The consensus framing for AI infrastructure has been power: gigawatts, PPAs, grid strain. The under-covered angle is real estate. Building data centers consumes enormous quantities of staged industrial space — for switchgear, generators, cooling, cabling, and contractor fit-out — and that demand is now the single largest category in new manufacturing leasing.

For industrial owners, this rewires the demand map. The strongest sub-markets are no longer only e-commerce nodes; they are the logistics rings around hyperscale data-center campuses. Absorption in those corridors is partly a function of someone else’s AI capex budget.

Underwrite industrial near data-center clusters to a different demand curve — one tied to AI build cycles, which are lumpy and capex-driven. The upside is real; the risk is concentration. When the buildout pauses, that incremental 27% is the first demand to soften.

➤ TAKEAWAY

The AI boom doesn’t just need power — it needs warehouses, and it’s now the biggest one leasing them. For owners and allocators, the most defensible basis is shifting from proximity to ports toward proximity to compute. Buildings that will never sit in a data-center fund are now underwritten, in part, to a hyperscaler’s capex calendar — demand that pays like a tenant but moves like a balance-sheet decision made three states away. The edge belongs to whoever prices that exposure before it shows up in the comps.

Source: CLink Logistics, CoStar (Prologis), CBRE — Q1 2026 data, recirculated June 2026