79% of Global Data Center Capacity Faces Climate Risk as Big Tech Reworks AI Cooling
Updated
Updated · CNBC · Jun 29
79% of Global Data Center Capacity Faces Climate Risk as Big Tech Reworks AI Cooling
1 articles · Updated · CNBC · Jun 29
Summary
79% of global data center capacity faces elevated risk from flooding, extreme winds and wildfires, according to First Street, as heatwaves expose how vulnerable AI infrastructure is to weather shocks.
40% of a data center's energy use already goes to cooling, and extreme heat pushes that higher just as air-conditioning demand strains grids, raising blackout risks and forcing operators to curb loads.
Zurich said severe weather has become the top source of loss in its U.S. data-center builders' risk portfolio over the past three years, accounting for about one-third of losses as projects spread into frontier markets.
64% of capacity under construction this year is outside traditional hubs, moving into places such as West Texas, Tennessee, Wisconsin and Ohio, where tornadoes, hail, high winds and heat can hit exposed cooling systems.
Microsoft said it is using site selection, redundancy and real-time monitoring to manage heat risk, while Nvidia said its new AI servers can run cooling liquid at 45C and each 1-degree chiller increase can cut cooling energy costs about 4%.
AI can predict climate disasters, but can its own power-hungry data centers survive them?
As tech giants build private power grids for AI, are they creating a resilient future or leaving the public in the dark?
The 'data heat island' theory is a myth. So what is the real environmental time bomb ticking inside the global AI boom?
The AI Data Center Surge: Climate Vulnerabilities, Resource Strain, and the Race for Sustainable Infrastructure
Overview
The rapid growth of artificial intelligence is driving an unprecedented demand for data center capacity, which in turn is escalating their exposure to critical climate risks. As data centers face increasing threats from a changing climate, such as heatwaves and water stress, traditional risk assessment models that rely on historical climate data are proving inadequate. The climate is no longer behaving as past records would predict, leaving these facilities vulnerable to operational shortfalls and higher costs. This creates significant challenges for the digital infrastructure supporting AI, highlighting the urgent need for updated risk models and resilient strategies.