BasicsConcernsA wet-cooled 300 MWe reactor running at full output would consume roughly 4.8 million gallons of water a day. That figure is this article’s own arithmetic, built on the generic factor of about 670 gallons per megawatt-hour for nuclear plants with cooling towers, which comes from NREL’s operational water-use factors. It is a ceiling for one cooling choice, not a forecast for any project. Developers are now proposing air cooling, small intakes, and sodium-cooled reactors with steam loops, and the water answer changes with each.
This guide is for the person who lives near a proposed site and wants numbers to ask for, not adjectives. It covers what the terms mean, what four named projects have disclosed, what dry cooling costs, and what drought has done to river-cooled plants this summer.
Withdrawal and consumption are different numbers
Withdrawal is the water a plant takes from a river, lake or aquifer. Consumption is the part that evaporates and does not come back. NREL’s researchers use both measures, and the gap between them is large. The University of Arizona’s Water Resources Research Center, summarizing the same NREL data, puts withdrawals at tens of thousands of gallons per MWh for nuclear and coal plants on once-through cooling, and at hundreds of gallons for plants with recirculating towers. Once-through plants return most of what they take, warmer than they took it. Tower plants take far less and lose most of it to the sky.
Neither number is the whole story. A plant that withdraws little but consumes nearly all of it can strain a small aquifer. A plant that withdraws a lot and returns it can harm fish if the discharge runs too warm. When a developer says a project “uses little water,” ask which of the two it means. The answer should arrive as a figure in gallons per day, with the source named.
What size changes and what it does not
A 300 MWe plant is smaller than a 1,000 MWe plant, so its total draw is smaller. Per megawatt-hour, size does not help on its own. Water use tracks the cooling system and the thermal efficiency of the cycle. A WNA-sourced explainer in Nuclear Engineering International notes that a plant at 33 percent efficiency discharges about 14 percent more heat than one at 36 percent. Several small water-cooled designs run at the lower end of that range.
So the claim that SMRs use less water per unit of power has no support in the figures I found. The defensible claim is narrower: a smaller plant draws a smaller absolute volume, and some designs avoid water cooling altogether.
Four projects, four cooling answers
The projects below are at different stages, and the stage matters as much as the design.
TVA Clinch River, GE Vernova Hitachi BWRX-300 (300 MWe, water-cooled): The NRC issued a construction permit on 29 September 2026. Construction has not been announced as started, and no operating licence exists.
TerraPower Natrium, Kemmerer, Wyoming (345 MWe sodium-cooled fast reactor, 840 MWt, with molten salt storage to 500 MWe): The NRC approved a construction permit in March 2026. An operating licence is still required, with completion targeted for 2030.
Dow and X-energy Xe-100, Seadrift, Texas (four 80 MWe high-temperature gas-cooled units, 320 MWe): The NRC completed its environmental assessment on 18 May 2026. A construction permit had not been issued in the sources I reviewed.
UAMPS and NuScale, Idaho (six 77 MWe modules, 462 MWe, air-cooled condenser): The project was terminated in November 2023 and never reached construction.
For Seadrift, Dow and X-energy state in their May 2026 release that the plant has minimal water requirements and no cooling towers or water intake structures. Those are company claims. The same arithmetic applied to a wet-cooled 320 MWe plant gives about 5.1 million gallons a day, which shows what the claim is measured against.
For Kemmerer, TerraPower CEO Chris Levesque said at site selection in 2021 that cooling water availability was one of the factors weighed. The NRC’s final environmental impact statement, NUREG-2268, published in October 2025, covers water resources. I did not find a published gallons-per-day figure in the sources reviewed, so read that chapter before quoting one.
For Clinch River, the permit announcement describes a water-cooled natural circulation design. Find the consumption and withdrawal figures in TVA’s environmental report for the site. The context that makes numbers like these comparable across projects is what SMRbrief Pro was built to provide.
Dry cooling saves water and costs output
Air cooling removes most of the water problem and adds a different one. NuScale said its air-cooled condenser for the Idaho project would cut water use by more than 90 percent, but electrical output would drop 5 to 7 percent, according to NEI’s 2020 account. That project was cancelled before construction, so no grid-scale dry-cooled SMR has operated. Argonne National Laboratory says dry cooling is less efficient than wet cooling but suitable where wet options are unavailable, in work with Energy Northwest.
A 5 to 7 percent output loss is a real cost over a 60-year plant life, and it explains why developers use dry cooling only where water is scarce or expensive. If a project near you claims air cooling, ask for the output penalty in the same sentence.
Drought tests the river-cooled plants first
This summer showed what low water does to existing plants. Carbon Brief’s review found that even a 15 C rise in cooling-water temperature costs a large reactor only about 6 percent of output. It also found that low river levels can last far longer than hot spells. In 2026, Hungary’s Paks plant halted for the first time in 44 years because the Danube fell to record lows, according to AFP reporting. One Cernavoda unit in Romania was also taken off the grid.
Those are large reactors, but any river-cooled SMR faces the same exposure. The neighbor’s risk is not only how much water the plant takes. It is also whether the plant keeps running when the river drops, and what the operator does with its water draw in a drought year.
So here is the question to bring to the next public meeting: if the river hits its record low, does this plant keep its water, or does the town?