Infrastructure risks
Climate (physical and transition), environmental and social risks for data centers, each with its sources.
Data centers are exposed to climate risk on two fronts: physical hazards that threaten the site and its power/water/network lifelines, and transitionpressures as prices, regulation and expectations shift under decarbonization. Alongside sit direct environmental footprints and social licence-to-operate tensions.
This is a framing to read siting exposure: not a hazard model. Many of the transition risks below are already concrete rules in the Regulation section (PUE caps, grid moratoriums, water limits), and the Atlas Map shows where 8,460 assets concentrate: often in exactly the water-, grid- and heat-stressed regions these risks describe.
Acute & chronic hazards to the facility and its power / water / network lifelines.
Heat & cooling stress
Higher ambient temperatures and heatwaves reduce cooling efficiency.
▸ Impact Raises PUE, power draw and water use; can force throttling or shutdowns. Hot climates cost more to cool year-round.
Example: Cooling failure took a Google Cloud region offline (DCD)Water stress & drought
Evaporative cooling depends on water; droughts and local limits reduce it.
▸ Impact Cooling-water restrictions (e.g. Chile, Netherlands) can curb operation or block new sites in dry regions.
Example: Google pauses a Chile data center over water (PBS)Flooding & sea-level rise
Coastal and riverine sites, and their power/fiber feeds, sit in flood zones.
▸ Impact Inundation risk to the facility and its lifelines; drives siting, hardening and insurance cost.
Example: Climate threats to data centers set to surge (DCD)Storms & extreme wind
Cyclones and severe storms hit structures and, above all, the grid.
▸ Impact Connectivity and utility power, what the site depends on, fail before the building does.
Example: Data centers built in extreme-weather zones (Bloomberg)Wildfire & smoke
Fire and smoke threaten facilities, air-cooling intakes and transmission lines.
▸ Impact Growing exposure in dry regions; smoke fouls free-air cooling and can trip the grid feed.
Example: Data centers and wildfires (Uptime Institute)Cost, policy & market shifts under decarbonization: incl. electricity prices, regulation and water.
Electricity price & volatility
Data centers are power-price takers; carbon costs and demand surges push prices up.
▸ Impact Electricity is the dominant operating cost, rising, volatile prices hit margins and siting economics hardest.
Example: Data centers added ~$23bn to US power bills (Fortune)Efficiency mandates & regulation
Binding PUE caps, KPI reporting and heat-reuse quotas (EU EED, Germany EnEfG).
▸ Impact Raise compliance cost and can rule out inefficient designs, see the Regulation section.
Example: EU mandates data-center energy reporting (DCD)Carbon pricing & renewable/CFE rules
Carbon levies and 100%-renewable obligations (Germany, Ireland).
▸ Impact Reshape power sourcing (PPAs) and run-cost; can gate approval of new load.
Example: Germany: 100% renewable + PUE caps (Columbia Law)Grid access constraints
Connection moratoriums and large-load tariffs (Ireland, Netherlands, Ohio, Virginia).
▸ Impact Delay, cap or block new capacity, increasingly the binding limit on where MW can land.
Example: No new Dublin grid links until 2028 (DCD)Water pricing & restrictions
Water-use limits and cooling-water rules (Chile, Netherlands, Malaysia).
▸ Impact Raise operating cost and constrain siting toward water-rich or water-light-design locations.
Example: Water-stressed regions push back on DC water (TechRepublic)Refrigerant phase-down
HFC phase-downs (EU F-Gas, US AIM Act) push low-GWP refrigerants.
▸ Impact Forces chiller/refrigerant choices and retrofit of existing cooling plant.
Example: Low-GWP refrigerant solutions for data centers (ACHR News)Energy & carbon footprint
Large, continuous electricity demand; carbon depends on the local grid mix.
▸ Impact AI-driven growth strains grid decarbonization; emissions scrutiny rises with scale.
Example: AI set to drive data-center power demand (IEA)Water consumption
Cooling can draw significant potable water.
▸ Impact Competes with local supply: acute in water-stressed regions and a growing licence-to-operate issue.
Example: The AI boom is draining water where it's needed (Bloomberg)Waste heat & land
Facilities reject large amounts of low-grade heat and occupy land/space.
▸ Impact A liability unless reused: heat-reuse mandates (Germany, Nordics) turn it into district-heat value.
Example: Turning data-center waste heat into home heating (WEF)E-waste & materials
Short hardware refresh cycles and embodied carbon in equipment and refrigerants.
▸ Impact Generates electronic waste and material/refrigerant demand across the supply chain.
Example: AI will add to the e-waste problem (MIT Tech Review)Community opposition & moratoriums
Local pushback over land, power and water has produced formal moratoriums and permit fights.
▸ Impact The Netherlands, Ireland and many US counties have paused, capped or confined new builds, a real limit on where capacity can land.
Example: Why communities are pushing back (Harvard Gazette)Competition for power & water
A single campus can out-compete whole towns for scarce grid capacity and water.
▸ Impact Raises affordability, equity and reliability concerns, and is driving special large-load electricity tariffs onto operators.
Example: How AI data centers affect bills & water (Consumer Reports)Noise, land use & amenity
Chillers and backup generators run 24/7; the low-frequency hum carries for hundreds of metres.
▸ Impact Neighbours report sleep and health effects and have taken operators to court, pushing stricter zoning and setback rules.
Example: ‘Living in hell’: neighbours on data-center noise (US News)Few local jobs for the footprint
Data centers are highly automated, even a vast campus may employ only a few dozen permanent staff once built.
▸ Impact Yet they take large tax breaks and heavy land, power and water, so communities increasingly ask whether the local benefit justifies the burden, and some are pausing incentives.
Example: Tax breaks for data centers bring few jobs (Forbes)Peer-reviewed literature and authoritative reports underpinning the risks above (each link verified). Figures and framing here are synthesised from these.
Framing only: a qualitative synthesis, not a facility-level hazard or financial model. Physical exposure depends on local hazard data; transition exposure depends on each jurisdiction's rules (see Regulation) and power market. Re-assess against the sources above before relying on it.