The rapid buildout of AI infrastructure is creating a new class of complex exposures that may have large implications for the insurance industry. As technology firms invest heavily in AI-enabled data centers, insurers are facing risks that are larger, more interconnected, and more technically challenging than traditional commercial property exposures.
According to data collected by Epoch AI, using SEC filings, capital expenditures (CAPEX) on AI-related infrastructure (i.e., large data centers) by the five largest cloud providers (or hyperscalers) reached $448 billion in 2025 – nearly tripling their spend over 2023.
The scale of investment alone is significant, but there are broader implications for the insurance industry, for both brokers and insurers. Global insurance premiums associated with data centers are projected to more than double from $10.6 billion in 2026 to $24.2 billion by 2030, according to Swiss Re.
Most AI data centers will likely require coverage for:
- Property risks, including construction delays, property damage, and business interruption after operations begin.
- Project cargo insurance to protect equipment in transit.
- Liability coverage can include third-party claims, workers’ compensation, professional indemnity, environmental risks, and evolving technology and cyber exposures.
Increase in AI infrastructure is creating more complex risks
Historically, data centers – physical buildings that host IT infrastructure necessary for storage, data processing, and high-powered computational activities – were often viewed as relatively stable property risks. But AI-focused facilities are different. These hyperscale campuses can require hundreds of millions of dollars or more in capital spending. Once high-density GPUs, backup power systems, networking infrastructure, and cooling technologies are operational, total insured values can increase even more. Many of these facilities also support mission-critical cloud operations, where even short service interruptions may create significant business interruption risk.
In addition, the operational risk profile intensifies materially after occupancy because service continuity becomes essential. A power interruption, cooling failure, or fire event can impact not only physical infrastructure, but also downstream customers and cloud-dependent operations simultaneously.
All of this is creating a new category of interconnected risk that extends beyond traditional property insurance. As a result, insurers face growing risk across property, cyber, business interruption, engineering, and reinsurance lines.
The insurers with strong underwriting discipline are best positioned to succeed in the AI data center space, given the limited historical loss data and constantly evolving risk environment. Compared to traditional infrastructure, hyperscale projects add complexity through faster construction timelines, high asset values, and several forms of aggregation risk, including supply chain disruptions and natural catastrophes, tied to the number of stakeholders involved at each site.
Challenges and opportunities for brokers
The rapid expansion of AI-enabled data centers creates several key industry challenges, which may also represent meaningful opportunities for insurance brokers. These include:
- Evolving operational and engineering risks.
- Limited actuarial and claims history.
- Large insured values and catastrophe concerns.
- Greater demand for underwriting flexibility and technical expertise.
- Increased pressure on reinsurance and capital management.
- Growing need for specialized underwriting organizations.
Large brokers such as Marsh, Aon, WTW, Brown & Brown, and Gallagher stand to benefit as AI data center projects require sophisticated cyber expertise, global placement capabilities, and strategic risk advisory services. These firms can help clients manage complex exposures such as cloud concentration, business interruption, and cyber-physical risks while also structuring alternative risk transfer solutions like captives and parametric insurance.
For example, Aon’s Data Center Lifecycle Insurance Protection Program (DCLP) (launched in 2025 and expanded in 2026), with capacity of $3.5 billion, is a multi-line insurance solution for data centers that serves operational, construction, cyber and financial risks.5
At the same time, AI infrastructure is creating significant accumulation and operational risk concerns for the broader insurance industry. A meaningful portion of U.S. data center capacity is located in regions exposed to natural catastrophes such as severe convective storms, including tornadoes. The clustering of facilities within these geographic areas increases the potential for correlated losses from a single catastrophe event. Furthermore, separate carriers may insure the building, equipment, and business interruption exposure independently, making total portfolio risk harder to monitor. This means a single event could lead to claims across different insurance programs.
Drawing on the historical analogy of Moore’s Law
While the current risk landscape for AI data centers may appear elevated, it is important to recognize that technology-driven industries often follow a pattern of rapid early-stage complexity followed by standardization and efficiency gains. Moore’s Law – which essentially says that as computing power increases, costs will decline – suggests that the AI infrastructure ecosystem may similarly benefit from innovation and efficiency over time. Because of this theory, today’s risks and uncertainty may reflect a transitional phase rather than a permanent change in exposure, suggesting that underwriting approaches and pricing models could evolve alongside the technology itself.
Ultimately, risks associated with AI are becoming more interconnected, technologically complex, and difficult to evaluate through traditional underwriting models alone. For carriers and brokers, that complexity may create opportunities. But as the industry evaluates the next generation of AI-related exposures, success will likely depend less on predicting exactly how these risks evolve and more on formulating the new underwriting models, acquiring the required technical expertise, and building the management capabilities necessary to respond effectively as the market develops.
Contributions to this article by: Eric Hallinan, Managing Director, MarshBerry.
