The United States still has the largest network of nuclear power plants in the world — bigger even than France, which gets about three-quarters of its power from nuclear. But the U.S. nuclear supply is shrinking fast. Plants constructed during the building spree in the 1960s and ’70s are being retired as they reach the end of their planned operating lifespan, while they simultaneously come under powerful price pressure from natural gas and ever-cheaper renewables. Recently the utility FirstEnergy announced plans to shut down three nuclear power plants in Ohio and Pennsylvania, following an announcement from Exelon that it would close the Three Mile Island plant next year.

This is bad. Whatever you think of nuclear power, it is still the largest zero-carbon portion of our existing energy infrastructure. We should wring every last kilowatt out of that infrastructure until renewables (or perhaps future superior nuclear tech) are ready to take up the slack. Reducing carbon dioxide emissions should be an overriding goal for all energy policy.

Now, that is not to say that we should prioritize building new nuclear power plants. Contrary to the stereotype of nuclear being blocked by dimwitted environmentalists, the big problem with constructing nuclear power is that it is stupendously expensive and complicated, and American institutions — whether they’re public or private — have developed severe problems with executing that type of project. Indeed, recent construction on a reactor in South Carolina got so over budget and behind schedule that it bankrupted the contractor Westinghouse and the whole thing was abandoned. (That’s just one of many such stories.)

Additionally, the American nuclear industrial and regulatory framework could badly stand an overhaul. Nuclear technology has advanced considerably over the past 40 years, but the existing overall nuclear system of technology, plants, and regulators hasn’t been changed much over that time. Nuclear reprocessing (taking spent nuclear fuel and running it through a procedure to extract more useable isotopes) is worth examining, as it would produce a lot of fuel and sharply cut high-level waste (which takes hundreds of thousands of years to decay into safe material) at the cost of a lot more low-level waste (which only takes hundreds of years). Most promisingly, advanced reactors using thorium and other materials are theoretically vastly superior to existing nuclear designs. It would probably take billions upon billions of dollars to spin up an entire industrial ecology, but if it works out, the payoff would be tremendous.

But all that is ancillary to the question of what to do with existing nuclear reactors. The truth, uncomfortable to some liberals and environmentalists, is that nuclear power is right now the most vital part of the American electrical grid. It’s old, a bit rickety, and overpriced, but produces a huge amount of zero-carbon power — 805 terawatt-hours (TWh) in 2017, or about 20 percent of the total. Renewables still only make up 17 percent, and the plurality of that capacity is hydropower.