Written by: Kamil Ponicki, Director of Talent Acquisition, Digital Colliers
Swiss Re forecasts global data-centre insurance premiums will hit $20-30B a year by 2030. At the same time, about 40% of US data-centre capacity sits in tornado-prone areas. If you are underwriting loans for data-centre buildout, you need a site-risk model that prices weather exposure, insurance cost, and single-site dependency before you sign the term sheet. The operators who skip this step are learning the hard way when the first big storm hits.
Why lenders need to price site risk now
Data-centre operators are chasing cheap land and cheap power. That combination often lands them in the tornado belt or near flood plains. The insurance market is responding with steep premium curves. If you underwrite a loan without modeling the site risk, you are assuming the operator can absorb rising insurance costs and weather downtime without defaulting. That assumption does not hold.
The pattern I keep seeing: lenders price the construction loan and the operating line based on current insurance quotes. Two years later, the operator faces a 40% premium increase or a coverage reduction because the site sits in a high-risk zone. The loan was structured assuming stable insurance costs. Now the operator has a cash-flow problem and the lender has a stressed loan.
The three layers every site-risk model needs
A working site-risk model for data-centre loans has three components. You need all three before you underwrite.
First, weather exposure. Map the site against tornado, hurricane, flood, and wildfire zones. Do not rely on the operator's insurance broker to surface this. Pull FEMA flood maps, NOAA storm data, and state wildfire risk assessments. If the site sits in a high-risk zone, model the insurance premium trajectory and the expected downtime cost for the operator.
Second, insurance trajectory. Insurance premiums for data centres in high-risk zones are climbing faster than the market average. Model the premium curve over the loan term. If the operator is projecting 5% annual insurance cost growth but the site sits in tornado alley, you need to stress-test at 15-20% growth. The gap between the operator's model and the real premium curve is your credit risk.
Third, single-site dependency. If the borrower's revenue model depends on one data centre and that site goes down for two weeks, can they service the debt? Most operators have business-interruption insurance, but the coverage limits and waiting periods vary. Model the revenue impact of a two-week outage and a four-week outage. If the operator cannot cover debt service during that window, the site risk is a credit risk.
Map site risk before underwriting
The winning move is simple. Before you underwrite the next data-centre loan, run the site through a weather-exposure map, model the insurance premium curve, and stress-test the operator's revenue against a multi-week outage. If the numbers do not work, walk away or reprice the loan.
Operators who are serious about site risk are mapping this before they break ground. Lenders who are not mapping it are discovering the exposure when it is too late to reprice. The insurance market is repricing data-centre risk in real time. Your underwriting model needs to do the same.