AI-driven hyperscale data centers are creating a new generation of risks and challenges that extend well beyond power shortages and chip supply, according to a new report from Zurich North America.
The unprecedented scale, speed and complexity of AI data center construction are exposing the industry to new threats previously unknown to the older generation of data centers, ranging from severe weather and energy constraints to insurance capacity, labor shortages and geopolitical disruptions, in its report, “Data Center Risks Right Now: Six Critical Questions to Enable a Resilient Buildout.”
The report states that hyperscalers are prepared to spend an estimated $710 billion in capital expenditures during 2026, and global investment in data centers is projected to top $7 trillion by 2030. New capacity added between 2026 and 2030 is expected to total roughly 100 gigawatts, equivalent to the peak electricity demand of about nine New York Cities.
Much of that is because these new data centers are not like any previous generations of data center development in that modern AI campuses can span up to 20 buildings, consume as much as 2,000 megawatts of electricity and house billions of dollars’ worth of servers and cooling equipment.
In addition, insurance providers are struggling to keep pace with the massive growth in projected value of these data centers. Zurich says the average value of the data centers it insured has jumped from roughly $150 million five years ago to about $3 billion today, while the largest campuses can be measured in the tens of billions of dollars.
The report cited six areas of concern, the primary of which is growing: severe weather. Severe weather has surpassed fire as leading construction threat due to changing geography. Zurich states that 64% of U.S. data center capacity currently under construction is located outside traditional markets such as Northern Virginia, in areas are known for bad weather.
They include West Texas, Tennessee, Wisconsin and Ohio, where tornadoes, hailstorms and high winds present new hazards. Zurich says severe weather has become the largest source of losses in its U.S. builders-risk portfolio over the past three years, surpassing fire as the industry’s dominant construction threat. Weather accounts for 32% of losses in Zurich’s data center portfolio, followed by fire and equipment damage.
The second issue is compressed construction schedules. Operators are increasingly beginning operating portions of a campus where construction is complete and while construction continues elsewhere. That means welding and other, heavy equipment plus incomplete fire protection coexist with active server halls containing sensitive computing equipment.
Beyond construction and to absolutely no surprise, Zurich identifies energy infrastructure as one of the defining challenges facing AI expansion. The report notes that U.S. data center electricity demand increased roughly 22% in a single year and is expected to nearly triple to approximately 134 gigawatts by 2030.
Likewise, water availability is becoming equally important as power as AI workloads generate more heat and require increasingly sophisticated cooling systems. The report notes that many operators are deploying closed-loop water recycling systems or shifting toward air cooling where possible in a bid to reduce water consumption.
Downtime has become more expensive because so much expensive equipment is involved, even minor operational failures can become multimillion-dollar events.
The report also warns that replacement equipment often requires months to arrive, citing industry estimates that switchgear may take up to 85 weeks to replace and generators as long as 100 weeks.
Workforce shortages have raised operational concerns as the AI construction boom is straining the labor market, and they don’t mean The IT staff to run the place, they need people to build it. Zurich cited a report from Associated General Contractors of America that 92% of U.S. construction firms are having difficulty finding qualified workers.
The report concludes that geopolitical tensions, regulatory scrutiny and emerging technologies will increasingly influence where and how data centers are built. Among the trends Zurich identifies are growing political concern over electricity prices and water consumption, increased reliance on nuclear energy, interest in integrating future quantum computing systems and even early proposals for orbital data centers supported by solar power.