First Street, a climate risk modeling firm, analyzed 97 markets and finds 79% of global data center capacity sits in elevated risk areas, with growth hubs like Northern Virginia, Johor, and Marseille most exposed.
The world's largest data center hub sits 30 miles outside Washington, D.C. It also sits on a flood plain that climate models now flag as one of the more exposed pieces of real estate in the global digital economy.
According to a June 18, 2026 analysis from First Street, the climate-risk modeling firm, 79% of the world's installed data center capacity sits in areas facing elevated exposure to at least one of five hazards: flooding, extreme heat, wildfire, wind, or drought. The figure comes from First Street's "Climate Risk in Global Data Center Markets: Implications for Investment and Performance," which examined 97 markets on six continents.
First Street sells its climate models to investors, lenders, and operators, a commercial interest that does not invalidate the numbers but is worth stating outright. The headline 79% figure is a publisher-released finding; the underlying report is referenced by title only in the press release, and the figures have not been independently corroborated in the reporting available here. The rest of this piece treats First Street's analysis as a starting point and flags where independent validation would change the picture.
The geographic collision is sharper than the global number suggests. First Street's analysis breaks the 97 markets into regional groupings, and the hubs driving much of the current AI construction boom sit in the higher-exposure columns: Northern Virginia, Johor in Malaysia, and Marseille in southern France sit in the highest climate-risk tier globally, while lower-risk Nordic markets rank among the least exposed. The full report's regional breakdown is referenced in the press release; the specific figures are not reproduced in the available excerpt and would need to be confirmed against the underlying study before publication.
Chronic and acute exposure are different problems, and they hit operators' books in different ways. Chronic exposure is the slow drip: rising cooling costs as average temperatures climb, falling efficiency as outdoor air can no longer be used for free cooling, and a steady upward pressure on water use in drought-prone regions. Acute exposure is the spike: a flood that takes a building offline, a heat dome that pushes a chiller plant past its design limit, a wildfire that forces a multi-day shutdown or a power cut. First Street's framing collapses both into the 79% number, but the two categories carry different risk profiles and different underwriting consequences.
The financial channel is where the gap is widest. Trillions of dollars are flowing into cloud and AI infrastructure on the assumption that the underlying buildings are long-lived, low-depreciation assets. The press release argues that climate risk "remains largely absent from many underwriting and valuation frameworks" and treats that as a flaw to be corrected. Whether underwriters, asset managers, and reinsurers will adjust premiums, covenants, and discount rates to reflect First Street's risk scores is the open question. Global data center capacity is expected to nearly double again by 2030, which means a great deal of new build is being financed before that adjustment, if it comes, takes hold.
For operators and planners, the practical answer is the part of the report the headline skips over. First Street's data identifies which growth hubs are most exposed and which are less so, and the gap between those lists is large enough to sit on. The Nordic markets exist; the cooling trade-offs in Marseille and Johor are documented; the flood plain in Loudoun County has been mapped for years. The question is whether the capital following AI demand will price those facts into its underwriting, or treat the data center as a piece of real estate that does not have a flood plain.