same map, and insurance is the discipline that reads both layers together.

There is a constructive side, and it is where parametric products and better data re-enter the story. Because climate perils are increasingly modeled with satellite observation, dense sensor networks, and improving physical simulation, they lend themselves to objective triggers — which is part of why parametric cover and climate risk have grown up together. Faster, index-based payouts are a natural fit for perils that are becoming more frequent and more correlated. And the modeling itself is improving: catastrophe models are the industry's attempt to replace a stale historical record with forward-looking physical simulation, pricing the future rather than merely extrapolating the past. The frontier is a race between deteriorating stationarity on one side and improving models and instruments on the other.

The durable lesson is a shift in mindset. Historically, you could treat insurability as a fixed background condition — cover existed, you bought it, you moved on. Climate turns insurability into a dynamic variable that can degrade over an asset's life, sometimes faster than the asset depreciates. For long-lived infrastructure, that means climate risk is not an environmental footnote to be handled by a sustainability team; it is a core financial and siting question that determines whether the asset can be insured, financed, and operated for its full intended life. The past has stopped being a reliable guide, and pretending otherwise is the most expensive assumption an operator can make.