Artificial intelligence is turning data centres into one of the fastest-growing classes of critical infrastructure. Electricity demand from data centres rose sharply in 2025, while investment in AI capacity, power generation, cooling and grid connections continued to accelerate. The result is a project pipeline with exceptional asset values and unusually concentrated dependencies.
A data centre loss is rarely limited to damaged servers. A power interruption, cooling failure, construction defect or supplier delay can affect customers in several countries, create contractual penalties and interrupt digital services far beyond the physical site.
This guide explains how owners, operators, investors and tenants can structure insurance around the full lifecycle of an AI-focused data centre.
The expansion of AI capacity is increasing both the scale and density of data centre projects. A single campus may combine high-value computing equipment, substations, transformers, generators, battery systems, liquid cooling and complex fibre connections. Many of these components have long replacement times and limited alternative suppliers.
Local power markets are also becoming part of the insurance discussion. A technically sound facility can still face delayed energisation, restricted grid capacity or dependence on a small number of substations. These constraints affect completion dates, revenue forecasts and the size of a credible business interruption loss.
Property insurance should reflect the interaction between electrical equipment, cooling, fire protection and information technology. Traditional building values may represent only a fraction of the total exposure.
Machinery breakdown wording, electrical disturbance exclusions and testing conditions should be reviewed together. The policy must also clarify whether damaged equipment is valued on a replacement, agreed-value or depreciated basis when technology has changed.
Many major losses emerge before operations begin. Data centre construction combines civil works with specialised electrical and mechanical systems, and commissioning may expose defects that were not visible during installation.
A construction programme should identify critical equipment, manufacturing locations, shipping routes and realistic replacement periods. Delay in start-up insurance must align with the revenue model and recognise that grid connection, customer acceptance and performance testing may all sit on the critical path.
The indemnity period should be based on the time required to restore full contracted capacity, not merely repair the building. Operators should model simultaneous loss of power, cooling and network connectivity, together with the time needed to replace transformers or specialist computing hardware.
Coverage for utilities, cloud platforms, telecommunications providers and other dependent services requires precise definitions. A provider may fail without suffering physical damage, so non-damage extensions and cyber policies should be compared carefully.
Operational technology, building management and cooling controls create a bridge between cyber events and physical operations. A malicious or accidental configuration change can increase temperature, interrupt access or shut down a facility.
The insurance programme should allocate scenarios between cyber, property, technology errors and omissions, and general liability policies. Ambiguous hand-offs can create disputes over whether the event was a security failure, equipment breakdown or professional service error.
For each scenario, calculate property damage, expediting expense, revenue loss, contractual liability and the effect of policy waiting periods and sublimits.
Data centre insurance is an infrastructure exercise, not simply a technology placement. The strongest programmes connect engineering, construction, energy supply, cyber controls and customer contracts to a single loss model.
Next step: Request a data centre risk and insurance programme review from Kompetenz.