The surge of insurance capital into AI-driven data center development is drawing fresh scrutiny from regulators, raising questions about whether the risks behind these fast-growing investments are being fully understood—or properly rated.
The National Association of Insurance Commissioners has begun examining how insured data centers are being evaluated, according to what sources have told the Financial Times.
"The NAIC's Securities Valuation Office reviews investments across a wide range of asset types and sectors, including data centers, as part of its ongoing work on behalf of state insurance regulators," the organization told the FT.
While insurance regulation is set at the state level, the NAIC plays an outsized role in shaping how those rules are applied, particularly regarding capital requirements tied to investment risk.
At the center of the review is a fundamental question: whether insurers—both as underwriters and as investors—are getting a complete picture of the credit risks embedded in data center projects. The concern comes as insurance companies take on an increasingly prominent role in financing AI infrastructure, often backing early-stage developments that rely on investment-grade ratings to attract capital.
Those ratings, though privately issued, must be filed with the NAIC, which uses them to determine how much capital insurers are required to hold. Lower-rated investments carry higher capital burdens, making the accuracy of those ratings critical.
Regulators are expected to focus on several pressure points within the data center model, including tenants' financial strength, lease structures that allow early exits, and construction firms' track records, particularly regarding delays and cost overruns.
Some of those risks are already surfacing. Moody's Ratings reported in February that Amazon, Meta, Alphabet, Oracle, and Microsoft collectively hold an estimated $662 billion in contracted lease obligations tied to data centers, even as many of those projects have yet to break ground, according to its analysis.
At the same time, the structure of those leases is shifting. To keep pace with rapidly evolving AI technology, hyperscalers are increasingly favoring shorter lease terms—often 5 years instead of the traditional 10 to 15 years. That shift introduces greater uncertainty for developers and lenders, particularly given the high upfront costs of building these facilities.
To offset that risk, landlords are turning to residual value guarantees, which compensate owners if a tenant walks away and help preserve a property's value. But those arrangements can also obscure the true scale of financial commitments.
"When the financial commitment made by the hyperscalers is in the form of an RVG that is part of a lease, the accounting can defer the reporting of the expected obligations," Moody's wrote at the time.
The broader backdrop is adding to the unease. According to the Financial Times, a surge in data center construction, combined with questions about underwriting standards and exposure to other tech sectors vulnerable to AI disruption, has contributed to a rise in redemption requests from private funds. That, in turn, is heightening concern among insurers with growing allocations to private credit, including data center debt.
© Arc, All Rights Reserved. Request academic re-use from www.copyright.com. All other uses, submit a request to TMSalesOperations@arc-network.com. For more information visit Asset & Logo Licensing.