A $14B El Paso campus is fully financed and only partly insured

Meta and BlackRock announced a joint venture on July 28, 2026 to build a $14 billion AI campus on about 1,000 acres in El Paso, targeting 1 gigawatt of capacity, with construction already more than six months underway when the deal was made public ^[1][2]. According to a report from cryptobriefing.com, the project is now large enough that conventional insurance markets can only cover part of it, and the campus is proceeding with partial rather than comprehensive coverage ^[11].

The capital stack explains why that matters. BlackRock funds hold 80% and Meta 20%, with Meta the sole initial tenant leasing capacity back from the venture ^[4]. BlackRock is contributing roughly $4.9 billion in cash ^[6]; Meta is contributing land and work in progress valued at about $2.3 billion, plus a distribution of roughly $1 billion to square the ownership split ^[5]. Underneath sits a $12.5 billion debt package ^[7], equal to about 89% of the headline project cost ^[1]. Meta has also written a residual value guarantee with a threshold near $13 billion, declining as the asset depreciates, which the report describes as a floor giving BlackRock's investors downside protection ^[8]. That floor is about 104% of the debt ^[2], which means the lenders are, in substance, underwriting Meta's credit rather than the building.

One caution on the numbers as published: $4.9 billion plus $2.3 billion plus $12.5 billion is $19.7 billion, some $5.7 billion above the $14 billion headline ^[3], and the source does not reconcile the difference. Read the components, not the round number.

The underwriting problem is structural, not clerical. Insurance prices by pooling many comparable assets, and a one-off asset at this scale has almost no comparable pricing data ^[9]. Marsh is involved in the project's risk analysis, and the report cites industry observers calling this a "data center insurance supercycle," with specialised products such as Marsh's Nimbus line built to fill gaps that standard commercial policies leave open ^[10]. The outcome at El Paso is coverage that exists but does not span every risk category, because full cover is either unavailable or prohibitively expensive ^[11].

Texas sharpens it. ERCOT runs in near-total isolation from the rest of the US grid, which limits its ability to import power during emergencies ^[12], and Winter Storm Uri in 2021 produced prolonged outages, cascading failures and billions of dollars in economic damage ^[13]. Non-damage business interruption tied to grid failure is among the hardest categories to price, because the loss is not a physical event with a clear dollar figure attached ^[14]. A $13 billion value floor does not answer that: on the report's own description it protects what the asset is worth, not the revenue from a gigawatt sitting idle ^[8][14].

The consequence is financial. Lenders and institutional investors require coverage as a condition of financing, and when coverage is partial or non-standard it creates friction in deal structuring and can raise the cost of debt ^[15]. Power and silicon are procurable at a price. Risk transfer capacity is not, and until the market develops products that reliably cover $10 billion-plus assets including ERCOT-specific grid risk, the report argues every project in this class will meet some version of the same shortfall ^[16].

Watch whether the 2028 in-service date ^[3] holds while coverage is negotiated, and whether the next gigawatt campus arrives with a named insurance consortium attached rather than only a named asset manager.