AI boom exposes challenge of insuring infrastructure against physical risks

Key points

• AI infrastructure is increasing insurers’ accumulation risk as data centres, power grids and energy projects become more interdependent.

• Climate hazards, including hail, fire, floods, windstorms and earthquakes, could trigger widespread physical damage and business interruption.

• Insurers will need improved modelling, technical underwriting and alternative risk-transfer solutions to cover these emerging exposures.

In light of the AI boom, insurers are reassessing their coverage of large-scale infrastructure projects, including against increased physical risks, new research points out.

With the buildout of AI infrastructure driving up electricity demand, billions of investment are currently directed towards capital-intensive physical assets, such as power grids, battery storage and data centres.

This rapid expansion poses several risks to insurers providing cover for the construction and operations of such assets, insurance giant Swiss Re explains in its latest sigma research report.

Many of these assets are built around access to energy, land, water, skilled labour, logistics and policy incentives. Several are interdependent through common critical supply chains, electricity needs, telecommunications and cloud infrastructure, increasing contagion risks. 

According to Swiss Re, this interdependence leads to increased “accumulation risk” compared to previous generations of infrastructure because diversification is reduced and the potential for losses to spread across multiple companies, sectors and business lines increases.

Moreover, the reinsurer argues competition for labour, specialised equipment and critical components has raised replacement costs and business interruption losses, resulting in higher claims being filed than would otherwise have been the case.

Physical climate risks

The research report emphasises that changing natural catastrophe risk can further “accentuate” the sources of accumulation.

For instance, data centres, solar photovoltaic farms and electricity grid infrastructure are all vulnerable to hail, strong wind and flash floods.

Swiss Re has found that, in the US alone, over a quarter of data centre capacity is in areas that could experience more than three days of large hail per year, which is seen as an elevated risk. 

Data from traditional data centres further suggests that fire is the dominant source of insured losses, in part due to the growing use of lithium-ion batteries. But losses have also resulted from wind, hail and utility outages, with business interruption often being more financially material than direct physical damage, the reinsurer noted. 

Meanwhile, 64% of China’s offshore wind capacity is based in the provinces of Guangdong and Fujian, which are both -ironically- exposed to high windstorms. Swiss Re says the assets’ higher vulnerability to typhoons carries the risk of potential knock-on effects on energy markets and industrial production in China.

While earthquakes have generally benefitted from more advanced modelling, Taiwan’s high exposure is also flagged as a risk, particularly to the country’s globally dominant semiconductor industry.

“About 88% of Taiwan’s semiconductor fabrication plants sit in extreme to very extreme seismic risk zones, built around access to talent and infrastructure,” the report notes.  

Despite strong anti-seismic construction norms, if a strong earthquake were to hit Taiwan, it would likely compromise the country’s chip making production, triggering a large downstream impact, Swiss Re adds.

Insurance solutions

Still, the build out of AI and energy infrastructure also provides a clear opportunity for the insurance industry with demand rising for innovative solutions, the report states.

“This shift could represent one of the largest commercial property and casualty (P&C) opportunities in decades,” Swiss Re claims.

The reinsures estimates that AI data centre construction and operation could generate approximately US$91bn in cumulative premiums by 2030, while renewable energy investment could exceed that, generating around $111bn over the same period.

While construction cover would account for the first wave of insurance demand, the larger and more durable growth opportunity lies in decades of insuring operational risk, Swiss Re says. “As new infrastructure enters operation, insurers will underwrite more high-value property, business interruption and liability exposures,” it argues.

Future coverage against extreme weather events is already being developed, with brokers and reinsurers reportedly exploring the issuance of catastrophe bonds for data centres. Catastrophe bonds have historically been used to transfer natural disaster risks, such as hurricanes and floods, from insurers and governments to outside investors.

The main challenge for insurers now is to understand the risk landscape, through engineering-led underwriting, technical due diligence and improved modelling, to make risks linked to modern infrastructure assets insurable.

Given that hyperscale data centres and next-generation energy infrastructure have limited operating histories, loss frequency and severity are difficult to estimate, Swiss Re stresses, adding that meaningful data will only emerge over the coming years as the first wave of facilities enters operation.

Still, while risks such as pandemics and terrorism would require government intervention, large commercial risks linked to AI data centres and energy infrastructure remain, at present, “fundamentally insurable” through private-market solutions, Swiss Re states. Layered insurance programmes, syndication, reinsurance and alternative capital are expected to enable insurers to assemble capacity for more complex projects.

Written by