Why America Stopped Building Reactors and What the NRC Actually Does

A nuclear power plant is the most expensive tea kettle ever built. Strip away the concrete, the control rooms and the acronyms, and the thing is designed to do exactly what a pot on your stove does: boil water. The steam turns a turbine, the turbine turns a generator, and your lights come on. Humanity has been using steam to make power since before the Civil War. The only exotic part is the fire under the pot.

So why is it that the country that built the first commercial reactors, that put a reactor in a submarine and a carrier before anyone else, struggled to finish a pair of new ones in Georgia, and abandoned a pair in South Carolina after spending a fortune? The answer is not that the physics failed. The answer, I will argue, is a licensing and cost-overrun spiral that Washington made and Washington could fix.

Let me take you through how a reactor works, how we got the rules we have, and what a sensible reform looks like.

How the kettle works

The reactors that make nearly all of America's nuclear electricity are called light-water reactors. "Light water" is just ordinary water, as opposed to heavy water. Here is how one works, in five steps.

1. Fuel. Uranium is mined, processed, and enriched so that a few percent of it is the fissionable isotope uranium-235. It is formed into ceramic pellets and stacked in long metal rods.
2. Fission. When a uranium-235 atom absorbs a neutron it splits, releasing heat and more neutrons, which split more atoms. That is a chain reaction.
3. Control. Rods that absorb neutrons can be pushed in or pulled out to slow or speed the reaction. Water also slows the neutrons to the speeds that make fission more likely, and it carries away the heat.
4. Steam. In a pressurized water reactor, the hot water stays under high pressure in a closed loop and heats a second loop of water to make steam. In a boiling water reactor, the water boils right in the reactor vessel. These are the two designs in the American fleet.
5. Electricity. The steam spins a turbine connected to a generator, just as in a coal or gas plant.

Notice what is the same as every other plant, and what is different. The turbine hall of a nuclear plant looks much like the turbine hall of a coal plant. What is different is the heat source and everything that surrounds it to keep that heat source safe: thick steel vessels, a containment building, redundant cooling systems and backup power. Engineers call the approach "defense in depth," meaning layers of protection so that one failure does not become a catastrophe.

1954: the government opens the door

In the first years after the Second World War, atomic energy was a government monopoly. The Atomic Energy Act of 1946 kept the technology under federal control. President Eisenhower's "Atoms for Peace" speech to the United Nations on December 8, 1953, signaled a different course, and the Atomic Energy Act of 1954 let private companies own and operate reactors under a federal license, with the Atomic Energy Commission as regulator.

The first full-scale commercial plant in the country, at Shippingport, Pennsylvania, began delivering power in 1957. The industry was on its way.

Here is a feature worth noticing. The AEC had two jobs that pull in opposite directions: it was supposed to promote nuclear power, and it was supposed to regulate its safety. Critics of the arrangement argued for years that the same agency should not be both cheerleader and referee. We'll see what Congress did about that.

1957: Price-Anderson and the liability bargain

Private companies were wary of building reactors because of a risk they could not insure. No private insurer was willing to cover the full liability of a severe accident. So in 1957 Congress passed the Price-Anderson Act. It sets a framework in which the operator carries required insurance and, beyond that, the industry shares in a pooled fund for the largest accidents. Congress has renewed the law more than once.

You can debate whether a liability limit is a subsidy. I think it is closer to a bargain: the industry pays for the first layers of coverage and, in the larger layers, pays into a pool. If your reaction is that nuclear power should have to bear every theoretical risk in full, apply the same standard to every other industry whose worst case is hard to insure. The more important point is that the structure existed to make private ownership possible, and it did.

1975: a new referee

By the early 1970s the double role of the AEC looked untenable. Congress passed the Energy Reorganization Act, signed in October 1974, which abolished the Atomic Energy Commission and split its functions. Promotion and research went to a new energy agency that eventually became part of the Department of Energy. Regulation went to a new body, the Nuclear Regulatory Commission, which began work in January 1975.

What does the NRC actually do? It does not build or run reactors. It does not set electricity prices. It licenses nuclear plants, writes safety rules, inspects them and enforces those rules. It is led by five commissioners nominated by the President and confirmed by the Senate, serving staggered five-year terms, with no more than three from the same political party. Independence is the design.

That structure is sound, and I am not here to attack it. Having a referee who does not also own a team is a conservative idea if anything is. The question is how the referee behaves.

1979: Three Mile Island

On March 28, 1979, a reactor at Three Mile Island near Middletown, Pennsylvania, suffered a partial core meltdown. A relief valve stuck open, the operators misread what the instruments were telling them, and part of the core was damaged. The release of radioactivity outside the plant was small. The accident did, however, shake public confidence, and it shook the regulatory system.

President Carter appointed a commission led by John Kemeny, then president of Dartmouth College, to investigate. It reported in October 1979 and found that the problems lay in people, procedures and organization as much as in hardware. The response included new requirements for operator training, emergency planning, control room design, and plant instrumentation. The industry also created its own peer-review organization, the Institute of Nuclear Power Operations, in late 1979.

Some of these reforms made the plants safer, and that is not nothing. But there were also costs. Requirements changed while plants were already under construction, and each change meant redesign, rework and delay.

The spiral: how a project goes from expensive to impossible

Now we come to the heart of it. Let me give you the economics in plain terms, because this is where nuclear power's reputation was made and broken.

A reactor is a high-fixed-cost project. Almost all of the expense is building it, and the fuel is cheap by comparison. That means the cost of money matters enormously. Every year of delay is a year of interest on billions of dollars with no electricity to sell.

Here is a hypothetical. Suppose a utility spends $10 billion to build a plant and borrows at 8 percent. A year of interest on $10 billion at 8 percent is $10,000,000,000 × 0.08 = $800,000,000. A three-year delay adds about 3 × $800,000,000 = $2,400,000,000 in simple interest, before counting any compounding or the price of the changes that caused the delay. The delay costs more than money. It makes the next utility think twice.

Add to that the problem of one-of-a-kind designs. In the 1970s, American utilities built reactors from several different vendors and architect-engineers, with each plant customized. Nobody learned from the last build the way a factory learns. Regulatory requirements shifted in the middle of construction. Lawsuits and hearings held up approvals. Interest rates soared in the late 1970s and early 1980s, and electricity demand growth slowed after the 1973 oil shock. Utilities canceled many planned reactors. It was a perfect storm in which the technology was only one of many ingredients.

Then the story repeated itself. At Plant Vogtle in Georgia, two new reactors were completed years behind schedule and billions of dollars over budget. In South Carolina, the V.C. Summer project was abandoned in 2017 after the designer, Westinghouse, went into bankruptcy. Those were the first new builds in a generation, run by teams that had not built a reactor in decades. That is the cost of stopping.

The licensing maze

The traditional process was in two steps. A utility first applied for a construction permit and, if it won one, built the plant. Then it applied for an operating license. Each step involved review and hearings, and the second step meant that a utility could spend billions before learning whether it would be allowed to run the plant.

Reformers saw the problem. In 1989 the NRC created a streamlined path in a part of its regulations known as Part 52. A company can seek an early site permit to approve a location, a design certification to approve a standard reactor design, and a combined construction and operating license, often called a COL, that approves both building and operating in one go. The Energy Policy Act of 1992 wrote the one-step idea into statute.

On paper that was a major improvement. In practice, the projects that used it were the first of a kind and ran into the problems I just described. The process did not fail so much as it was never given the chance to become routine.

The strongest objection

Let me state the opposition's case fairly. A nuclear accident can have severe and long-lasting consequences, and the public has little tolerance for it. Three Mile Island proved that operators can fail, and a rule written after an accident exists for a reason. A regulator that cuts corners to meet a deadline invites the next accident. The costs are high because safety is expensive, and the people who tell you otherwise are selling something.

Much of that is right. Safety is expensive, and I am not arguing for fewer protections. A conservative has no business trading away the safety of the family that lives near a plant.

But here is the answer. The question is not whether to have rules but whether the rules are predictable, finite and proportionate. Consider these points.

One: a requirement that changes mid-construction does not increase safety in proportion to its cost if it forces years of redesign. Finality matters.

Two: a principle called ALARA, "as low as reasonably achievable," governs radiation exposure. The word "reasonably" is doing a lot of work. The argument I'd press is that reasonableness should include cost, so that a dollar spent for a tiny gain in safety is weighed against the dollar's alternative uses, including cleaner or safer power that never gets built.

Three: delay has victims too. A reactor not built is a coal or gas plant run longer, and a family that pays more for power. The cost of saying no is real even when it is hidden.

What a sensible reform looks like

Here is what I think a reform would include.

1. Firm deadlines. Set statutory time limits for decisions on applications, with the commission accountable for meeting them. Congress has moved in this direction, including in the ADVANCE Act of 2024. Check Congress.gov for what it requires.
2. Standardize. Approve a design once and build it many times. The goal is to turn each plant from a prototype into the tenth of its kind. The experience of France, laid out in France Built Reactors, Germany Shut Them, and America Argued in a Side-by-Side, shows what repetition does.
3. Finality. Once a design and a license are approved, hold the rules steady for the project unless there is a real safety reason. Do not reopen settled questions.
4. Keep the independence. The five-member commission, with Senate confirmation and bipartisan limits, is worth protecting.
5. Fix the waste promise. A plant that cannot say where its used fuel goes is a harder plant to finance, a story told in Yucca Mountain and the Nuclear Waste Promise Washington Broke.
6. Cut the duplicate paperwork. The environmental review law is a big part of the delay on many projects, as NEPA Explained, the 1970 Law Where Paperwork Outlasts the Project explains.

None of that weakens the safety case. All of it makes the process quicker and more certain.

Why it matters for your family

Nuclear power supplies about a fifth of America's electricity, and it does it without regard to the weather. It runs day and night, in July and in January. If you want to understand why that matters on the cold morning when demand spikes, How the Power Grid Really Works and Why Supply Must Match Demand Every Second explains how the grid must balance supply and demand every second. And if you are weighing the whole menu, All of the Above, Graded on What a Family Actually Needs From Power grades each source by what a household actually needs.

The bottom line

• American nuclear power stalled because of cost overruns and a changing, slow licensing process, not because the technology failed. Three Mile Island in 1979 was a turning point, but rising interest rates and slowing demand also canceled many plants.
• The Nuclear Regulatory Commission, created in 1975, is the independent referee. The goal is to keep the referee and make the game predictable: firm deadlines, standard designs, and finality once a design is approved.
• Safety and speed are not opposites. A standardized plant built the tenth time is likely to be safer and cheaper than a custom plant built the first time.
• Ask your members of Congress whether they back statutory deadlines and a durable waste solution. Then ask your utility what its plan is for reliable power on the coldest night of the year.