Data-center supply chain takes 10M SF in DFW

Dallas-Fort Worth's industrial market just found a tenant it didn't have to chase: the data-center supply chain. PWD's tracking counts 10 million square feet of logistics space occupied by suppliers and contractors supporting data-center development since 2025, more than any other major U.S. market—not a rounding error in a market that measures industrial take-up in the tens of millions of square feet a year, but a new demand channel arriving just as traditional distribution leasing cools and changing what a warehouse landlord owns.

The number matters because of what it excludes: data-center operators signing long-term ground leases for powered shells or buying land near transmission lines—the core data-center trade with its own capital stack, zoning fights, and queue for transformers and switchgear. This is the layer behind it, the contractors, equipment handlers, fabricators, and logistics providers that build and outfit those facilities, taking conventional warehouse space in the same industrial parks that have served consumer goods and auto parts. They likely pay warehouse rents rather than data-center rents and need no substation capacity, which for a landlord is the cleanest kind of data-center exposure there is.

Two years ago, Dallas-Fort Worth's industrial market was still a distribution story—e-commerce, third-party logistics, and retail restocking—where big blocks were measured by clear height, trailer parking, and door counts; now the same buildings are measured by proximity to a different kind of construction schedule, because a supplier staging equipment for a data-center campus doesn't need a cross-dock near an intermodal yard, just space within a short drive of the job site on a lease that can flex with the project. That shift is precisely what the 10 million square feet captures.

The 10 million square feet is the visible tip of a re-rating: a warehouse landlord that was underwriting a generic distribution tenant two years ago is now, without changing the asset, underwriting a tenant whose revenue depends on power-generation buildout. The lease still reads like a warehouse lease, but the tenant's balance sheet now tracks data-center capex—a second-order correlation that moves a market when it scales.

The contrast with Milwaukee makes the DFW figure sharper: CoStar counts 8.7 million square feet of signed industrial space in Milwaukee in the past year, a gain of 13 percent over the prior year, and that healthy figure is the whole market, whereas Dallas-Fort Worth's 10 million square feet comes from one supply-chain segment alone. When the data-center buildout reaches the point that its contractors occupy more logistics space than all tenants in Milwaukee, the cycle has moved from the power grid into the warehouse.

The mechanics are worth being precise about, because they explain why this is not just demand substitution: CoStar's Milwaukee tally is leases signed, while the 10 million square feet in Dallas-Fort Worth is occupied space since 2025 by a defined tenant class. The two aren't directly comparable in vintage but are in direction—one market fills buildings with whoever signs; the other fills them with a capital cycle that has a multi-year runway, which is what the warehouse landlord is actually buying.

Strip out the data-center operator and the trade becomes clearer: the operator takes the land, the power, the zoning risk, the construction risk, and the equipment procurement risk, while the warehouse landlord takes none of it. It owns a conventional warehouse with loading doors and a roof, leases it to the company that supplies the operator, and gets derivative but cheap exposure to the boom—no need to win a capacity auction or wait three years for a transformer, only to be within the logistics radius of a project that has already been financed. That is the position the DFW market has been accumulating since 2025.

The risk is not the data-center boom ending but its becoming too successful for the warehouse landlord. If data-center demand pushes industrial land prices to the point where a warehouse's residual value is really as a future data-center site, the landlord ceases to be a second-order play and becomes a land bank, and the trade shifts from leasing to assemblage. The 10 million square feet of occupancy suggests the market is still in the leasing phase, where net operating income is driven by tenants rather than land re-zoning—an earlier, less speculative stage that has historically been most generous to existing owners.

The aggregate number does not say which submarkets are doing the work or name the landlords capturing the leases, because the tracking has the market-level figure rather than the tenant roster. That absence suggests the absorption is dispersed enough that no single institutional owner has been able to market it as a strategy; the trade is being built quietly, lease by lease, across the same parks originally built for retail distribution. The day a major landlord starts breaking out data-center supply-chain occupancy as a percentage of net operating income is the day the market has named the trade.

The case for Dallas-Fort Worth as the first major market to show this pattern is simpler than a ranking of data-center pipelines: one clean number, 10 million square feet occupied by one supply-chain segment since 2025, more than any other major market. If another major market were absorbing its data-center supply chain at the same rate, it would show up in the same tracking; it hasn't. The absence of a rival data point is not proof, but combined with the scale it is enough to call DFW the leading indicator.

What this means for cap rates and valuations is less obvious than the bull case suggests: a warehouse lease to a data-center contractor is not a data-center lease and is unlikely to command a data-center multiple. The tenant's credit may be tied to a single project or a single hyperscaler's build schedule, which means the lease term could be shorter and the recovery profile lumpier than a 10-year deal with a national third-party logistics provider. Underwriting that revenue as if it were stable distribution income would be a mistake; underwriting it as zero would be a bigger one, because the demand is real and concentrated. The right answer is somewhere in the middle—a moderate uplift to market rent, a modest discount to the headline data-center trade, and a hard look at the tenant's contract backlog.

The Milwaukee figure supplies the other half of the contrast: there, 8.7 million square feet of signed leases have not yet reached rents, meaning tenants are signing but landlords are not yet pricing the gain, while in Dallas-Fort Worth the 10 million square feet has already moved into occupied space and passed through the leasing pipeline. The next step is renewal or expansion, where rents get set; a market that absorbs first and prices later is, in the near term, a better place to own than a market that signs first and waits, and DFW has the former profile where Milwaukee has the latter.

There is a version of this story where the 10 million square feet is a one-time pull-forward—a surge of construction that burns through local warehouse vacancy and then leaves behind short-term leases that roll in 2028. The way to test that is to watch the renewal rate and lease term: five-year deals with options mean recurring occupancy; two-year deals tied to a single campus mean the landlord owns a construction cycle, not a tenant base. The source data doesn't yet break that out, and that missing variable is the difference between a secular trade and a cyclical one—the market seems to be pricing the secular outcome already, but it has not resolved the question.

The second-order trade has one advantage over the direct data-center trade: it is easier to exit. A warehouse occupied by a data-center contractor can be sold to a conventional industrial buyer if the boom stalls, because the asset doesn't have the specialized power infrastructure that would make it a stranded cost. That optionality means the downside is not a binary bet on electrification—the landlord gets a call option on data-center demand without the put risk of owning a powered shell, an asymmetry that makes the 10 million square feet matter more than the square footage itself.

For industrial owners outside DFW, the takeaway is less to buy DFW warehouses or convert buildings into supplier parks than to find the next market where the same two ingredients are present: enough data-center construction to create a supplier base, and enough existing warehouse stock to absorb that base without forcing it into purpose-built facilities. The first ingredient is visible in half a dozen markets; the second is the binding constraint. Where both exist, the 10 million square foot pattern is likely to repeat; where one is missing, the supply chain will leak across state lines and the landlord will never get the lease.

The warehouse landlord who can distinguish a switchgear supplier from a furniture distributor has a more valuable underwriting model than one who cannot, and the 10 million square feet since 2025 is the first real test of it. The answer will come at the first renewal cycle, when the market learns whether it owns a tenant base or a construction cycle.