Meta and BlackRock form $14 billion joint venture to build 1 GW AI data center in El Paso

Meta and BlackRock, the world's largest asset manager, announced a venture on July 28, 2026, to develop and operate a $14 billion data center campus in El Paso, Texas, with BlackRock-managed funds taking an 80% ownership stake and Meta retaining the rest.

What's new

According to Reuters' reporting on the deal: "Meta said BlackRock-managed funds will take an 80% ownership stake in the venture, with Meta retaining the remaining 20%." Meta is contributing land and in-progress construction assets worth about $2.3 billion, while BlackRock is putting in roughly $4.9 billion in cash, with $12.5 billion of BlackRock's investment financed through debt. The campus, already under construction, is designed to deliver 1 gigawatt of compute capacity: "The El Paso data center campus, already under construction, is designed to provide 1 gigawatt of compute capacity, essential for Meta's AI technologies and supporting its core business. Operations are expected to commence in 2028."

Context

This isn't Meta's first data center project in El Paso — the company broke ground on an AI-optimized campus there in October 2025, and this joint venture builds directly on that site. It's part of a broader pattern of hyperscalers turning to outside capital to fund AI infrastructure at a pace their own balance sheets can't match alone: Reuters noted that "the race to build out AI infrastructure has prompted tech giants to turn to debt sales worth tens of billions of dollars and seek external capital from fund managers such as BlackRock due to an unprecedented scale of investment." Meta has used similar capital-partnership structures elsewhere, including with Constellation on clean energy and with Reliance on a data center in India, alongside its own direct investments — including a January commitment of up to 6.6 gigawatts of nuclear power specifically to feed AI compute growth.

The 80/20 equity split, with BlackRock taking the majority stake while Meta keeps a fifth of the equity and presumably operational control, is a financing structure spreading fast across the industry: off-balance-sheet joint ventures with asset managers and private credit funds let hyperscalers build gigawatt-scale AI campuses in parallel without carrying the full debt load on their own books.

Why it matters

At roughly $14 billion for 1 gigawatt of capacity, the deal is a concrete, recent data point on what frontier-scale AI compute now costs to build — useful context for sizing up the wave of multi-gigawatt announcements coming out of every major AI infrastructure buyer this year. It also marks another step in institutional capital treating AI data centers as a distinct, investable infrastructure asset class: BlackRock, managing trillions in assets globally, is putting real equity and debt structuring behind AI compute the same way it would a toll road or utility, a bet that the demand for AI training and inference capacity will be durable enough to justify multi-decade infrastructure-style returns.