The Retired Power Plant Is the Shortcut, But Only If You Site It Right!
Every serious conversation about powering data centers now runs into the same wall: interconnection. It is the single hardest thing to get on the American grid, and it has become the binding constraint on the entire build-out.
The numbers tell the story. Data centers are forecast to need on the order of 80 gigawatts of new power in the U.S. by the mid-2030s — roughly tripling today's load. The grid they want to plug into is already gridlocked: more than 2,600 gigawatts of capacity sits waiting in interconnection queues, over twice everything currently connected. A large project now waits four to eight years to get through. Only about 14% - 20% of projects that enter the queue reach commercial operation; roughly 70% withdraw along the way. For a data center racing a competitor to stand up compute, that timeline isn't a delay. It's the whole game.
This is why developers are circling retired and retiring power plants. A decommissioned plant already owns the thing the queue rations: a large, permitted, high-capacity connection to the transmission system, built over years, sitting idle. Reuse that connection and you skip the line that kills most projects. That is a genuine, quantifiable advantage, and it is the right instinct.
The pool is large, but the plants are not interchangeable
The good news is that the inventory is deep. Since 2011, the U.S. has retired about 145 gigawatts of coal capacity, down from a peak of 318 gigawatts — and another 11 to 12 gigawatts of fossil capacity is retiring every year, split across coal, gas, and oil. In 2025 alone, planned retirements ran roughly 8 gigawatts of coal, 2.6 of natural gas, and 1.6 of petroleum. One industry analysis puts the retired coal capacity realistically convertible to serve new load at around 70 gigawatts — close to the scale of the demand coming at the grid.
But here is where the "just reuse the interconnection" story quietly breaks, and where the real work begins. That 145-plus gigawatts is not a stack of interchangeable sites. A mothballed coal unit, a still-dispatchable gas peaker, and a shuttered oil plant each hand a developer a completely different starting position — different interconnection rights, different reconstruction cost, different permitting exposure, different community risk. Pick the wrong one and the queue time you thought you saved comes back, this time as a demolition schedule, a failed air permit, or a four-year zoning fight. The advantage is real; capturing it depends entirely on which plant you choose, and that is a question of due diligence, not luck.
Two screens separate a genuine head start from a multi-year trap.
Screen 1: What's still running, and why a live peaker beats a dead plant
Not every candidate is a cold, decommissioned site, and the difference is larger than most developers assume. A gas peaker that is still commissioned and ready to dispatch is often the strongest site on the board, because almost nothing has to be rebuilt. The generation is already there, already permitted to run, already connected and injecting to the grid.
That matters in three compounding ways.
It powers the data center now, and flexibly. A data center load that can flex — ramping down or shifting during grid stress — turns a live peaker into a dual-purpose asset: it powers compute when demand is high and stays available to support the grid when the grid is tight. You get speed to power without waiting on new generation.
It carries live interconnection rights, not just idle wires. A running peaker holds an active Large Generator Interconnection Agreement. And because peakers run at very low capacity factors, that agreement is usually far larger than the plant's real-world use — which is exactly the headroom a new resource can tap through Surplus Interconnection Service, a FERC mechanism that lets a new project connect at an existing plant's point using its spare capacity. The savings are not marginal. Surplus interconnection reviews typically finish in under a year, versus the four-to-eight-year main queue; in one documented Kansas case, a surplus-interconnection solar project faced interconnection costs of $0.71 per kilowatt against $333 per kilowatt for a comparable standard project nearby — a 99.8% reduction. MISO and SPP alone are now studying well over 15 gigawatts of these surplus requests.
It has the fastest credible path to clean. This is the argument that should carry the most weight, and it is the one usually skipped. A live peaker site is the single best place to phase in solar and storage, precisely because of that surplus interconnection headroom and the shared substation, transformers, and land already on site. Battery storage is already replacing gas peakers directly — the duty cycle matches, and the existing industrial permitting and grid hardware transfer over. So the live peaker is not a bet on gas; it is a bridge that can be walked down to clean power far faster than a greenfield build, because the connection the clean resource needs is already in the ground. A dead coal plant offers none of this: the interconnection may be lapsed or derated, the site needs demolition and remediation before anything is built, and the clean-energy path starts from zero.
The screen, in one line: favor a plant that is still connected and dispatchable over one that is merely large and dead, and value most the site that gets you to clean power fastest — not the one with the biggest nameplate on a plaque.
Screen 2: The site was permitted for a world that no longer exists
The second screen is the one most likely to cost years, and the one a spreadsheet of megawatts will never surface. Retired plants were sited, zoned, and permitted decades ago, under the orders and standards of their era. The surroundings changed; the old entitlements did not. Before committing, a developer has to know not just what the site was permitted for, but what has grown up around it since — and whether a modern data center clears the current permitting, zoning, and environmental bar.
Critically, that check has two dimensions, and screening for only one means walking into the other.
Residential encroachment. An industrial site that stood alone when it was built may now sit among homes. Markley Group's Lowell, Massachusetts data center occupies a former pasta factory but the surrounding neighborhoods are densely residential, with homes in some cases less than 100 feet from the facility, a public ballfield directly behind it, and an early-learning center nearby. As the company moved to expand, that proximity became the fight: the state's first municipal data center moratorium, in March 2026, and resident-led litigation over the plant's diesel generators and air permits. No amount of interconnection value offsets a conflict like that.
Resource sensitivity. Even a site with few immediate neighbors can sit beside a treasured shared resource that mobilizes regional opposition. The Greenidge plant on Seneca Lake, in the small New York town of Torrey, would pass a "surrounded by homes" test far more easily than Lowell. Yet it spent four years in conflict — not over a fence line, but over the lake: thermal discharge, coal-ash handling, and emissions, culminating in a 2022 air-permit denial, a 2024 court ruling, and a 2025 settlement imposing binding emissions cuts. The siting liability was the water body, not the neighbors.
The screen a single-axis analysis misses: a real assessment asks both who lives here now? and what sensitive resource does this plant touch, and who speaks for it? — against the rules as they stand today, not as they stood when the plant was first permitted.
The point: the connection is the beginning of the work, not the end
Reusing a retired plant is a real advantage on the one problem that matters most where the grid is constrained — time to power. But the inherited interconnection is the start of the project, not a finished site. The developers who treat it that way — who screen the fleet for the right power source and the right land-use fit before the money is committed — capture the years of saved queue time, avoid the demolition and remediation costs of the wrong site, and sidestep the community and environmental fights that come from a plant permitted for a vanished world. The ones who mistake an old interconnection for a finished project meet the four-year timeline anyway.
That plant-by-plant assessment — which site, which power source, which zoning and environmental exposure — is precisely the analysis our work at EcoDe Carb is built around. If you're weighing a decommissioned site for a data center or large load, it's worth running these screens before you sign.