The infrastructure race fuelling artificial intelligence's explosive growth masks a critical vulnerability that investors and technology companies are only beginning to confront: the vast majority of data centres powering the sector sit in locations exposed to serious climate risks. A new analysis by climate-risk analytics firm First Street reveals that nearly four-fifths of global data-centre capacity operates in regions facing elevated threats from flooding, extreme wind, wildfires, and other acute climate hazards, upending assumptions about the stability of AI infrastructure investment.
The findings arrive at a particularly acute moment for the technology industry. Amazon, Microsoft, Google, Meta, Oracle, and other major corporations are racing to secure computing capacity to support AI applications, driving an unprecedented surge in data-centre construction and expansion. Competition for land, electricity, and grid access has already intensified dramatically. Climate exposure now represents an additional complexity that will shape investment decisions, operational costs, and ultimately the geography of artificial intelligence development. The International Energy Agency projects that global data-centre electricity consumption will more than double by 2030, underscoring the scale of expansion underway and the magnitude of climate risk exposure.
Asia-Pacific emerges as the region facing the most severe vulnerability, with 89% of data-centre capacity located in areas experiencing elevated acute climate risk. This finding carries particular significance for Malaysia and the broader Southeast Asian region, where rapid data-centre development is outpacing comprehensive climate risk assessment. The Americas face substantially lower exposure at 50%, while Europe, the Middle East, and Africa experience 46% exposure. Within Asia-Pacific, fast-growing markets including Johor in Malaysia and other strategic locations have been identified as simultaneously experiencing significant climate vulnerability and rapid capacity growth. Meanwhile, Nordic markets have emerged as the lowest-risk geography, attracting operators seeking long-term stability.
The risk profile extends beyond acute climate events. Just over half of the data-centre capacity facing acute climate hazards also sits in locations experiencing chronic climate stress, including prolonged extreme heat and drought conditions. These chronic pressures operate on a different timeline than sudden disasters but carry equally consequential economic implications. A facility that appears economically attractive today because it offers cheap land, readily available power, and expedited permitting could become substantially more expensive to operate if heat waves intensify, water availability declines, or storm frequency accelerates over the facility's 20 to 30-year operational lifespan. This temporal mismatch between initial site selection and long-term operating costs represents a significant valuation risk that many investors have not adequately priced.
Climate risks damage data-centre economics through distinct mechanisms. Acute hazards such as flooding and wildfires create sudden operational downtime and trigger insurance claims that raise premiums or exhaust coverage limits. Chronic stressors including heat and drought gradually erode profitability by increasing cooling costs, raising water expenses, and straining power supply reliability. Jeremy Porter, chief economist at First Street, characterises acute risk as primarily a volatility problem affecting operational stability, while chronic risk functions as a margin problem that slowly compresses profits. Both pathways ultimately converge on the same outcome: reduced net operating income, weakened loan coverage ratios, and diminished asset valuations. This analysis suggests that climate risk pricing, currently undervalued across much of the market, will become a more prominent driver of real estate values and investment returns.
The vulnerability transcends individual facility boundaries in ways that complicate traditional risk management. Data centres depend on extensive supporting infrastructure including roads, electrical substations, transmission lines, water systems, fibre optic connections, and local electrical grids. A facility engineered to withstand direct climate impacts may still experience extended downtime if surrounding infrastructure fails. A flooded transmission line, overwhelmed water treatment system, or damaged fibre connection can paralyse operations even if the data centre itself remains undamaged. As Porter notes, while operators can construct flood walls around individual buildings, they cannot isolate their facilities from the broader grid infrastructure that delivers power and connectivity. Effective risk management therefore requires underwriting not merely the asset but the entire location and its supporting systems.
For businesses relying on cloud computing services or data-centre colocation arrangements, climate exposure operates as an indirect but material threat to operations. Climate-related disruptions at major data-centre facilities could manifest through higher hosting fees as operators pass increased costs to customers, reduced regional capacity as damaged facilities leave the market, intensified insurance pressures, or degraded service uptime during extreme weather events. As data-centre exposure becomes more constrained by climate risk, companies dependent on cloud infrastructure may face restricted choices and elevated prices, transforming climate resilience into a fundamental competitive factor.
Some technology leaders have begun implementing design innovations to mitigate climate vulnerability. Microsoft has developed AI-optimised data centres featuring chip-level cooling systems that eliminate water consumption for cooling purposes entirely, a significant advancement given that data-centre operations consume vast quantities of water for temperature management. Google conducts watershed assessments before site selection and prioritises air cooling or recycled water systems in regions where fresh water availability faces acute stress. These engineering solutions represent genuine progress but introduce additional capital requirements that strain operators already contending with extraordinary demand for computing capacity and constrained power supplies in many regions.
The economics of climate resilience present a stark challenge for the data-centre sector. Implementing robust cooling systems, hardening buildings against extreme weather, installing redundant power connections, and conducting thorough climate risk analysis all require substantial capital investment. These expenses accumulate within an industry already struggling to keep pace with soaring demand for AI infrastructure and competing fiercely for limited power supplies and skilled labour. The intersection of climate risk and infrastructure scarcity creates a dual constraint that will force difficult choices about where new capacity can be built and what premium companies must pay for locations offering genuine long-term stability. For Malaysia and the Southeast Asian region, this dynamic will shape whether the region can sustain its position as a growing data-centre hub or whether climate vulnerability diverts investment toward more resilient jurisdictions.
