AI data centres are becoming too complex for standard insurance

A new Allianz Commercial report finds a single hyperscale event can now prompt claims across six lines simultaneously

AI data centres are becoming too complex for standard insurance

Transformation

By Mark Rosanes

The rapid global build-out of artificial intelligence infrastructure is generating one of the most complex insurance challenges the market has faced in decades - and a new report from Allianz Commercial makes clear that the industry is still catching up with the risks it is being asked to cover.

Allianz Commercial's data centre construction boom: risks and claims trends report finds that annual investment in data centres is projected to nearly double from around US$500 billion in 2024 to more than US$1 trillion as early as 2027. The global data centre insurance market is expected to more than double in value from around US$11 billion today to more than US$24 billion by 2030, driven by rising insured values, rapid capacity expansion and increasing operational complexity.

For insurers and brokers, the scale of the opportunity is matched by the complexity of the risk - and claims data already shows that complexity can be expensive.

Fire leads losses, but nat cat risk is the structural problem

Allianz Commercial's analysis of insurance industry data centre claims found that fire is the leading driver of loss severity, accounting for well over 50% of around €700 million (US$800 million) in analysed losses. Natural catastrophe activity ranks second, followed by deliberate acts - including crime and cyber incidents - and power failure. Water damage, by contrast, is the most frequent cause of claims by volume, followed by deliberate acts, fire, and equipment breakdown.

Business interruption is the primary driver of claims severity by line of insurance, a finding that underscores just how financially damaging even short periods of downtime can be when facilities are running mission-critical AI workloads.

The nat cat dimension carries a structural warning that is easy to overlook amid the pace of construction. Around 79% of global data centre capacity is already located in areas with heightened natural catastrophe exposure, while 54% faces chronic heat and drought stress.

In the UK, the government has conceded planning errors in approving a hyperscale facility at Iver, Buckinghamshire, pending a proper environmental review. That is a sign that the collision between data centre growth and planning and environmental constraints is already reshaping how projects are approved and insured in Britain.

Germany, the UK, and Ireland remain Europe's major data centre markets, but faster expansion is expected in Spain, Finland and Denmark, where power availability and permitting conditions are comparatively more favourable. As projects move beyond established hubs, they move into less well-understood risk environments, which is precisely where claims surprises tend to emerge.

A single event, multiple claims

What makes the modern data centre particularly difficult to underwrite is not the size of any individual risk, but how quickly a single event can cascade across multiple lines simultaneously. Hyperscale and colocation campuses bring together multiple tenants, construction works, servers, utilities and on-site infrastructure in one physical or operational space.

Allianz Commercial's claims analysis shows that in hyperscale facilities, damage to external cooling systems, hot works-related fire damage and a delay in start-up caused by power disturbances have each resulted in losses in the US$50 million to US$100 million range. Construction costs for a single AI campus can exceed US$20 billion, with insured values rising substantially once high-performance computing equipment is installed.

Christian Kolbe, global head of construction claims at Allianz Commercial, said the underwriting question was shifting from property value to value concentration. "For insurers, the key question is not only the value of the building, but the concentration of value and dependency inside and around it," he said. "Power, cooling, batteries, fibre routes, testing and commissioning, and business continuity planning are all part of the same risk picture."

That concentration problem is already pushing the market toward integrated solutions. Aon expanded its Data Centre Lifecycle Insurance Programme to US$5 billion in July, extending coverage from construction through to long-term operations across property, construction, business interruption, cyber, and liability lines. Marsh's Nimbus facility, backed by Lloyd's and company market insurers, provides up to €1 billion in construction all-risks cover and up to €350 million for delay in start-up for UK and European projects.

Where the coverage gaps actually open up

For brokers, the coverage complexity of data centre risk does not sit neatly within a single line. A large campus carries property, construction, business interruption, supply chain, cyber, and liability exposure, often simultaneously. Underwriters are increasingly requiring detailed technical information before agreeing terms.

The construction-to-operational handover is where gaps most commonly open. That transition should be managed as a specific, dated checkpoint rather than a general area of concern: brokers should confirm the exact handover date at which a construction all-risks policy ends and the operational property and business interruption programme begins, and verify there is no coverage lapse or overlap dispute at that boundary. Given that Allianz Commercial's own claims data identifies power disturbances during commissioning as a distinct, high-severity loss driver, the same review should confirm whether cyber and liability coverage are already active during the testing and commissioning phase specifically, rather than only from the point a facility is declared fully operational, since that phase is precisely when several of the largest losses in the analysis originated.

The Swiss Re Institute's sigma insights report on insuring AI data centre risks, published in July, noted that large data centres are sometimes presented to insurers through separate programmes - covering buildings, equipment and power plants independently - making it difficult for carriers to track overall exposure. A single loss event can therefore impact several insurance programmes simultaneously.

That fragmentation is a specific, checkable submission task, not just a structural observation. Brokers should request a client's complete data centre insurance programme structure at renewal and map exactly where the building, the equipment, and the power infrastructure each sit, whether under one combined placement or three separate ones, then confirm that limits, triggers and indemnity periods are consistent across all three. A programme that looks adequately covered when each component is reviewed in isolation can still leave a client materially exposed if a single event, say, a fire that damages both the building and on-site power infrastructure, hits gaps or inconsistent triggers between policies that were never designed to be read together.

Thomas Lillelund, chief executive of Allianz Commercial, said insurance had become a prerequisite for financing large-scale AI infrastructure. "Comprehensive insurance cover has become a prerequisite for financing many large-scale AI infrastructure projects," he said. "Success will increasingly depend on resilience: access to power, reliable supply chains, robust construction controls, as well as climate-aware site selection and insurance programmes that reflect the true accumulation risk."

The Allianz Commercial report's central message for the broker community is that resilience must be designed in from the earliest planning stage, not retrofitted after construction begins. For a market that is doubling in value by the end of the decade, that is a window that is closing faster than many clients realise.

Related Stories

Keep up with the latest news and events

Join our mailing list, it’s free!