Decommissioning a nuclear power plant

Decommissioning, in the US Nuclear Regulatory Commission's definition, means taking a facility out of service and reducing residual radioactivity to a level that allows the property to be released and the licence terminated, either for unrestricted use or under restricted conditions. It is not demolition: it is a regulated process with a numerical dose target, and knocking down the conventional buildings is only its visible part. The NRC allows sixty years from the end of operation to finish the job, and requires the licensee to have set up a decommissioning fund beforehand, a trust or a parent-company guarantee, because the bill arrives once the plant no longer sells electricity. The NRC puts the typical cost of a single unit at $300 million to $400 million.

Within that deadline there are three routes. DECON dismantles straight away: soon after closure the activated equipment and structures are decontaminated or removed, and the site is released within a decade or so. SAFSTOR does the opposite and defers: the plant is kept under care and surveillance for decades while the radioactivity decays on its own, and only then is it taken apart, with lower worker dose and less intermediate-level waste, in exchange for paying for maintenance and monitoring the whole time. The third one, ENTOMB, seals the activated material in place inside concrete; the NRC provides for it, but no US power reactor has used it and it is not available for today's large units. Choosing between DECON and SAFSTOR is a question of money and dose, not of safety: both end in the same place.

The point that usually gets lost is that finishing decommissioning is not the same as getting the land back. Maine Yankee completed its work in 2005, Connecticut Yankee and Yankee Rowe in 2007; all three released most of the property, and all three are still licensed nuclear sites, because the spent fuel never left. What remains at each is an ISFSI, the dry cask pad: a concrete slab with the casks, its fence, its security force and its licence, waiting on a federal destination that does not exist. The licence ends when that ISFSI is dismantled too, and that depends on a decision the operator does not control. It is the difference between dismantling the plant and closing the file.

Water is the other front, and Pilgrim, in Plymouth, Massachusetts, shows the whole shape of it. The plant shut down in May 2019 and Holtec, which bought it to decommission it, proposed discharging up to 1.1 million gallons of treated water into Cape Cod Bay. The obstacle was not radiological. The Massachusetts Department of Environmental Protection denied the permit in July 2024 and rejected the appeal in September 2026, holding that the release is not grandfathered under the Ocean Sanctuaries Act, the state law that bars industrial waste from an ocean sanctuary. Holtec argues the plant made equivalent discharges for more than forty-five years of operation, regulated by the NRC. Both things are true at once: what changed at shutdown was not the water, it was the legal category it is read under.

Behind that argument sits a hard physical limit. The isotope left in the water of a light-water plant is tritium, and tritium becomes part of the water molecule itself. In the NRC's words, tritiated water “is chemically identical to normal water and the tritium cannot be filtered out of the water”: not filters, not ion-exchange resins, not reverse osmosis, because there is nothing to separate. That leaves three options and no others: dilute and discharge under permit, evaporate, or hold the water until it decays, with a half-life of 12.3 years. When an operator pushes to discharge, it is because the other two cost time and money, not because filtering never occurred to anyone.

Europe runs the largest job of this kind, and it is useful for calibrating timescales. The Greifswald plant, at Lubmin, closed in 1990 with German reunification and is still being taken apart: some 1.8 million tonnes of material to sort, a cost estimated in the billions of euros, and work scheduled into the 2060s, roughly seventy years after closure. Its 5,037 irradiated fuel assemblies went into CASTOR casks in an interim store on the site itself, for the same reason as in the United States: Germany has no final repository either. Sixty or seventy years is the real order of magnitude of a decommissioning, and the fuel outlives the process.