The rapid expansion of artificial intelligence (AI) and generative AI technologies has triggered an unprecedented surge in demand for data centers. These facilities are now growing in scale, complexity, and risk profile at a pace that’s testing the limits of traditional infrastructure – and the insurance industry.
According to Goldman Sachs research, global data center power consumption is expected to grow by 160% by 2030, driven primarily by AI-related workloads. With ChatGPT-like queries requiring up to 10 times more power than a standard web search, the impact on power grids, emissions, and operational risk is undeniable.
But amid this transformation, a critical question remains: Is the insurance industry prepared to support the evolving needs of this sector?
“A new data center typically represents an investment of $500 to $700 million (in insurance coverage), but increasingly we’re seeing multi-building campuses – entire complexes of data centers built adjacent to each other,” said Randy Hodge, chief operating officer of FM, a commercial property insurance provider with decades of experience insuring complex facilities.
“For those, investors and lenders often require full coverage for the total exposed value, which can reach $2 billion or more. In today’s market, it’s very challenging for most data center operators to secure more than $700 million in coverage.”
Meta founder Mark Zuckerberg recently announced that the company plans to spend “hundreds of billions of dollars” to build massive AI-focused data centers across the United States.
Among them is Prometheus, a multi-gigawatt data cluster in New Albany, Ohio, expected to go online in 2026, according to a BBC report. Another facility, Hyperion, is being developed in Louisiana and will scale up to five gigawatts over several years, rivalling the footprint of Manhattan.
While the size of data centers is increasing, so too is their risk exposure and insurance demand. Hodge said FM can provide up to $2 billion in capacity for such sprawling multi-campus facilities. But broadly speaking, he said, there is a capacity gap in the insurance industry that must be addressed as these projects scale.
“Data centers have zero tolerance for downtime,” said Hodge. “They’re expected to operate 24/7, and with AI driving up the stakes, resilience has become the top priority for owners and operators.”
A primary exposure for data centers involves battery energy storage systems (BESS), particularly lithium-ion batteries. These are essential for keeping operations running during outages and for stabilizing energy loads, but they also pose significant fire hazards.
“When these batteries ignite, they’re extremely difficult to contain,” Hodge said. “The risk to the building and critical infrastructure is substantial.”
Other challenges stem from natural catastrophes, especially in regions such as the South and Midwest US, where many data centers are located. These areas are prone to hail, tornadoes, flooding, and convective storms.
Additionally, as more operators turn to self-generated power, including renewables, they also face risks tied to mechanical breakdowns and operational errors.
Beyond property risks, there’s also a growing need for more specialized forms of coverage. Data centers often operate as real estate hubs, leasing space to multiple clients who install their own hardware. These contracts typically include strict service-level guarantees, and failing to meet them can result in heavy financial penalties.
“Not all insurance policies clearly address these liabilities,” Hodge said. “Many providers, and even some clients, aren’t aware of the gap. That’s an area where we’ve developed tailored solutions to fill the void and help clients avoid unpleasant surprises.”
Over the next few years, the insurance industry must grapple with the rapidly evolving risk profiles for data centers. The shift will demand strategic partnerships, as well as a willingness for carriers and brokers to evolve as quickly as their clients’ technologies.
Earlier this year, FM launched a new initiative to help clients navigate risks driven by the AI and data center boom. The program expands its investment in loss prevention engineering and research, and introduces a virtual, on-demand expert resource unit to provide global, rapid-response support.
“Every one of our clients relies on these data center providers for their AI computation as well as their cloud services,” said Hodge. “So, if they are not resilient, then our clients will have more vulnerability to downtime. If data centers are resilient, our clients benefit. That’s where we see strong strategic alignment.”
