When a single infrastructure project costs more than the GDP of a small country, the usual rules start to break down. Meta and BlackRock’s joint venture to build a $14 billion AI data center campus in El Paso, Texas is running into exactly that problem: the project is so large that conventional insurance markets can only cover part of it.
The campus, spanning roughly 1,000 acres and targeting 1 gigawatt of computing capacity, was already under construction for more than six months before the partnership was formally announced on July 28, 2026. It is projected to come online in 2028.
How the deal is structured
The ownership split gives BlackRock funds an 80% stake, with Meta holding the remaining 20% and serving as the sole initial tenant by leasing capacity back from the joint venture.
Meta’s contribution to the deal includes land and construction already in progress, valued at roughly $2.3 billion, plus a distribution of approximately $1 billion to align the two parties’ ownership stakes. BlackRock is putting in around $4.9 billion in cash, with a $12.5 billion debt financing package underpinning the broader project budget.
Meta has also provided a residual value guarantee threshold of approximately $13 billion, which will decrease over time as the asset depreciates. That guarantee is essentially a floor on the asset’s value, giving BlackRock’s investors a degree of downside protection that pure equity ownership wouldn’t provide.
The insurance problem no one has fully solved
Building a data center that costs north of $10 billion introduces an underwriting challenge the insurance industry has not historically had to solve. Insurers operate by pooling risk across many similar assets. When a single asset is unprecedented in scale, comparable pricing data barely exists.
Risk advisory firm Marsh is involved in the project’s risk analysis, and industry observers describe the current moment as a “data center insurance supercycle,” where surging project valuations are outpacing the capacity of traditional insurance products to respond. Specialized offerings, such as Marsh’s Nimbus product line, have emerged specifically to fill gaps that standard commercial policies leave open on mega-scale infrastructure.
The result for the El Paso campus is partial coverage. The project carries insurance, but not comprehensive coverage across all risk categories, because the combination of high replacement cost and limited market capacity makes full coverage either unavailable or prohibitively expensive.
Texas adds a layer of complexity that goes beyond raw dollar size. The state’s ERCOT grid operates in near-total isolation from the broader US electricity network, a design quirk that limits its ability to import power from neighboring grids during emergencies. Winter Storm Uri in 2021 demonstrated what that isolation can mean in practice: prolonged outages, cascading failures, and billions of dollars in economic damage across the state.
For a data center drawing a full gigawatt of power, a multi-day outage is not a nuisance. It is a catastrophic business interruption event. Insuring against non-damage business interruptions tied to grid failures is one of the harder coverage categories to price, because the loss isn’t a physical event with a clear dollar figure attached to destroyed property.
What this signals for AI infrastructure investment
The insurance gap is a real constraint on how fast capital can deploy. Lenders and institutional investors require insurance coverage as a condition of financing. When coverage is partial or non-standard, it creates friction in deal structuring and can raise the cost of debt. Until the insurance market develops products that can reliably cover $10 billion-plus data center assets, including ERCOT-specific grid risk, every mega-project in this category will face some version of the same coverage shortfall that the El Paso campus is navigating now.