14 Billion Dollar AI Data Center Insured for Only $450 Million! Meta and BlackRock's Sopaipilla Project Reveals Less Than 3% Insurance Coverage

The 1GW AI data center 'Sopaipilla,' jointly developed by Meta and BlackRock in El Paso, Texas, has a total development cost of approximately $14 billion. Although it has secured hundreds of billions in debt financing and investment-grade ratings, insurance covers only a few hundred million dollars—far below the project's scale—revealing a structural insufficiency in the insurance market's underwriting capacity as ultra-large AI infrastructure expands.

Under the agreement, funds managed by BlackRock hold 80% ownership, while Meta retains 20% and serves as the sole initial tenant, utilizing all computing power through a long-term lease. Meta contributes land and in-progress construction assets worth approximately $2.3 billion and receives a one-time distribution of about $1 billion. BlackRock injects approximately $4.9 billion in cash, with the remaining $12.5 billion financed through debt.

In terms of insurance, the Sopaipilla project, following advice from Marsh—the world's largest insurance broker—has purchased only limited coverage: a $218 million cap on rent relief for construction delays, $645 million in terrorism insurance, up to $427 million in all-risk property insurance during construction, and $450 million after operations begin (with an annual premium of about $5 million). Commercial general liability insurance has per-incident and aggregate limits of $50 million each. The Sopaipilla project explicitly does not insure against total loss; losses exceeding these limits will be borne by the joint venture entity, potentially passing on to lenders and investors.

Valuations of mega-scale data centers often reach tens of billions of dollars, making full coverage unattainable for traditional insurance markets. Although dozens of commercial insurers participate, many are reluctant to accumulate large exposures at a single location, fearing concentrated risks from natural disasters, power outages, or litigation over construction delays. As a result, insurance brokers have generally shifted toward 'Probable Maximum Loss' (PML) assessments, advising stakeholders to accept losses beyond PML rather than pursue full insurance.

For the Sopaipilla project, Marsh's PML is based on a fire scenario occurring once every 250 to 500 years, corresponding to a 3.9% to 7.7% probability of occurrence during Meta's 20-year lease term.

BlackRock emphasized last month, prior to its bond issuance, that El Paso faces relatively low extreme weather risks and that modern disaster-resilient data centers rarely suffer total losses even when disasters occur.

S&P analyst Viviane Gosselin notes that if insurance payouts are insufficient, Meta would need to cover a gap of up to $450 million. If a severe incident causes project delays exceeding 18 months, Meta can terminate the lease without penalty—the lease being the core support for the debt's value. The debt maintains strong investment-grade ratings: S&P assigned A+, while Fitch and KBRA rated it AA-.

This insurance gap is not an isolated case. As AI infrastructure scales from billions to tens of billions of dollars, insurance underwriting capacity has become a new bottleneck in financing. Previously, lenders required full insurance for mid-sized data centers, but now they are beginning to accept partial coverage for mega-projects. Insurance brokers play a critical role in transactions, as their assessments directly influence credit ratings, but this also raises concerns about conflicts of interest—since the same broker may advise multiple parties and sell insurance products.

For the market, this means that the credit risks hidden within the AI capital expenditure boom are gradually becoming visible. While the current debt structure is supported by Meta's lease commitment and a declining residual value guarantee of about $13 billion, if an event exceeding the PML occurs, losses will fall more heavily on bondholders and infrastructure funds.

In the future, similar projects may require higher risk premiums, more complex layered insurance structures, or greater reliance on the balance sheets of tech giants for backing. Discussions on this topic are already circulating in the market this week, with investors beginning to reassess the tail risks of financing mega-scale AI data centers.