AI hyperscaler capital expenditure will reach $750 billion in 2026, triggering an insurance demand surge that could total $90 billion in premiums by 2030, Swiss Re's Sigma 2026 report warns. The re/insurance industry currently supports only a fraction of the limits required for these multi-billion-dollar data centre projects, creating a capacity crunch as AI infrastructure investment reshapes commercial risk pools.
Financing demands outpace traditional capacity
New AI data centres can cost more than $20 billion to build before equipment installation. Lenders increasingly require insurance limits that cover the full construction cost, even though maximum probable loss scenarios are far lower. Swiss Re said the re/insurance industry can only support a fraction of this limit at competitive rates for traditional construction risk policies.
Ivan Gonzalez, CEO of Swiss Re Corporate Solutions, said: "This unprecedented wave of investments into the data centres requires insurance, most of the time, across the entire life cycle of a project. We estimate that the demand could be, from an insurance premium standpoint, of up to $90 billion on an aggregated basis until 2030. These are premiums. If you think about the capital, the capacity associated with this demand, we're talking about 200, 300, 400 billion of capital that is required to support these projects."
Concentrated risks from campus-style builds
Modern AI data centres are larger and more complex than traditional facilities. Built as campuses with dense, interdependent systems, they concentrate risk within single sites. Once GPUs, tenants, and services are established, value and operational complexity surge. Continuous availability becomes critical, elevating business interruption, loss of rent, and service disruption risks.
These projects require advanced cooling, high-voltage power, backup infrastructure, and sophisticated hardware and security software. The globalised supply chain for components introduces another vulnerability, as delays in construction and repairs can cascade.
Natural hazard exposure compounds the challenge
Around 40% of US data-centre capacity could sit in significant-to-very-high tornado-risk zones, and more than a quarter faces elevated large-hail exposure, according to the report. Insured losses from natural catastrophes are rising by 5-7% annually on average in real terms over the long term.
Risk engineering and alternative solutions
Swiss Re's risk engineering team developed the 3W model - focusing on watts (energy requirements and sustainability), water (local supply for cooling), and waiting time (supply chain and repair delays). Gonzalez said a coordinated approach across prevention, insurance, and risk financing is needed to build resilience.
He outlined three priorities:
- Risk engineering to assess how data centres are constructed
- Alternative risk transfer, including parametric solutions and cat bonds, often paired with captive management
- Credit and surety to cover financing, construction, and operational risks
The investment cycle extends beyond AI into broader energy infrastructure and advanced manufacturing, creating new pools of insurable risk across property, engineering, cyber, liability, and business interruption lines.
Why this matters for insurance professionals
The data centre boom is pushing the boundaries of traditional underwriting. Brokers and underwriters must grapple with single-site exposures that demand limits far exceeding available capacity, while risk engineers need to assess highly technical, interdependent systems. Alternative risk transfer mechanisms - from parametrics to captives - are becoming central to closing the gap. As these trends accelerate, insurance professionals can deepen their expertise through AI for Insurance training to stay current with these evolving risks.
Your membership also unlocks:
