Data center capital splits: cheap shells, pricier ground

Colovore's Hutto, Texas, data center broke ground this week with a construction budget of $280 million, a building it had once priced at $500 million. The missing $220 million moved into the land under the slab, the grid connection that makes the building live. That single project captures a split now running through data-center capital: the utility shell is deflating while energized land inflates, one side commoditized down and the other bid up like a scarce resource.

Bitdeer's $100 million purchase of Alcoa's land works out to $500,000 an acre, but what the price actually bought was a right to draw grid capacity, secure water, and scale an AI cloud operation to 742 megawatts—an option premium on power that has not yet been contracted. GLP's RMB4 billion Foshan campus carries the same logic at a different scale: total potential of 200MW, with just 40MW committed to a single unnamed hyperscaler and the remaining capacity priced as an undeveloped option before a second tenant signs. In each case capital is advancing for the right to energize land, and the data center building is only the vehicle that turns that right into revenue.

Hutto cuts the other way. A construction budget that fell from $500 million to $280 million—a 44 percent drop—says the market has stopped paying for white space: the shell, the cooling loops, the security perimeter are now a cost to be minimized rather than a value to be accumulated. No design simplification explains a cut that large; it is a repricing of what actually appreciates, which is the site and specifically its position in the grid queue. The brick-and-mortar data center is a claim on that position, not the position itself.

The IIF's new analysis of site selection backs the trade, ranking power, land, and proximity ahead of tax abatements in the location decision. That demotion matters because abatements are the traditional battleground where municipalities compete for data-center projects; a tax deal is a negotiated cash-flow feature, while a substation is an existential condition. Utilities do not compete the way economic development offices do, and the finite capacity they control is increasingly spoken for. Incentives can close a gap at the margin, but they cannot create the connection that makes a site viable.

Underwriting in the next phase will resemble an options trade more than construction finance: energized land is the underlying asset, the grid application is the option, and the data center itself is the exercise. A developer who underwrites only the $280 million shell is pricing the vehicle while ignoring the asset. The landowner who holds grid access is selling a call on future power, and Bitdeer's $500,000 an acre is the first market-clearing price for how valuable that call has become.

For investors underwriting data-center funds, the underwriting question now centers on how many acres of energized land a sponsor controls and how far along in the interconnection queue those acres sit, rather than on how many megawatts it plans to build. Sponsors that own land with negotiated power and water can afford to build cheap shells; sponsors that buy shells without the grid position are buying a depreciating asset whose one scarce input is missing.

Bitdeer is emerging as a power company with an AI workload attached, while GLP is a logistics operator redeploying its balance sheet into a utility option with a data center attached. Both trades invert the old real estate logic in which the building was the permanent asset and the land a passive base. In today's data-center market the building is the depreciating wrapper and the land the appreciating core.

The Hutto groundbreaking marks where value now sits: the $220 million removed from the building budget now sits on the landowner's balance sheet. A developer paying $500,000 an acre for land and $280 million for a shell is writing a large check to the landowner and a smaller check to the contractor. The next data center budget will show whether 44 percent has become the market's starting assumption.