51 years ago, in reaction to the 1973 OPEC oil embargo against the United States, President Richard Nixon announced “Project Independence,” a plan to build one thousand nuclear reactors by the turn of the century. The power from these plants could virtually end U.S. reliance on global oil. But in 1979, after the Three Mile Island scare, that dream collapsed. The U.S. stopped building nuclear power for a generation, despite having the technologies for an abundant, sovereign energy future. We chose badly.
After the passage of the 2005 Energy Policy Act, the country seemed ready to change course. The law provided subsidies and incentives for nuclear development to reduce America’s reliance on Middle East oil amid the Iraq war.
In 2009, construction began on a nuclear reactor on a remote Savannah River site in Georgia, the first in over three decades. The project consisted of two units: Vogtle 3 and Vogtle 4, each capable of generating 1.1 gigawatts/GW (enough to power around 825,000 American homes). Today, the Vogtle plant has four reactors (the only one in the U.S.) and at over 4 GW production, is the second largest power facility in the country, after Washington’s Grand Coulee Dam.
The story of the Vogtle plant’s construction illustrates painfully why we no longer build new nuclear power plants in the U.S. Its completion after decades without new plants could have marked nuclear power’s rebirth in the U.S. But it did not.
Problems with the project began almost immediately. Soon after construction commenced, the Nuclear Regulatory Commission (NRC) published a rule requiring that all new nuclear power plants be able to withstand a direct hit from a commercial jetliner. The benefit of that rule was dubious, and the commission acknowledged that it wasn’t necessary for adequate safety. But the NRC required the engineering, procurement, and construction (EPC) company building the Vogtle plant to completely redesign the containment and shield structure, discarding years of planning and preparation.
It took over two years for the NRC to approve the unit’s reactor design, and construction began with only 40% design completion.
A further complication was the high failure rate of project components, built offsite at a Louisiana modular fabrication yard and shipped to Georgia as modules. Quality issues arose as the contractor was unaccustomed to nuclear-grade construction standards. The modules often arrived in Georgia late, incomplete, or both. Regulatory paperwork routinely held up shipments. The design of the modules was also problematic. Independent analysts later testified that components were hard or impossible to make as designed.
Construction work onsite faced issues. NRC inspectors found 750 tons of rebar (metal bars for concrete reinforcement) in the nuclear island area, where the nuclear reactor and containment dome would be, didn’t meet specifications. Instead of letting the EPC fix the issue in-place, the NRC required the work to be completely redone. A Department of Energy study later found test failure rates for project components ranged from 40-80% during some parts of the project.
Reworking construction increases costs and delays on any project. But on a nuclear project, where an accepted rule of thumb is that every piece of material is 10 times more expensive than a natural gas plant due to regulation, this rework caused catastrophic cost increases and years of delays. The spiraling losses forced the project’s EPC to file for bankruptcy in 2017.
The project’s issues worsened because the EPC had to rely on inexperienced workers building the project out of sequence due to design changes and constant rework. At the peak of construction, Vogtle required 7,000 on-site workers. Given the project’s remote location and lack of available skilled labor, the EPC often had to settle for inexperienced workers; a large percentage quit or went absent.
The project completed construction in 2024, seven years behind schedule and $21 billion over budget. Of the 31 new reactors announced after the Energy Policy Act of 2005, Vogtle’s are the only completed ones. The risk of building nuclear projects made them hard to finance. In the last fifteen years, the advent of hydraulic fracturing (fracking) has made natural gas cheaper than ever. To policymakers, the value proposition of nuclear can seem fleeting.
In September, Constellation Energy, a major power provider, announced it would restart a reactor at the Three Mile Island plant in Pennsylvania. In March 1979, the second reactor at Three Mile Island sustained a partial meltdown, costing billions and igniting a national panic about nuclear power. The plant operated through 2019 with one reactor, until it closed for financial reasons.
When Constellation announced its reopening, they said it would power just one customer: Microsoft.
Today, the U.S. faces a dilemma: political pressure to replace the nearly 20% of energy that comes from coal, and a massive need for new power generation capacity for AI data centers. McKinsey estimates that data centers could make up 12% of U.S. electric demand in 2030, compared to 3-4% today. $400 billion worth of data centers is under development in the U.S. with Microsoft and OpenAI reportedly considering another 5 GW AI data center dubbed Stargate at a cost of up to $100 billion. These data centers are extraordinarily energy intensive, and the cost of power represents 70% of their operating costs.
Solar is a poor substitute for nuclear in meeting power demand due to low energy density (power per unit of material) and low capacity factor (rarely operating near full capacity). To power a project like Stargate would require up to 25 GW of solar power, more than all the utility-scale solar capacity added last year, and take up to 200,000 acres. Nuclear provides an ideal solution. It is the densest energy source, and can operate at near total capacity, 24/7.
Amazon recently purchased a 100% nuclear-powered data center, while Oracle and Google are building data centers powered by small nuclear reactors.
As Mark Zuckerberg and others have stated, the energy requirements for training AI models could limit further development. If we don’t increase energy output and drive costs down, U.S. companies could fail to compete long-term.
In good news: a nascent bipartisan consensus exists on the need for more nuclear power construction, both to meet AI demand and catch up to China, which has a 15-year lead in developing advanced nuclear reactors. China is building designs based on U.S. concepts from decades ago, which are impossible to build in this country.
Congress passed the ADVANCE Act to accelerate new nuclear development, and former President Trump stated we need to double our energy production and invest heavily in new nuclear power. If elected, Trump has pledged to modernize the NRC and address regulatory obstacles to new nuclear development.
Still, two key impediments must be addressed: finance and labor. There aren’t easy solutions.
The first impediment is the capital intensity and high cost of capital for new nuclear projects. Per Department of Energy research, 21% of the Vogtle project’s cost increase came from financing costs due to construction delays (leading to inflated interest accrual). Up to 80% of a nuclear power plant’s levelized cost of electricity (LCOE), which is a metric that estimates the average cost of generating electricity over a plant’s lifetime.”
The second impediment is the lack of skilled labor that hurt the Vogtle project. The U.S. is short an estimated 550,000 construction workers, and up to 41% of the construction workforce will retire by 2031.
The lack of skilled workers could lead to a future where our ability to build complex infrastructure degrades and labor costs increase. We’ll pay more and get less in construction. Unfortunately, this trend has already begun. An analysis of Bureau of Economic Analysis data suggests the value added per worker in construction was about 40 percent lower in 2020 than in 1970, and construction sector productivity today is lower than in 1950.
The labor shortage and skills decline in construction have led to unprecedented decisions. A number of years ago, an EPC I worked with was contracted to build a large petrochemicals plant. The cost estimate for the project’s “cracker” units, which contain equipment to break down hydrocarbons into smaller molecules, was $2 billion. So, the decision was made to build those units in China. The Chinese contractor built 38 football-field-sized modules for under $400 million in total. The modules were shipped to Texas and assembled onsite like giant Legos. Unlike the Vogtle plant attempt, it worked beautifully.
China has thirty nuclear power plants under construction. Unlike Vogtle, these projects rely on a standard design, produced repeatedly, often by the same people in the same places. China saves a fortune and moves much faster.
The Chinese government recently approved the construction of 11 new nuclear reactors. China became the first country to build and operate a fourth generation nuclear reactor, proposed by American researchers in the 1940s, which offers advantages over the third-generation reactor that was built at Vogtle. It can’t melt down and can operate at much higher temperatures, ideal for providing carbon-free industrial process heat that drives over 20% of global energy demand.
The Chinese government plans to construct the world’s first thorium molten salt nuclear reactor to generate hydrogen. A molten salt nuclear reactor was first built by the U.S. in 1954. Such a reactor doesn’t require an expensive containment structure or fuel enrichment process, theoretically making it much less expensive than current nuclear reactor types. If China’s plants can be built cheaply enough, they could enable the country to produce synthetic natural gas at a cost competitive with or lower than fracked gas.
The U.S. has several fourth-generation nuclear reactor startups, like Oklo, which has agreements to provide gigawatts of power. Investors have poured over $5 billion into nuclear startups in the past two years. Earlier this year, the Bill Gates-backed startup Terrapower broke ground on a fourth-generation nuclear reactor project in Wyoming, which will become the world’s most advanced nuclear facility when completed.
Getting the Terrapower project to this stage has required one of the richest men in history to absorb the risk of cost overruns personally. It shouldn’t hinge on him. China’s ability to build projects at a much lower cost than the U.S. is aided by state-controlled banks financing Chinese nuclear projects at low interest rates (sometimes in the range of 2%). That can give nuclear power plants an LCOE of $47 per megawatt-hour instead of $100 per MWh if financed at the 10% rate common for U.S. projects.
A $47/MWh LCOE is cheaper than any renewable energy source in the U.S. accounting for intermittency. This understates the cost advantage because nuclear plants have an 80-year lifespan which DOE research indicates can safely be extended to 100 years. This creates the possibility of a 70-year period after construction loans have been paid off when electricity can be generated at a negligible cost.
If the U.S. wants to control the defining technologies of this century, it must address two issues. First, it needs private capital to embrace a longer-term view and accept more risk for potentially lower and uncertain returns, which requires a big financial commitment from the U.S. government. A major reason the U.S. technology sector leads the world is the availability of risk capital. There’s a lack of capital for building in the physical world, where costs and risks are higher. In nuclear, it’s even scarcer.
The second issue is the need to improve efficiency in building new nuclear power. The country that unlocks energy abundance and combines it with AI will win the future. There should be no other peripheral goals associated with U.S. industrial policy.
This will require extreme efficiency gains in every step of the building process, from permitting to engineering design to construction. There are ideas for a step-change improvement in building new nuclear projects. For example, there could be an intensive R&D effort to build more automated and advanced modular fabrication facilities to produce modules for standardized nuclear power plant designs with a high precision, like China. The problem with such an idea in a country without command-and-control is the need for a guaranteed pipeline of projects to justify the investment in modular manufacturing. Due to the sheer amount of capital and risk tolerance required — and the regulatory nightmare — an experimental fabrication yard is uninvestable in the U.S. at the moment.
We could change that, but change doesn’t happen overnight. There is an almost complete lack of R&D spending in construction, contributing to productivity declines in that industry. An ambitious R&D program focused on nuclear reactor design, knowledge transfer, automation, robotics, modularized construction, combined with a government effort to create a repeatable pipeline of projects at a lower cost of capital, could be transformative.
Still… such an approach would require profound changes in U.S. strategy and philosophy on the role of the state and private capital in industrial development. However, accepting our current approach to building nuclear projects as immutable means accepting a decline that has already begun. I say we shouldn’t accept it.