Power, not land, decides data-center winners

Texas has 474 gigawatts of projects waiting to connect to the grid, and a new state audit has just attached a price to jumping the queue—the single most important number in data-center site selection. The interconnection backlog has moved from an engineering constraint to a capital-allocation problem, so the underwriting question for every developer, lender, and offtaker is no longer where the land is but where the substation is.

The 474-gigawatt queue measures how far supply has run ahead of the grid's ability to absorb it, and the state audit that landed this month is Texas's attempt to make that backlog liquid—to assign a cost to holding a queue position itself. Details are still being digested, but the direction is clear: the audit is expected to penalize speculators who hold interconnect slots without building and to reward projects with signed offtake agreements and real generation behind them. In effect, the audit prices patience, raising the carrying cost for developers who sat on queue positions as a land-banking strategy while giving the strongest hands to sponsors with contracted power and a committed offtaker.

That pricing is the binding underwriting constraint for the $73 billion construction pipeline PWD tracks across the data-center sector, and the capital chasing that pipeline is no longer betting on land or steel but on substations. A sponsor who secures substation capacity before the site has a permitting date holds a real asset; one who secures land without the power allocation holds a liability, and the differential will show up in land prices, in joint-venture structures, and in lenders' willingness to advance capital on projects without a firm interconnection position.

The clearest evidence that the market's most sophisticated buyers have already reached this conclusion came on Aug. 31, when Amazon announced a multi-party renewable-energy agreement with Eolus, OX2, Mirova, and Dala Vind that secures generation directly, bypassing the queue. It is the hyperscaler version of vertical integration: rather than wait for the grid's schedule, Amazon is contracting with the generators themselves, in effect bringing its own power to the meter. For the rest of the market, that is both a template and a warning—the template is the contract structure, and the warning is that a hyperscaler of Amazon's scale no longer relies on the grid's schedule. Every data-center developer without a similar agreement is now competing for a shrinking pool of accessible interconnection capacity.

The underwriting risk has migrated from the water table to the meter—at Amazon's Shreveport campus, the question was whether the local aquifer would serve the cooling load; a year later, it is whether the interconnection queue will carry the electrons. The first was a physical limit; the second is an administrative one, which means it can be priced, traded, and arbitraged, and because it can be arbitraged it will attract capital. Private credit funds that once wrote loans against leasing velocity are now learning the vocabulary of interconnection agreements, while institutional investors that once underwrote data centers on power purchase agreements are now underwriting the queue positions themselves.

The audit's timing is no accident: it lands just as the $73 billion pipeline is moving from development plans to construction starts and lenders are being asked to fund projects still waiting on grid connections. The collateral value of a data center is no longer the building but the power contract, so the due diligence list has changed—site plans are now measured against substation capacity, and land options are being written with contingencies keyed to interconnection milestones.

If the audit works as intended, the queue position itself becomes a tradeable asset, and a secondary market in interconnection slots will develop—much like the market in Scottish salmon farms or New York taxi medallions. That will draw infrastructure funds and credit funds comfortable pricing optionality, a new class of investors alongside the traditional real estate crowd, and they will be the natural counterparties for developers who hold more queue positions than they can build out. The spread between a site with a queue position and one without it will become the market's clearest measure of the cost of power allocation.

For the week ahead, the data-center market will be digesting the audit's implementation timeline and the next wave of power purchase agreements. Watch for the first land deals that trade on the new pricing and for the first developer retreats from positions that no longer clear their hurdle. The developers who deliver the $73 billion pipeline will be the ones who treat the interconnection queue as part of the site plan; the rest are holding land without a route to the grid.