Can America Build Nuclear Again? Part 2
A Half Century of Asymmetric Regulation Explains Why Nuclear Is So Expensive to Build
Tomorrow I’m participating in an event at AEI — How Will the US Unleash Commercial Nuclear Power? — for which you can register to attend in person or online. As prep, I am publishing a three-part series on U.S. nuclear power. You can find Part 1 here. Today, in Part 2, I compare federal regulation of natural gas versus nuclear power. With Part 3, coming later this week, I’ll share an XLS file with all data, sources, and links of the series.
Today’s installment in this series is motivated by the figure below, which shows U.S. electricity generation trends for natural gas and nuclear starting in 2000: Nuclear generation has remained constant while natural gas generation more than tripled. This post reports what I learned when I looked to better understand the reasons for this divergence.
One big part of the answer: asymmetric federal regulation.
The story in a nutshell: Nuclear power plants are cheap to run, but expensive to build, whereas natural gas plants can be expensive to run, but cheap to build. The cost of capital underlying power plant construction — and the asymmetry between nuclear and natural gas — is a major explanation for the divergent outcomes shown in the figure above.
Let’s dive in . . .
The existing fleet of U.S. nuclear power plants produces some of the cheapest electricity in the country: NEI/EUCG data, compiled from utility financial filings with FERC, put the average total generating cost for the fleet at $33.74/MWh in 2024 — covering fuel, operations and maintenance, and capital expenditures. Nuclear fuel accounts for only a small fraction of that total: uranium’s extraordinary energy density (a single fuel pellet contains roughly as much energy as 17,000 cubic feet of natural gas) means that fuel cost is insulated from commodity price swings, unlike fossil fuels.
Nuclear power’s price stability is compounded by the fleet’s 92% capacity factor in 2024 — the highest of any generating source in the country — which spreads fixed costs across the maximum possible output. Those costs are largely on the capital side. The only nuclear plant completed in the U.S. since 1996, Vogtle 3 & 4, cost about $35 billion for ~2,200 megawatts of capacity — approximately $15,700 per kilowatt all-in — while the EIA’s AEO2025 puts the overnight construction cost estimate for a new large light-water reactor at $7,821/kW in 2023 dollars, before financing costs are added.
An investment that large, requiring ten to fifteen years to complete before a single kilowatt-hour is sold, defines the economic challenge faced by new nuclear power.
Combined-cycle gas turbine (CCGT) plants occupy the opposite position. Construction costs are modest and well-documented: EIA’s generator construction cost data put the capacity-weighted average for new CCGT plants at $722/kW in 2022 and $898/kW in 2023, about one-twentieth of Vogtle’s all-in cost. A CCGT can be financed, permitted, and generating revenue within two to three years of a final investment decision. Fuel is essentially the entire operating cost, and natural gas prices vary. The Henry Hub annual average ranged from $1.86/MMBtu in 2020 to $6.41/MMBtu through the first eight months of 2022 — a more than threefold swing within two years — before falling back to $2.21/MMBtu in 2024.
That range translates to fuel costs swinging from roughly $13/MWh at 2020’s annual average to $44/MWh at 2022’s average, with the 2022 peak periods pushing considerably higher.1 Variable operations and maintenance (O&M) costs add about $3–4/MWh.
The consequence is that a gas plant’s economic viability is essentially a bet on the future price of natural gas — a commodity whose price is set by weather, storage levels, LNG export demand, wars, geopolitics, politics, and associated production from oil wells — none of which the plant operator controls. When gas is cheap, CCGTs undercut almost everything else on the grid. When it isn’t, they don’t.
What explains the capital cost difference: natural gas vs. nuclear?
My review of the literature2 came up with seven factors that help to explain the huge difference in capital costs between natural gas and nuclear.
1. Federal Licensing Timeline and Full-Cost Fee Recovery
The fee-recovery mandate — the requirement, codified at 10 CFR Part 170, that the Nuclear Regulatory Commission (NRC) recover 100% of its annual budget directly from licensees — traces to a 1950s policy: nuclear power was a private commercial enterprise, so its licensees, not taxpayers, were expected to bear the full cost of federal oversight. The practical consequence is that a combined license application for a new reactor accumulates $50–100 million in regulatory fees before construction begins, on top of a review process that averages six years and sometimes exceeds ten — costs that translate directly into capital: at 7% cost of capital on a $10 billion project, each additional year of pre-construction review adds roughly $700 million, or about $315/kW on a 2.2 GW plant. The Federal Energy Regulatory Commission (FERC) gas certificate process, by contrast, is funded through congressional appropriations; a gas developer pays filing fees in the thousands of dollars, not tens of millions. The ADVANCE Act (2024) capped some NRC fees, but the underlying statutory fee-recovery mandate remains.
2. Mid-Construction Backfitting Authority
This mechanism drove more cost escalation than any other single regulatory instrument in the post-Three Mile Island (TMI) era. As discussed in Part 1, the NRC retains authority to require design modifications to plants already under construction. Regulators exercised that authority broadly after Three Mile Island. Lovering et al. (2016) document that among reactors under construction when TMI occurred and were subsequently completed, median costs were 2.8 times greater than for earlier plants, traceable to physical rework rather than financing. No federal equivalent applies to gas. Executive Order 14300 (May 2025) directed the NRC to set clearer limits on when mid-construction design changes can be required, but Congress has not yet passed legislation locking those limits into law — meaning a future administration could revisit them.
3. Quality Assurance
Federal quality assurance rules for nuclear construction, set out in 10 CFR Part 50, Appendix B and written largely in the 1960s, require that every weld, cable installation, and component specification be documented and traceable through a formal chain — meaning any deviation, however minor, must be recorded, reviewed, and typically corrected before work can continue. A 2010 NRC staff study estimated these requirements add 15–25% to construction costs compared to a risk-based alternative. At Vogtle, the consequences were severe: the DOE Liftoff Report (2023) documented component test failure rates of 40–80% at various points during construction. Those rates were driven by three compounding problems: (1) construction began from an incomplete design, so specifications kept changing mid-build; (2) the US hadn't built a nuclear reactor in decades, leaving the skilled labor pipeline atrophied and forcing Vogtle to serve as a training ground for roughly 30,000 workers with no nuclear construction experience; and (3) suppliers who hadn't produced nuclear-grade components in years had to relearn the strict QA requirements from scratch. Each failed component test triggered Appendix B's documentation and rework cascade, turning what might have been manageable construction errors into systematic, costly delays. Natural gas plants face no equivalent federal mandate — they comply with standard ASME and ANSI commercial codes administered by licensed professional engineers.
4. NEPA Environmental Review
Both technologies fall under the same law — NEPA (National Environmental Policy Act, 42 U.S.C. § 4321) — but they are treated very differently, for at least two reasons. First, the technologies both produce electricity but do so very differently:
a nuclear plant produces long-lived radioactive waste, requires exclusion zones and evacuation planning for surrounding populations, and creates cumulative radiological impacts extending well beyond the plant's operational life;
a gas plant's environmental impacts — air quality, water use, local ecology — are real but bounded and have long been familiar to regulators.
Second, a 1971 court ruling locked in the differential regulatory treatment. In Calvert Cliffs' Coordinating Committee v. AEC, 449 F.2d 1109 (D.C. Cir. 1971), the D.C. Circuit held that the Atomic Energy Commission (AEC) must apply environmental review "to the fullest extent possible" at every stage of licensing — a ruling that forced the AEC to halt all nuclear plant licensing for eighteen months to overhaul its procedures. No equivalent ruling ever reshaped FERC's gas-plant NEPA process. The result: nuclear construction triggers a full Environmental Impact Statement governed by NRC NUREG-1437, typically running 3–5 years and exceeding 1,000 pages. FERC's NEPA review for gas plants typically completes in 12–18 months. The requirement has a direct capital cost: at 7% cost of capital on a $10 billion, 2.2 GW project, three additional years of pre-construction review adds ~$1,000/kW in interest costs.
5. Decommissioning Mandates
When a natural gas plant closes, the cleanup is straightforward — remove the equipment, remediate any minor hydrocarbon contamination, and then the site is largely reusable. When a nuclear plant closes, it leaves decades of accumulated radioactive contamination requiring licensed radiological remediation, controlled dismantlement, and long-term waste management — a process the NRC estimates generally costs $300–400 million per reactor, with site-specific figures often running higher. Because that liability falls due 40–60 years after the investment decision, and because an insolvent or defunct operator would leave the cleanup to the public, legislation requires nuclear licensees to secure the full decommissioning cost — through a trust fund, surety bond, or parent company guarantee — before the plant begins operating. In practice, most utilities accumulate funds over the plant's life through an external trust, contributing roughly 0.1–0.2 cents per kilowatt-hour generated. The capital cost consequence is twofold: the decommissioning obligation is a contingent liability that project lenders must underwrite alongside construction debt, increasing total financial exposure; and where upfront prepayment is required, it ties up hundreds of millions of dollars before the plant generates any revenue, raising the weighted average cost of capital for the entire project. No equivalent federal obligation exists for gas plant developers — whatever cleanup costs arise are unregulated and treated as a normal operating expense near end of plant life.
6. Price-Anderson Liability Insurance
The Price-Anderson Act (42 U.S.C. § 2210) was enacted in 1957 because private insurers refused to underwrite nuclear risk. The potential magnitude of a reactor accident was too large, too uncertain, and too correlated — one event could produce catastrophic claims across an entire region — for commercial providers. Without liability coverage, no utility would build a reactor, so Congress stepped in. The resulting two-layer system requires each plant site to carry the maximum commercially available primary insurance, raised to $500 million per site as of January 2024; if an accident causes damages beyond that, every other covered reactor in the country is assessed a retrospective premium of up to $158 million per reactor per incident, payable at no more than $24.7 million per year. With 95 covered reactors, the total pool stands at roughly $15.5 billion. Annual primary insurance premiums run $2–4 million per reactor, with no gas equivalent — gas plants carry standard commercial liability through ordinary property and casualty markets. The capital cost impact results from the direct cost of $2–4 million per year in mandatory premiums, and a subtler indirect effect — the retrospective premium is a contingent liability owed if any U.S. reactor has a major accident, regardless of fault, creating an industry-wide correlated tail risk that lenders and equity investors must price into required returns on any new nuclear project.
7. IRA Incentive Structure
The Inflation Reduction Act (2022) gave the existing nuclear fleet significant new support — but its credit structure for new construction embeds a timing asymmetry that disadvantages nuclear relative to technologies with shorter build times. The differential exists because the law was written primarily in support of specific technologies — solar, wind, and battery storage — that complete in one to three years. Its supposed technology-neutral credit structure treats nuclear identically to solar on paper while producing radically different economic outcomes. Calling these credits “technology neutral” is accurate in the narrow sense that the rate per kilowatt-hour is the same regardless of technology; it is misleading in any economically meaningful sense, because when you receive that credit depends entirely on when you complete construction — and applying the same nominal credit to technologies with a two-year construction timeline and technologies with a twelve-year+ construction timeline is neutral in name only. A dollar of credit deferred twelve years is worth roughly half a dollar of credit deferred three years at an 8% discount rate. Applied to the Section 45Y production credit of 1.5¢/kWh on a 2.2 GW plant running at 92% capacity factor, that timing gap reduces the present value of the credit stream by approximately $350/kW3 relative to a technology completing in three years — a structural penalty baked into the IRA’s architecture and unaddressed by any subsequent legislation. The One Big Beautiful Bill Act (2025) improved existing nuclear's relative position by terminating 45Y/48E credits for wind and solar while preserving them for nuclear — but left the timing asymmetry for new nuclear construction unaddressed.4
Taken together, these seven mechanisms help to explain why building a nuclear plant in the United States currently costs roughly fifteen to twenty times more per kilowatt than building a natural gas plant.
Why the Regulatory Asymmetry?
Some of the regulatory asymmetry came from nuclear-specific congressional choices, including the Price-Anderson Act, the fee-recovery mandate, and the Appendix B QA framework. Congress could in principle revisit or revise each of these provisions in a way that reduces capital cost escalation for new nuclear plants.
Other aspects of the regulatory asymmetry arrived via the administrative state. NEPA carried no nuclear-specific provisions, and Congress never voted on the scope of environmental review that ended up applying to reactors. The 1971 Calvert Cliffs decision did that, and Congress never revisited the issue.
No comparable ruling reshaped FERC’s natural gas-plant NEPA process. The backfitting authority that applies to nuclear but not gas, similarly, came from agency rulemaking after Three Mile Island — not a congressional vote.
Once cost escalation became apparent, why didn’t Congress act?
One answer to this question is that the JCAE — the Congressional committee that might have explored and perhaps addressed this outcome — was terminated in 1977, and no committee with similar nuclear energy jurisdiction was ever established in its place. One reason for the broader reform was concern that nuclear policy had fused promotion and regulation in a single institutional structure — a concern that Congress had already acted on in 1974 when it abolished the AEC for that reason and separated its functions into two agencies. Those concerns were justified, but the abolition of the committee did much more than open the door to stronger regulation — whether intended or not, it arguably helped to put the brakes on new nuclear energy.
The figure below shows the current fragmented way that Congress oversees nuclear energy — currently spread across 5 Senate committees and 6 House committees. This decentralized structure makes it difficult — maybe impossible — for Congress to consider nuclear energy comprehensively, much less legislate comprehensively.
The Joint Committee on Atomic Energy (JCAE) held combined authorizing-and-oversight jurisdiction over essentially everything related to nuclear power from 1946 until its abolition in September 1977. Its dissolution split nuclear jurisdiction across Armed Services, Energy and Natural Resources, Environment and Public Works, and Foreign Relations committees — none of which inherited JCAE’s combined promotional-and-technical mission, and none of which had jurisdiction over NEPA. The termination of the JCAE occurred as cumulative licensing delays from the Calvert Cliffs era became visible in plant economics, leaving no successor institution with both the jurisdiction and the mandate to revisit NEPA’s application to nuclear.
Congressional amendment of NEPA requirements for nuclear has never been seriously considered. A nuclear-specific NEPA carve-out would invite every other regulated sector to seek their own preferential treatment. That said, Congress — during both the Biden and Trump administrations — has not shied away from picking energy technology winners (and losers). Congress has treated nuclear like a loser for a half century.
Another factor was Three Mile Island (1979), which foreclosed the notion of promoting nuclear energy for a generation — a political third rail.
When Congress did eventually act on nuclear energy, it acted marginally rather than comprehensively. For instance, Title XXVIII of the Energy Policy Act of 1992 created the combined construction-and-operating license to stop intervenors from relitigating settled construction-stage issues at the operating stage — but did not revisit NEPA requirements as defined by Calvert Cliffs. Similarly, the ADVANCE Act (2024) and Executive Order 14300 (2025) work within the existing framework rather than revisiting the bigger picture.
One reason that the regulatory asymmetry between nuclear and natural gas persists is that the institutional capacity to revisit it — concentrated nuclear-specific jurisdiction spanning both promotion and the general statutes constraining it — has not existed since 1977.
Beyond the Feds: State Authority
The seven explanations for the asymmetric regulation of nuclear versus natural gas all are at the federal level. A separate layer of asymmetry operates at the state level.
In Pacific Gas & Electric Co. v. State Energy Resources Conservation and Development Commission, 461 U.S. 190 (1983), the Supreme Court further tilted the scales in favor of natural gas: the Atomic Energy Act gives the NRC exclusive jurisdiction over radiological safety, but states retain traditional authority over the economics of power plant construction — need determinations, certificates of public convenience, and moratoria framed as economic rather than safety judgments. The Court upheld California’s statutory moratorium on certifying new nuclear plants based on economics.
Since the 1970s, sixteen states have enacted nuclear-specific restrictions of this kind — outright moratoria, voter-approval requirements, or conditions tied to federal progress on permanent waste disposal (via DOE Office of Nuclear Energy tracker). No comparable state-level barrier exists for natural gas. A developer choosing between a gas CCGT and a nuclear plant in Massachusetts or Minnesota will face a legal question for one technology that simply does not arise for the other — in addition to all seven federal mechanisms above.
Mind the Gap
The regulatory gap between nuclear and natural gas is not an accident of nature or the inevitable product of the technologies’ different risk profiles. It is the cumulative output of specific federal decisions — some deliberate, some inadvertent — made mostly between the 1940s and the 1970s, and then largely frozen in place by the politics and institutions that followed. The seven federal mechanisms discussed in this post and an additional state-level layer combine to produce a capital cost differential of roughly fifteen to twenty times between natural gas and new nuclear construction.
What would reversing this asymmetry actually require?
Not one legislative fix but several simultaneous ones, including:
a fee recovery reform that puts nuclear and gas on equal regulatory cost footing;
a statutory design-finality standard that limits mid-construction backfitting;
a risk-informed revision to the 1960s-era QA framework;
a NEPA scoping approach that calibrates review depth to actual radiological risk rather than applying blanket EIS requirements to every reactor;
and an IRA credit structure that accounts for nuclear’s construction-timing disadvantage rather than treating a twelve-year investment horizon as equivalent to a two-year one.
This is not a short list, and it does not include the state-level changes that would be required. Whether the current policy environment and growing bipartisan interest in nuclear deployment might actually move the needle on these structural issues is the question the AEI discussion tomorrow aims to address.
In the closing installment of this series, I’ll take a deeper dive into policy options that Congress might consider as part of a nuclear forward agenda.
Before you go, one request — Please click that “❤️ Like” button. Thanks!
Comments, critique, other views are all welcome!
Posts like this one exist because of THB’s free and paid subscribers.
Thank you!
Calculated at a representative CCGT heat rate of 6,900 Btu/kWh.
The literature I reviewed was policy reports and grey literature — I did not find any comparative analysis of nuclear versus natural gas regulation in the academic literature, a gap which was one motivation for this post. If I’ve missed something, do let me know!
The $350/kW is my estimate— It assumes the Section 45Y clean electricity production credit at 1.5¢/kWh ($15/MWh) — the rate available to facilities meeting prevailing wage and apprenticeship requirements — running for 10 years from the placed-in-service date (both per IRS). Applied to a 2,200 MW plant operating at a 92% capacity factor (EIA, 2024 fleet average), annual credit value = $265.9M. Discounted at 8%, the present value of a 10-year credit stream beginning in Year 12 (nuclear, consistent with the Vogtle timeline) is approximately $348/kW, versus $696/kW for a credit stream beginning in Year 3 (a technology completing construction in two to three years). The gap is $348/kW, rounded to $350/kW. At 7% — the rate used elsewhere in this post — the gap narrows to approximately $290/kW; at 10% it widens to approximately $410/kW.
The One Big Beautiful Bill Act (July 2025) improved nuclear’s relative position somewhat: it preserved Section 45U for the existing fleet through 2031, kept Sections 45Y and 48E available for new nuclear beginning construction before 2034, and added a new 10% credit adder for facilities in “nuclear energy communities,” while terminating 45Y and 48E for wind and solar placed in service after December 31, 2027, and effectively ending the Section 45V clean hydrogen credit for most announced projects. What it did not touch is the structural timing problem: credits still accrue only post-completion, and nuclear’s 10–15 year construction timeline still means those credits arrive at a deep discount relative to their nominal value compared with any technology that completes faster.
And all of this is in service of a vanishingly small gain in safety. At TMI the containment vessel did its job and nobody was harmed by radiation (although a great many were harmed by the fear and stress from the release of the film China Syndrome 3 weeks earlier, and wall-to-wall media coverage as a result). There is such a thing as excess safety; we could make cars almost perfectly safe if we required all drivers to wear 5-point racing harnesses, helmets, Nomex suits, limited horseless carriages to a maximum speed of 12 mph, and required a licensed safety officer to go ahead of each one in a residential area to warn pedestrians of its approach. This is roughly what the NRC attempts to do through ALARA, "as low as reasonably achievable" where "achievable" is defined solely on safety with no consideration of cost.
Follow that up with the realization that long-term, low-intensity radiation is far less dangerous than the official risk model suggests. We know that because regions in India with everyday background radiation nearly 8 times the US EPA annual limit were studied over 20 years and showed not only no increase in cancer, but a statistically insignificant decrease for a linear fit.
The same result comes from accidental exposure in some Taiwan apartment buildings inadvertently built from contaminated recycled steel. The worst exposed of that population had a cumulative dose over 20 years roughly half of what would kill you outright if you received it all at once, with no increase in cancer. This is irrefutable evidence that our assumption that any amount of radiation is hazardous, and that damage is cumulative over a lifetime, is flat wrong.
The Navy has quite a few nuclear-powered vessels including submarines. I assume none of these laws actually apply to the ones they build. Relying on defense of country arguments, it can't take tens of years to build a submarine. Why is nuclear power on land not afford the same rationale to use Navy nuclear technology to build on land? Taking into account the Navy has secrets about it's nuclear program it must keep secret. I'm sure they probably also use only one design modified for different vessels.
It seems we have the answer but no one wants to acknowledge it.