TechTalk: Hyperscale data centre boom triggers ‘call to action’ to break down insurance silos

As the internet caught the world in its web in the early 2000s, the first hyperscale data centres emerged to support the expansion of cloud and digital infrastructure.

These facilities were substantially larger than their predecessors. According to market research firm Blackridge Research and Consulting, hyperscalers often exceed one million square feet in size and are typically built by tech giants to “manage vast amounts of data and handle extensive traffic with enhanced efficiency and security by utilising numerous servers”.

And, alongside the artificial intelligence (AI) boom in recent years, hyperscale campuses have evolved into multibillion pound assets that combine real estate, energy generation, technology infrastructure and complex contractual structures within a single site.

With capital pouring into AI infrastructure, the demand for data centre insurance has risen. According to S&P Global Ratings’ recent report entitled ‘The Insurance Gap Is Reshaping Hyperscale Data Center Finance’, released on 8 June 2026, data centres could generate around $10bn (£7.9bn) in new insurance premiums this year.

Speaking to Insurance Times, Sam Tiltman, digital infrastructure leader for the UK at Marsh Risk, said that the broker’s discussions with over 25 insurers over the last six months revealed that data centres are either “the number one priority” or among the top three areas of focus for carriers – many are musing how insurance can be “best deployed to support the digital infrastructure ecosystem”, he explained.

But, with this rise in opportunity, the transition toward mega AI data centres has also fundamentally reshaped the risk landscape of digital infrastructure.

This is the overarching view shared by Patricia Kwan, senior credit analyst at S&P Global Ratings, who explained that this “shift has been driven by the integration of high density computing components, such as graphics processing units (GPUs), and the resulting complexity of energy requirements”.

Hyperscaler risk, Kwan said, is “intrinsically linked to computing capacity, power utilisation and the high degree of facility interconnectedness”.

This dynamic has opened “significant opportunities for insurers”, she explained, in lines including contractors’ all risks, delay in startup (DSU), service level agreement (SLA), marine cargo, commercial auto liability, parametric insurance, workers’ compensation and surety bonds.

However, Kwan noted that the insurability of emerging data centre risks depends on “the development of specialised underwriting expertise for these complex assets, the refinement of policy wording, the scope of business interruption (BI) coverage and the availability of reinsurance or alternative third party capital”.

She continued: “Consequently, effective underwriting of these massive campuses requires precise risk pricing that accounts for asset concentration and, crucially, the potential for cascading failures.”

Evolving risk profile

Assessing the systemic exposure of interconnected risks is increasingly important as coverage pressure points evolve.

Chris Ives, partner at Fenchurch Law, told Insurance Times that data centres’ “risk profile is evolving in a way which is likely to generate more complex and higher value disputes”.

He explained that these disagreements tend to include “the growing aggregation risk – as multiple customers rely on a single facility or cloud region – increased dependency on uninterrupted power and cooling infrastructure and the convergence of cyber and physical perils where a cyber event can trigger tangible asset damage and wide-scale outages”.

Given the concentration of exposure and the increasingly critical infrastructure status of data centres, Ives believes that “it is likely only a matter of time before a systemic loss event tests both policy wordings and insurers’ appetite to respond”.

Most major brokers and carriers, Ives noted, “have developed integrated or nose-to-tail insurance programmes” which are designed to provide continuous cover from construction – including DSU – through to operational risks.

However, the emergence of more integrated insurance programmes is no panacea for data centre risk. Ives identified a newer emerging issue as “the classification of loss, a distinction which can significantly affect both limits and applicable exclusions”.

For example, this encompasses whether a loss is properly characterised as a first party property BI loss of the operator or a third party liability exposure. This distinction becomes particularly significant when outages trigger SLA penalties – when a service provider fails to meet agreed terms and conditions outlined in a service level agreement with its client.

Ives added that “SLA liabilities are a challenging area for data centre operators from an insurance perspective” as, in practice, cover is often limited because many SLAs “operate as strict liability or service credits, triggered by downtime, regardless of fault”.

‘Multiparallel perils’

The financial impact of digital infrastructure outages is severe, making continuous performance of paramount importance.

According to a white paper by parametric insurance provider Parametrix, entitled ‘An Introduction to Data Centre SLAs: Key Terms, Financial Risk and Underwriting Implications’, released in June 2026, a 45-minute outage of a 100 megawatt data centre with $144m (£113m) in annual rent could result in a service credit loss of $24m (£18.9m) – reducing annual cash flow by approximately 42%.

A service credit is effectively a discount or reduced pricing used to compensate a client’s account if SLA metrics are missed. Other options commonly used to address SLA violations include providing liquidated damages or a pre-agreed compensation amount, or the ability to terminate the service contract without any financial penalties. All of these routes have a financial bearing on the service provider in question.

Speaking to Insurance Times, Sharon Haran, chief commercial officer at MGA and Lloyd’s coverholder Parametrix, said that “one of the unique aspects about SLA violations” is that this risk does not fit neatly into a single insurance category of cover. In his opinion, insurance policies typically ”work in verticals” concerning siloed lines of business, while SLA violations can be triggered by ”multiparallel perils coming from different disciplines”.

For him, today’s available coverage normally excludes contractual liabilities, meaning that “the customer has no remedy for SLA penalties”.

He continued: “They might be compensated for BI and for the revenue loss that they had during this event, but the penalties could be 30 times higher because if you are out of operation for a day, the penalty that you’re going to get could be two months of rent.

“Those SLAs are also coming with an early termination right to the tenant in their agreement, which is a death penalty to the project.”

Seeking to address this exposure, Haran explained that Parametrix’s main product when the MGA was founded in 2019 was cloud outage insurance – this used in-house monitoring technology to “constantly track the availability of hundreds of data centres worldwide”.

Within this data set, Haran said Parametrix has approximately “10 years of data that indicates the likelihood that a data centre will go down”, enabling the business to price and underwrite this specific exposure accurately.

He added: “Once there is an SLA violation that’s leading to a penalty, as long as that violation is caused due to power issues, cooling issues or connectivity issues – we are paying.

“Our policy is also offering a remedy for [the financial impact of early termination], where we are offering the operator a few months of lease to enable them to find an alternative tenant or negotiate with the tenant to stay.”

‘Call to action’

As data centre exposures become increasingly interconnected, insurers’ ability to close current coverage gaps relies on obtaining sufficient capacity to absorb these risks.

While the global insurance market is reacting to surging demand for coverage, Kwan noted that a significant capacity gap remains among traditional insurers.

Indeed, S&P Global Rating’s aforementioned report revealed that insurance markets are adjusting to a step change in asset scale. This is because hyperscale data centre campuses can extend into the $20bn to $50bn (£15.7bn to £39.4bn) total insured value range during the construction phase alone, which is well beyond traditional single risk benchmarks.

As a result, Kwan anticipates that certain material risks, particularly those emerging during the operational phase, such as BI and technology driven IT equipment losses, “will likely remain self-insured or only partially covered”.

She continued: “This shortfall is expected to drive an increased reliance on captive insurance structures, [a form of self-insurance to retain risk in-house rather than buying all cover from the open market], and may catalyse the entry of alternative capital into the sector.”

For Tiltman, the challenge for insurance or reinsurance carriers is that they have “fixed capital”, which makes it harder to be flexible in the face of complex projects that span multiple exposures.

“I’m saying insurers need to work together and bring capabilities,” he asserted. 

“The call to action is more flexible capacity. It’s the trend I see in the near future. Deploying more flexible blocks of capacities to support these projects, which break down speciality silos as well as product silos.”