The AI Boom Faces a New Problem: Extreme Weather Hitting Data Centers

The race to build the infrastructure powering artificial intelligence is running into an unavoidable and increasingly unpredictable obstacle: severe weather causing damage to data centers.

As Europe swelters through a record-breaking heatwave, the same soaring temperatures are straining power grids and public infrastructure and exposing a growing vulnerability for the massive data centers that fuel AI models. Climate-related hazards, from extreme heat and flooding to tornadoes and wildfires, are becoming one of the biggest operational and financial risks facing the sector.

Technology companies are pouring billions of dollars into hyperscale data centers equipped with thousands of advanced graphics processing units (GPUs), whose enormous computing power also generates tremendous amounts of heat that must be continuously managed.

According to insurance company Zurich, severe weather has become the leading cause of losses within its U.S. data center builders’ risk portfolio over the past three years, now accounting for roughly one-third of its losses in the sector.

“Severe weather is no longer something that can be treated as a background exposure,” Patrick McBride, Zurich’s head of international construction, told CNBC. “It is one of the first things we and the owners we work with look at.”

The risks are expanding as developers increasingly move data centers beyond traditional technology hubs into suburban and rural regions where land is cheaper and electricity is more readily available. Those locations, however, often have less historical infrastructure designed to withstand increasingly frequent extreme weather events.

McBride noted that approximately 64% of data center capacity currently under construction is being built outside established markets such as Northern Virginia. Instead, companies are expanding into so-called frontier markets, including West Texas, Tennessee, Wisconsin, and Ohio, where facilities may face greater exposure to severe weather conditions like tornadoes, hailstorms, and damaging winds capable of affecting vast rooftop cooling systems and energy installations.

Research suggests the risks are already widespread, as a recent study by climate risk analytics firm First Street found that 79% of global data center capacity is exposed to elevated climate hazards, including flooding, wildfires and extreme winds.

Extreme heat presents a particularly complex challenge because it simultaneously increases electricity demand while making cooling systems work harder. According to Mishal Thadani, CEO of AI software company Rhizome, cooling already accounts for roughly 40% of a typical data center’s energy consumption under normal conditions. During heatwaves, that percentage climbs sharply at the very moment electric grids are strained by widespread air conditioning use.

The problem becomes even more significant as AI facilities consume electricity comparable to that of entire communities. “Now add facilities that each pull as much power as a hundred thousand homes,” Thadani said. “The heat and the load hit the same wires at the same time.”

Major technology companies are already adapting their infrastructure strategies, with Microsoft telling CNBC it designs its data centers to operate “reliably in a wide range of environmental conditions, with site selection, redundant systems, and real-time monitoring helping manage risks from extreme heat and severe weather.”

Meanwhile, Nvidia recently announced that its newest AI servers can safely operate with cooling liquids at temperatures reaching 45 degrees Celsius, allowing facilities to reduce cooling demands. According to the company, increasing chiller temperatures by just one degree Celsius can lower cooling energy costs by approximately 4%