01 · Tornado · Hyperscale
A hyperscale campus is not a point on the map
AI infrastructure is expanding across campuses worth billions of dollars. We simulated tornado risk at 16 campuses over 4 million years to examine how their size, layout and equipment change the frequency and severity of physical damage.
AI is driving a generational expansion of computing infrastructure, with data centres at its heart. Their scale makes tornado damage an important question for insurers: buildings and equipment worth billions of dollars can be spread across several square kilometres.
Meta's Hyperion development in Louisiana illustrates that scale. Originally announced on 2,250 acres of land, about 9.1 square kilometres, it is planned as the largest campus in Meta's fleet. Louisiana Economic Development now reports a planned investment of more than $50 billion [1a][1b]. That is a project investment figure; the total insured value (TIV) depends on the assets and coverage included in an insurance programme.
At Vayuh, we simulated tornado risk across a broader sample of 16 US hyperscale data-centre campuses to examine how their size and layout affect physical damage. A tornado crosses the landscape as a swath. The buildings, substations, generators and cooling equipment in its path determine how much of a campus is damaged.
Same value. Different footprint.
The same tornado width crosses a different share of the site.
Compact site
A hit can affect a larger share of value.
Spread-out campus
A larger target, with more value outside the swath.
more frequent encounters when the same value is spread from 550 m to 3 km in the study. [V]
Continue reading
Read the full study
Explore how campus layout and equipment shape tornado risk for hyperscale data centres.
Inside the study
- How campus size changes loss frequency
- Which equipment drives physical damage
- What the findings mean for insurance and resilience
Discuss the campus assumptions, component exposure and validation behind this study.