A wave of investment in data centres, energy infrastructure and advanced manufacturing is enlarging the stock of insurable assets and opening new lines of demand. Data centres are the sharpest illustration, capturing both the scale of the opportunity and the complexity of the risk.

While construction risk is primarily about creating the asset (challenges include physical perils, subcontractor interdependencies and delay), operational risk is about keeping a high-value, multi-tenant critical system continuously available. Once GPUs, tenants, and services are in place, both the value and the operational complexity increase, making business interruption (BI), loss of rent, and service interruption critical.

A single large AI campus can cost billions to build, and that figure can roughly double once the high-value chips inside are installed. The hazards are distinctive, too: sprawling buildings with large flat roofs and humidity-sensitive equipment prove unexpectedly vulnerable to water damage, on top of the wind and hail exposure of their locations.

Location is where the real concentration lies. Operators cluster campuses where power and land are cheap, and several of those hubs rank among the most catastrophe-exposed in the United States. Our analysis finds that around 40% of US data-centre capacity could sit in zones with significant tornado activity, and more than a quarter in areas exposed to frequent large hail. Because the sites are packed so closely, one severe storm can strike several at once, converting what looks like a series of separate risks into a single, concentrated regional loss. Pricing that accumulation correctly and underwriting it becomes essential as the build-out accelerates.

US data centre capacity (GW) by ≥EF1 tornado days per year

US data centre capacity (GW) by large-hail days per year