Co-Locating Data Centers With Power Plants

The idea is simple. A power plant already has generation, a transmission interconnection, a switchyard and usually water and fuel infrastructure. Put a data center next to it, and the campus can draw power without waiting for new transmission to be built across many miles.

In practice, co-location covers several different arrangements with different technical, commercial and regulatory consequences. The first task in evaluating a co-location opportunity is to pin down which one is actually on the table.

The configurations are not equally available everywhere. In vertically integrated states the local utility often owns or controls both the plant and the retail service, which shapes what is possible. In restructured markets with independent generators, a direct deal with a plant owner is more common, but the grid operator and state regulators still have a say. Our guide to behind-the-meter vs. front-of-the-meter power explains the metering distinction in more depth.

A co-located campus inherits the plant’s strengths and weaknesses. Diligence on the plant is as important as diligence on the land.

• Availability and outages. Every plant has planned maintenance and unplanned outages. Thermal plants have periodic major overhauls, and nuclear units have refueling outages. Know how the campus is served during each.
• Unit configuration. A plant with several units can keep serving part of a load when one unit is down. A single-unit plant cannot.
• Remaining life and retirement plans. A plant that may retire within the data center’s planning horizon is a poor anchor unless replacement generation is part of the plan.
• Fuel and water. Gas supply firmness, coal or fuel logistics, and cooling water rights all carry over to the data center’s reliability.
• Environmental permits. Changes in how a plant operates, such as higher run hours, can trigger air permit review.
• Ownership and contracts. Existing offtake contracts, capacity obligations and lender consents can limit how much output is available to a new load.

Plant sites are often large, but much of the acreage is spoken for. Switchyards, transmission corridors, fuel handling areas, ash or water ponds, buffer zones and land the owner holds for future units can all reduce what is available. Nuclear sites add security perimeters and exclusion areas, which our guide to nuclear and SMR siting covers.

• Confirm the boundary of the parcel offered and whether it is owned, leased or subject to the plant’s easements.
• Map transmission rights-of-way, pipelines and water lines that cross or border the land.
• Review environmental history, particularly at older thermal plants with ash or fuel storage. A Phase I ESA is a baseline, and retired or active plant land often needs more.
• Check zoning. Plant land may be zoned for utility or heavy industrial use that does or does not allow a data center.
• Look at access, flood exposure and room for the data center’s own substation, generators and cooling equipment.

Retired plant sites are a related but different opportunity. They may offer an existing interconnection and industrial zoning without the live-plant constraints, which our guide to repurposing industrial and power plant sites explores.

Regulatory treatment is the least settled part of co-location. Several questions come up repeatedly, and the answers depend on the state, the utility, the ISO or RTO and the specific configuration.

• Whether a grid-connected co-located load is using, and should pay for, transmission service and other grid costs.
• How removing a plant’s output from the market affects resource adequacy and reliability for other customers.
• Whether the arrangement is a retail sale subject to state utility law, and who has the right to serve that load.
• What interconnection and metering changes are needed at the plant, and whether they trigger new studies.
• How the load must behave during grid emergencies, and whether it can be curtailed.

FERC, state utility commissions, grid operators and NERC reliability standards can all touch these questions, and policy in this area has been evolving. A co-location plan should be reviewed with regulatory counsel in the specific market before the deal terms are fixed, and revisited as rules change during development.

Co-location tends to fit when the plant has multiple units and a long remaining life, the land is genuinely available and buildable, and a grid backup path exists or can be arranged. It fits less well when the plant is near retirement, the site is boxed in, or the regulatory treatment in that market is unclear and the project cannot absorb a change.

It also helps to compare co-location against a standard large-load request at the same location. Sometimes the plant’s switchyard makes ordinary front-of-the-meter service fast and simple, and the extra complexity of a behind-the-meter deal is not worth it. When we screen a plant-adjacent site, we evaluate both paths side by side.