The factory side of the AI real estate trade

Ten to 30 percent of recent leasing across Hines' U.S. industrial portfolio now comes from component makers, infrastructure suppliers, and advanced manufacturers building the supply chain for the data center buildout. Three years ago, many of these names barely registered as industrial tenants, Hines says in a commentary published by IREI; the firm calls them "New Companies Doing New Things," a label that applies especially to advanced manufacturing.

The commentary arrives as the AI buildout captivates commercial real estate, with conference agendas heavy on power, hyperscalers, and compute capacity; the enthusiasm is deserved, since the infrastructure opportunity is one of the most compelling in decades. But the buildings housing the AI buildout are only part of the picture: the rest is the factory base feeding them, a new class of occupiers forming before they become obvious tenants.

That sequence is the data center capital cycle expressed one step earlier in the value chain: power and water are the gating constraints on AI infrastructure, and the tenants Hines describes are the industrial response to that constraint. Investors underwriting industrial should read the tenant roster as a leading indicator rather than a fixed assumption, since a warehouse that looks like a standard distribution box today may already be a critical link in the AI supply chain.

The trend also blurs the conventional line between warehouse and manufacturing, because a building leased to an AI supply-chain component maker functions closer to light manufacturing than to pure logistics, with consequences for tenancy length, fit-out costs, and exit assumptions. Industrial investors who treat every warehouse tenant as fungible distribution demand will misprice that.

The 10 to 30 percent band is wide, and Hines' portfolio is a single institutional owner's sample, not a national statistic, but a tenant class that barely registered three years ago has become a double-digit share of deal flow. RIAs and institutional allocators should segment industrial demand beyond e-commerce and traditional logistics: at a double-digit share, the AI supply chain is already too large to be a subplot, and underwriting models that treat it as one are pricing the last cycle.

The sharper call is durability: data centers are point assets, built, powered, and left in place, while the supply chain feeding them rebuilds with each technology cycle, which makes the factory trade the more recurring real estate exposure in the AI story. The next industrial capital cycle may belong less to the speculative box than to the specialized manufacturing and warehouse infrastructure that supplies the buildout—one building type over from the one everyone is watching.