AI data centers' explosive growth creates insurance risks

Row of ten executives in suits with shovels breaking ground, with construction equipment behind them
Chris Wright, U.S. energy secretary, fourth right, and Masayoshi Son, chairman and chief executive officer of SoftBank Group Corp., third right, during a groundbreaking ceremony at the Portsmouth Site, a data center project in Piketon, Ohio, on March 20, 2026.
Brian Kaiser/Bloomberg

Takeaways:

Processing Content
  • Technology may require updates mid-construction
  • Financing construction and supply chain disruptions are potential risks
  • Cooling water lowers fire risks, but increases flood risk

The massive expansion of AI is leading to massive data-center demands that carry unique risks related to technology and natural hazards. This creates both an opportunity and a challenge for commercial insurers.

Spurred by the computing demands of AI, data centers have operational risks around planning, construction, location, sourcing power and even supply-chain disruptions, according to executives of companies directly engaged with building these facilities.

Alastair Swift of Willis
Alastair Swift, head of global specialties at Willis.

First, data centers can become outdated before they're even finished, explained Alastair Swift, head of global specialties at Willis, the insurance brokerage and risk advisory firm that started a digital infrastructure group in February.

"You can start one of these projects and very quickly, the technology that you've got in place could be — depending on how fast it evolves — not out of date, but not as efficient as new stuff," Swift said. "So you might have to change on the fly."

Construction requires financing, which must be covered with credit insurance, Swift noted. Banks backing data-center-construction projects want their credit risk covered, and usually have to spread this out among multiple insurers, according to Swift. 

Supply chain disruptions, most recently because of the war in Iran, can delay projects, he added. Normally these disruptions might be due to direct damage to materials being shipped, but now it can be because materials cannot reach the construction site. This situation might not be covered in shipping insurance, Swift noted.

Tom McFarland of Novidea
Tom McFarland, director of product management and head of analysis, Novidea.
Alexandru Cristian

However, it is possible that such an issue would qualify under business interruption insurance, according to Tom McFarland, director of product management and head of analysis at Novidea, an insurance platform management company. Larger corporations, on the scale of Google or Amazon, can likely absorb such an interruption to data center construction, but others may not be able to, he explained.

Data centers are often built close to sources of water that can be used to cool off the heat generated by their computing hardware. This can introduce greater flood risks. Swift advises data center builders to look carefully at construction locations. 

"One of our clients slightly changed the location of the asset, all within the bounds of the site they had, but they were able to change the location of the build just fractionally," he said. "All of a sudden, that put them outside of a flood plain versus in it."

The heat itself can increase a data center's fire risk, according to McFarland. Data centers are looking at ways to generate power on-site, he said. Some use diesel generators as a backup to electricity from the local power grid. 

"The deployment of on-site power generation is a relatively new thing in this industry," he said. "Instead of [connecting] into the national grid or whatever is market supplied, they can either have the primary energy source as solar panels or a wind farm, or actually even nuclear. Small nuclear reactors are a potential avenue to explore."

A data center operating its own power source would be adding that as a risk, especially for nuclear power, McFarland noted. "A lot of the large insurers are now hiring people or training people with their entire role and responsibility being to understand those risks," he said.


For reprint and licensing requests for this article, click here.
Commercial insurance Artificial intelligence Insurtech Risk management
MORE FROM DIGITAL INSURANCE
Load More