Data center projects are drawing billions in insurance capacity, but overlapping exposures and less recurring construction premiums complicate the growth outlook.

Data centers are creating a major source of insurance demand, but the central market challenge is tracking how much exposure insurers accumulate across construction, operations, equipment and power. That distinction matters: executives at the Insurance Leadership Forum in Colorado Springs said capacity remains available for many large projects, while concentrations at individual campuses and in storm-prone regions require careful management.
Large projects can attract substantial capacity
Thousands of data centers are expected to be built worldwide, with many projects valued at $3 billion to $5 billion, said Ryan Barber, Aon’s New York-based global head of property. He described the facilities as highly engineered risks and said the market has been able to assemble substantial limits for them.
“We’ve done placements now with limits as high as $8.5 billion, and that was over a year ago; we could probably do placements in excess of $10 billion now,” Barber said.
Aon is developing an operational property insurance facility for data centers. The broker estimates the sector could produce more than $100 billion in new insurance premiums over the next five to 10 years, Barber said. Those figures point to a sizable market opportunity, though the premium outlook depends partly on the pace and shape of development.
Executives discussed the sector in early October at the annual Insurance Leadership Forum, sponsored by the Council of Insurance Agents & Brokers and held at The Broadmoor resort in Colorado Springs. The event brings senior leaders from brokerages, insurers and related businesses together for individual meetings and industry discussions.
Accumulation is the central underwriting concern
Large campuses are often built in phases. That can leave insurers covering construction and completed operations at the same location at the same time. High-value computer equipment and dedicated power generation can add further exposures tied to one development, making it harder for insurers and reinsurers to understand their total potential loss from a site or portfolio.
“The biggest struggle right now for insurers and reinsurers, for that matter, is to just track and understand the accumulation of risk that they’re taking on from these data centers,” Barber said.
Pat Donnelly, Willis’ Chicago-based head of risk and broking for North America, said aggregation also shapes decisions about sites and insurers. Some carriers are willing to cover several phases of a development, while others prefer to limit their participation to specific segments. That affects how a project’s insurance program is assembled and how exposures are distributed among markets.
Insurers must weigh data center risks against the property exposures they already carry, said Marc Orloff, Boston-based president of global risk solutions North America at Liberty Mutual. The issue is not simply whether a project can be insured; it is how much additional concentration a carrier is willing to hold and how that concentration could affect its balance sheet.
“We all write a lot of property already … so it’s all about managing tolerances. How much of our exposure do we want? How much volatility do we want on our balance sheets?” Orloff said.
Geography adds another layer. Concentrations of expensive facilities in areas exposed to severe convective storms can increase uncertainty for insurers, Orloff said. The combination of large values and shared regional hazards means that individual placements need to be viewed alongside broader portfolio exposure.
Coverage availability varies by line
Property and business interruption capacity remains plentiful for most data center projects, according to Marc Kunney, San Francisco-based chief client officer at EPIC Insurance Brokers & Consultants. Professional liability presents a tighter market, he said, underscoring that the insurance opportunity is not equally broad across all lines.
“Where some of the challenges are is on the professional liability, which is a much more limited marketplace,” Kunney said.
That distinction matters for buyers arranging coverage across complex projects: ample property capacity does not guarantee similar options for every liability exposure. The mix of construction, operations, technology and power-related risks calls for insurers to assess each coverage line on its own terms.
Construction premiums may not become lasting revenue
The volume of development is producing significant premium today, but some of that revenue is tied to the construction phase. Builders risk premiums can fall if building activity slows, while operational property coverage may generate lower premium volumes after projects are completed, Orloff said.
“It’s not a recurring revenue stream,” he said.
That makes the sector’s long-term value to insurers less straightforward than the project pipeline alone suggests. The number and scale of facilities create a large near-term opportunity, while the duration and composition of premium depend on how much construction continues and how operational coverage develops. For insurers, the practical test is whether growth can be matched with clear exposure data, sound accumulation limits and coverage across the lines projects require.
AI exclusions face resistance as insurers favor underwriting
The same Insurance Leadership Forum discussions also highlighted a different challenge for insurers: how to handle artificial intelligence exposures when the technology is embedded across business operations. Broad AI exclusions have struggled to gain traction, with brokers and policyholders wary that sweeping language could remove coverage for risks that are difficult to separate from ordinary business activity.
Kunney said exclusions are not widely accepted among policyholders. “AI isn’t the risk per se. AI is an acceleration or augmentation of the delivery, and so it’s really hard to exclude AI,” he said. With organizations across industries using AI, he questioned what would remain covered under a broad exclusion.
Some insurers have offered affirmative AI coverage, but Kunney said buyers should read the terms carefully. Policies that name covered AI-related perils while leaving others outside the grant could operate as “cloaked exclusions,” he said.
Donnelly said he has not seen AI exclusions become prevalent in placements. Because AI is integrated into many business activities, broad restrictions can be difficult to apply in practice. “AI is too embedded,” he said.
Zurich U.S. CEO Alex Wells said the company has reviewed optional general liability exclusions addressing generative AI exposures that ISO, the insurance industry standards organization, filed last year. He believes they are more likely to be adopted by smaller insurers that routinely use ISO forms. Zurich does not use blanket AI exclusions, he said.
“We prefer to think about AI as an underwriting question rather than a contract question,” Wells said. He distinguished between businesses that use AI in their own operations and companies that develop AI services and products. Coverage questions may change as case law develops, he said.
Shawn Ram, chief revenue officer at cyber insurer Coalition, said he has seen numerous AI exclusions filed, but not widely adopted. Some brokers have told Coalition they would move business if insurers imposed such restrictions, he said. Carriers may also hesitate to be among the first to adopt exclusions that could put them at a competitive disadvantage.
Coalition has affirmative AI coverage in its policies. “In 2024, we announced that we are affirmatively covering AI exposures. We’ve only expanded that since then,” Ram said.
Across both topics, the market response points toward a closer examination of exposure than a single broad rule. Data center insurers are monitoring how risk gathers across phases, locations and coverage lines; for AI, several executives described underwriting as a more workable tool than blanket exclusions. The next measure of market growth will be how well insurers can price and track those exposures as projects reach operation and AI use continues to spread.
Discussion
Sign in to join the discussion.