Data-center land values now follow the power calendar

In the 2024-2025 delivery year, PJM capacity prices sat at $28.92 per megawatt-day; by 2025-2026 they had reached $269.92 and by 2026-2027 $329.17, a steepening that Commercial Observer's Sept. 7 report on data-center development reads as the shift that now orders everything else in the sector. Power has moved from an input in the data-center model to the product around which land, fiber, approvals and tenants get assembled.

Under that logic, a clean parcel with land, fiber, approvals and a committed tenant can be nearly worthless if electricity does not arrive on the tenant's timetable, which is why the old site-selection checklist has inverted: substation location, available capacity and the date the first block of power can be energized now rank above acreage and entitlement status. The deed matters less than the delivery schedule.

That split shows up most clearly in phasing, since a hyperscaler may need hundreds of megawatts eventually but not on day one; a site that can deliver 50 megawatts now and another 100 megawatts eighteen months later can be worth more than one promising 300 megawatts at some later point. In site-selection terms, the asset is capacity on a calendar, and the map comes second.

That calendar is pulling developers into energy work they used to leave to utilities, so utility relationships, transmission access, substations and gas connections now receive the scrutiny once reserved for land assembly and entitlements, according to the report. The underlying tension is a collision of timelines: AI tenants plan deployments in months, while transmission and generation additions take years. Behind-the-meter generation, cogeneration and direct energy partnerships fill that gap as bridge mechanisms.

The capital-stack reading follows directly: if scheduled power is the asset, then the contracts that fix a power delivery date to a parcel—an interconnection commitment, a transmission access agreement, a supply contract—are where equity value forms. A parcel without one of those is a bet that wires will eventually reach it, and buyers underwriting on that basis are speculating on utility construction schedules.

Commercial Observer names the layer digital dirt: former manufacturing centers, industrial corridors and legacy energy sites whose substations, transmission infrastructure, gas connections and power allocations are underused. The buildings on those sites may be obsolete even while the infrastructure underneath them is not, and that inherited stack is difficult and expensive to reproduce.

For private-market buyers, that distinction is more useful than the old core-versus-edge geography: powered land describes access to electricity, while digital dirt describes the entire inherited system that makes that electricity relatively easy to restore or expand. An old industrial site with a live substation and an established power allocation carries a different risk profile than greenfield next to a fiber line, even if the building on the old site is a teardown.

This publication has argued that the data-center trade has left the building shell and moved into the energy system around it; PJM's capacity curve now turns that argument into valuations. Usable square feet and zoning matter less than the ability to put power in service by a date certain, so buyers underwriting land on a price-per-acre basis are pricing the wrong thing. The scarce commodity is the megawatt-day secured by a delivery position, and sponsors who have one are buying the project while those who do not are buying a parcel with an indefinite wait.