AI Data Centres Strain Global Insurance Capacity Limits

The rapid proliferation of artificial intelligence (AI)-driven data centres is placing unprecedented pressure on the global insurance and reinsurance markets, as the scale and complexity of these facilities begin to exceed traditional coverage capabilities. A recent report by Swiss Re underscores how multi-billion-dollar developments are exposing structural limitations in underwriting capacity and reshaping the dynamics of risk transfer.

Escalating Project Costs Create Coverage Gaps

AI data centres have evolved into some of the most capital-intensive infrastructure projects in the world. Construction costs for a single site can reach as high as $20 billion, and the inclusion of high-performance computing hardware—particularly graphics processing units (GPUs) and advanced servers—can effectively double the insured value.

This dramatic rise in asset valuation has created a widening gap between insurance demand and available supply. In many cases, project financing requires full coverage for both construction and operational phases. However, reinsurance markets often struggle to provide sufficient capacity at commercially viable rates, leaving insurers unable to meet the full extent of coverage required.

As a result, developers and investors are increasingly confronted with a critical bottleneck: the inability to secure comprehensive risk protection for assets of such magnitude.

Rapid Growth in Insurance Demand

The surge in AI infrastructure investment is driving a corresponding increase in demand for specialised insurance products. Global data centre insurance premiums are projected to rise significantly—from $10.6 billion at present to approximately $24.2 billion by 2030.

This growth reflects not only the expansion of physical infrastructure but also the increasing complexity of associated risks. Insurers must now account for a wide array of emerging exposures, ranging from cyber-physical vulnerabilities to advanced cooling technologies, many of which lack long-term historical data for accurate risk modelling.

Geographic Concentration Amplifies Catastrophe Risk

A major concern identified by Swiss Re is the concentration of data centres in regions prone to severe weather events. In the United States, more than 25% of data centre capacity is located in areas experiencing at least three large hail events annually. Additionally, around 40% of facilities are situated in zones vulnerable to three or more EF1-strength tornado days each year.

This geographic clustering heightens the risk of correlated losses, where a single catastrophic event could simultaneously impact multiple high-value sites. Such scenarios challenge insurers’ ability to diversify risk and could result in substantial aggregate claims.

Major Risk Drivers in AI Data Centres

The evolving risk profile of AI data centres is illustrated in the table below:

Fire remains the most financially significant risk, accounting for a disproportionately high share of loss costs relative to its frequency. The increasing integration of lithium-ion battery backup systems within server racks introduces additional ignition sources, raising the likelihood of severe fire incidents within confined data environments.

Water-related risks are also becoming more pronounced. Modern AI servers generate substantial heat, necessitating advanced cooling solutions such as direct-to-chip liquid cooling. Whilst effective, these systems introduce new vulnerabilities, including leaks, installation defects, and maintenance failures, all of which can result in costly damage and operational disruption.

Power Demands and Business Interruption Exposure

Business interruption risk is another critical concern. According to Uptime Institute, power supply issues account for approximately 45% of data centre outages.

AI workloads significantly increase energy consumption. Traditional servers typically require between 5 and 15 kilowatts per rack, whereas AI servers can demand in excess of 100 kilowatts. This surge in power requirements is driving the adoption of on-site energy generation and battery storage systems, which introduce additional hazards such as fire, explosion, and toxic gas emissions.

Accumulation Risk and Systemic Complexity

Insurers are also grappling with accumulation risk—the potential for a single incident to trigger multiple claims across interconnected systems. Large data centres are frequently insured through separate policies covering buildings, equipment, and energy infrastructure. This fragmented approach can obscure the total level of exposure.

Moreover, shared infrastructure—such as power grids, cooling systems, and fire suppression mechanisms—means that a single failure could affect multiple tenants simultaneously. This interconnectedness increases the likelihood of cascading losses, complicating both underwriting and claims management.

Outlook: Urgent Need for Industry Adaptation

The findings from Swiss Re suggest that the insurance industry must evolve rapidly to keep pace with the demands of AI-driven infrastructure. Traditional models of underwriting, capacity allocation, and risk assessment may prove inadequate for assets of this scale and technological sophistication.

Innovations in risk-sharing mechanisms, capital deployment, and pricing strategies will be essential to bridge the widening gap between insurable demand and available supply. As AI data centres continue to expand globally, addressing these challenges will be critical—not only for insurers but also for the broader digital economy that increasingly relies on these high-value, mission-critical facilities.