Summary
- Meta and BlackRock have created a $14bn strategic venture for a 1GW AI data centre campus in El Paso.
- BlackRock-managed funds will own 80 percent of the venture, with Meta retaining 20 percent and leasing capacity.
- The structure shows hyperscale AI campuses moving closer to infrastructure-style ownership, debt, and long-term contracted capacity.
Meta and BlackRock have formed a $14bn strategic venture to develop and own a 1GW AI data centre campus in El Paso, Texas, giving BlackRock-managed funds a majority position in a major hyperscale asset.
BlackRock-managed funds will own 80 percent of the venture, while Meta will retain 20 percent. Meta is contributing land and construction assets, will be the initial sole occupant, and will lease the facility under agreements that could run for up to 20 years.
The campus is already under construction and is expected to begin coming online in 2028. Meta said the project will support its compute infrastructure and create thousands of construction jobs, with a smaller permanent operational workforce once live.
Hyperscale capacity adopts infrastructure finance
The structure points to the capital intensity of AI-era data centres. A 1GW campus requires land, substations, grid infrastructure, power procurement, backup systems, cooling plant, buildings, network capacity, controls, security, and long-term operations. Even the largest technology companies are looking for ways to share ownership and bring institutional capital into the physical layer of compute.
The El Paso venture resembles other infrastructure models where long-term contracted use supports outside investment. Meta secures access to capacity without owning the entire asset directly. BlackRock-backed investors gain exposure to a large digital infrastructure platform anchored by a creditworthy tenant.
Debt is also central to the model. Large AI campuses are beginning to look less like corporate real estate projects and more like power-backed infrastructure assets, where lenders and investors assess tenant strength, lease terms, energy costs, residual value, and the credibility of the construction and utility programme.
The 1GW scale is a reminder that a data centre campus of this kind is also an energy project. Grid access, transmission reinforcement, procurement of firm power, cooling design, water strategy, generator provision, and local infrastructure interfaces all sit behind the financing.
Europe will test the model harder
European data centre markets already use joint ventures, platform investments, sale-and-leaseback structures, and infrastructure funds. The Meta-BlackRock transaction shows how those models could grow as AI capacity becomes larger, denser, and more expensive to deliver.
The European version will face tighter constraints in many locations. Planning processes are often more restrictive, power connections can be slower, land is more politically contested, and water or heat-use conditions are becoming more visible. A financeable tenant does not remove those risks.
Institutional capital can accelerate campus delivery where power, planning, and construction routes are credible. It can also magnify exposure if capacity is financed before the grid path is firm enough. Debt-backed campus structures depend on predictable energisation, stable energy costs, and long lease commitments that survive changes in AI hardware, efficiency, and demand.
The ownership split also raises questions about asset flexibility. A campus designed around one hyperscale customer may have a strong initial contract but limited alternative use if density, cooling, network design, or control systems are highly specialised. Residual value matters more when capital structures become larger and longer-dated.
Meta and BlackRock have provided a clear template for financing enormous AI capacity. European developers and investors will not be able to copy it mechanically. They will need to match the structure with firm electricity, planning certainty, cooling resilience, and a credible route through public scrutiny.
