A new climate risk report finds that the majority of the world’s data center capacity is already operating under conditions that strain infrastructure, inflate costs, and expose investors to risks that current valuations do not reflect.
The global data center industry is in the middle of a construction boom. Driven by the rapid expansion of cloud computing and artificial intelligence, installed power capacity has grown from roughly 33 gigawatts in 2015 to approximately 114 gigawatts today, and is projected to nearly double again by 2030. Capital flows into the sector reached around $61 billion in 2025 alone, with McKinsey estimating that AI-focused data center infrastructure alone will require $5.2 trillion in capital expenditure through 2030, part of a broader projected total of up to $6.7 trillion across the full industry.
Approximately 54 percent of global data center capacity operates in environments that exceed defined thresholds for heat or drought
Against that backdrop, a report published by the climate risk analytics firm First Street argues that a critical variable is being systematically overlooked in site selection, underwriting, and asset valuation: physical climate risk.
Most capacity is already under stress
The 18th First Street Risk Assessment quantifies physical climate exposure across 97 investible data center markets identified by Cushman and Wakefield, evaluating both chronic risks, such as extreme heat and water scarcity, and acute hazards, including flooding, wind, and wildfire. The findings are substantial in scope. Approximately 54 percent of global data center capacity operates in environments that exceed defined thresholds for heat or drought. Acute risk exposure is even broader: 79 percent of global capacity sits in markets with elevated flood, wind, or wildfire risk.
The report draws a clear distinction between how chronic and acute risks affect financial performance
Regional patterns vary significantly. In the Asia-Pacific region, 89 percent of capacity faces chronic heat or drought stress, compared with 50 percent in the Americas and 46 percent in EMEA. For acute hazards, the Americas stand out most sharply, with 86 percent of capacity exposed to elevated flood, wind, or wildfire risk, against 60 percent in APAC and 25 percent in EMEA.
The report draws a clear distinction between how chronic and acute risks affect financial performance. Chronic stressors erode margins gradually. Higher ambient temperatures reduce cooling efficiency and increase energy consumption, while water scarcity constrains cooling capacity or raises operating costs. These pressures translate into higher and more volatile electrical loads, compressing net operating income and shortening effective asset life. Acute hazards, by contrast, introduce episodic disruption: physical damage, extended downtime, emergency capital expenditure, and insurance shocks that add volatility to cash flows.
The report illustrates these mechanisms through two case studies. In July 2022, record temperatures exceeding 40 degrees Celsius in the United Kingdom caused cooling system failures at Google and Oracle data centers in London, resulting in multi-hour outages across cloud platforms. The failure was not caused by physical damage but by ambient conditions exceeding design thresholds by more than 1.5 degrees Celsius. In 2025, severe flooding in Texas impaired substations and transmission infrastructure across one of the largest US data center markets, forcing extended reliance on backup diesel generation and delaying maintenance access. In both cases, the impact extended beyond the individual facility to the regional infrastructure on which the facilities depended.
A key finding of the methodology section is that redundancy systems, while designed to handle isolated failures, may not fully offset disruptions when multiple inputs are constrained simultaneously during a regional event. Grid instability, fuel supply constraints, and water shortages can coincide, increasing the probability of degraded performance or downtime beyond modeled expectations.
Water sits at the intersection of both risk categories in the report. As a chronic stressor, drought and water scarcity constrain the cooling systems that data centers depend on to manage heat loads. As an acute hazard, flooding directly threatens the power substations and transmission infrastructure that keep facilities online. What the aggregate risk figures do not capture is the scale of water demand that makes this exposure so consequential: US data centers consumed approximately 17 billion gallons of water in 2023, with hyperscale and colocation facilities accounting for roughly 84 percent of that total, according to US Department of Energy estimates. Projections for 2028 range from 16 to 33 billion gallons annually for hyperscale facilities alone. The markets the report identifies as facing the highest chronic stress, including large portions of the Asia-Pacific region, the southeastern United States, and parts of Southern Europe, are in many cases already facing basin-level water constraints independent of data center demand. As the sector continues to expand into secondary markets partly to escape permitting and power bottlenecks in established hubs, water availability is a variable that current site selection frameworks are not consistently pricing in.
Scale is being built in the riskiest places
The geographic distribution of risk in the report raises a broader structural concern. Among the ten markets ranked as having the highest composite climate risk, three, namely Northern Virginia, the Carolinas, and Atlanta, are among the largest and fastest-growing data center markets in the United States. Johor in Malaysia, Singapore, Batam in Indonesia, Marseille in France, Bangkok in Thailand, and Beijing and New York/Northern New Jersey round out the top ten.
At the other end of the scale, the ten lowest-risk markets are dominated by northern European cities: Stockholm ranks first, followed by London, Copenhagen, Amsterdam, Oslo, and Dublin. Toronto, the San Francisco Bay Area, Montreal/Quebec, and Bogota complete the group. These markets benefit from cooler climates, lower flood and wildfire exposure, and more stable infrastructure conditions.
The report frames this as a structural misalignment: the industry is concentrating significant new capacity in locations where operating conditions are among the most demanding, rather than the most favorable. The report notes that 30 to 50 percent of projects originally expected to come online in 2026 are at risk of delay, partly because established hubs face constraints on power availability, land access, and permitting. As developers shift toward secondary and emerging markets to overcome those barriers, climate exposure in those new locations is not consistently factored into site selection.
Climate risk needs to move from a secondary consideration to a core underwriting input
The central argument of the report is that climate risk is currently underpriced across data center markets. Markets with comparable power costs, connectivity, and demand fundamentals are being underwritten as equivalent despite materially different long-term risk profiles. The report identifies consequences for net operating income stability, insurance availability, debt capacity, refinancing terms, and exit valuations.
For investors and lenders, the report’s conclusion is that climate risk needs to move from a secondary consideration to a core underwriting input. This means adjusting baseline operating assumptions, stress-testing downtime scenarios and insurance availability, and differentiating markets based on long-term infrastructure reliability rather than treating location as a secondary variable after power and connectivity. Where that mispricing persists, the report argues, it represents both a source of downside risk for assets that carry more exposure than their valuations reflect, and an opportunity for investors who can identify and price that gap more accurately.
