The data centres powering the boom in artificial intelligence may be more exposed to climate risk than investors realise.
Nearly 80% of global data-centre capacity is in markets facing elevated risk from flooding, extreme wind, wildfire, and other acute climate hazards, according to a report released today by climate-risk analytics firm First Street. Just over half of these buildings are also in markets exposed to chronic climate stress, including extreme heat and drought.
The findings add another pressure point to the AI infrastructure race. Amazon, Microsoft, Google, Meta, Oracle, and other major technology companies are spending heavily to secure the computing capacity needed for AI. Developers are already competing for land, electricity, and grid access. Climate exposure may become another cost of doing business.
“Acute climate risk is now the base case, not the downside scenario,” Jeremy Porter, chief economist at First Street, told Inc. “The risk is mispriced today, which is downside for some and opportunity for others.”
First Street examined 97 data-centre markets globally. It found that 79% of capacity faces elevated acute climate risk. The report comes as AI demand is pushing data-centre construction into overdrive. The International Energy Agency expects global data-centre electricity consumption to more than double by 2030. CBRE has also warned that record-low vacancy rates and power constraints are already shaping where new projects can be built.
Data centres are usually expected to operate for 20 to 30 years. A site that looks attractive today because it has cheap land, available power, or fast permitting could become more expensive if heat, drought, storms, or flooding worsen over time.
An emerging risk
Porter said acute and chronic risks hit data-centre economics in different ways. Flood and wildfire can create downtime and insurance shocks, while heat and drought can raise energy and water costs over time.
“Acute risk is a volatility problem. Chronic risk is a margin problem,” Porter said. “Both end up in the same place: [net operating income], coverage, and valuation.”
That risk does not stop at the walls of the facility. Data centres rely on roads, substations, transmission lines, water systems, fibre connections, and local grids. A building can avoid direct damage and still face downtime if the infrastructure around it fails.
“You can flood-wall the building. You cannot flood-wall the grid,” Porter said. “You have to underwrite the location, not just the asset.”
For businesses that depend on cloud or colocation providers, the exposure may be indirect but still material. Climate-related disruption could eventually show up through higher hosting costs, tighter regional capacity, insurance pressure, or weaker uptime during extreme weather.
Why some AI hubs are more exposed
Some fast-growing markets are among the most vulnerable. First Street found that the Asia-Pacific region had the highest share of at-risk capacity, at 89%. The Americas stood at 50%, while Europe, the Middle East, and Africa stood at 46%. The firm identified Northern Virginia, Johor in Malaysia, and Marseille, France, as fast-growing markets with significant exposure. Nordic markets had the lowest climate risk.
Some operators are already changing the way data centres are cooled. Microsoft has said its newer AI-optimised design uses chip-level cooling and consumes zero water for cooling. Google says it assesses local watersheds before building new data centres and chooses air cooling or recycled water where water sources are at high risk.
But resilience costs money. More robust cooling systems, hardened buildings, redundant power connections, and careful site selection all add more expenses to a sector already under pressure from soaring AI demand. – Inc./TNS
