Three Mile Island is not generally mentioned cheerfully. A nuclear power plant capable of generating 1.6 Gigawatts of energy, it was in 1979 home to the worst nuclear accident in American history. While nobody is known to have died as a result, the area around the plant was irradiated for years. The cleanup cost around a billion dollars. As it happens, Three Mile Island is also due to reopen in 2028. Microsoft is investing in the plant in exchange for access to its energy for twenty years.
Three Mile Island is not the first nuclear power plant to be procured by a big tech company. Amazon this year bought a data center reliant upon the Susquehanna power station in Pennsylvania. Meta this month announced it will release a request for proposals (RFP) for nuclear energy developers to build reactors across the country.
Big tech’s courting of nuclear energy plants is the culmination of two trends reshaping technology and society. The first is decarbonization. Each company has pledged to reach carbon neutrality by at the latest 2040, with Meta and Microsoft promising net zero by 2030.
The second is the arms race in artificial intelligence. AI, due to its reliance on data centers, is extremely energy intensive. Goldman Sachs forecasts that data centers will grow to use 160% more energy by 2030, representing 8% of the US’s total energy use. This will represent the greatest increase in energy use since the start of the century.
Both of these trends will squeeze energy systems. Decarbonization requires that the US must eventually entirely replace the 84% of its energy which is produced via fossil fuels. Yet AI demands that the US produce more energy than ever. The energy which data centers use will be particularly difficult to wean off fossil fuels because data centers require reliable, consistent production of energy for 24-hour operation. Solar and wind energy, the two most important short-term sources of renewable energy, are highly variable in their energy production. If it is cloudy or there is no breeze, there is less energy.
Nuclear energy conversely, according to the Department of Energy, is the most reliable energy source, outperforming even fossil fuel power plants by a factor of two. Nuclear energy also produces no carbon in its operations, and across its life cycle is comparable to wind. It produces a third of the life-cycle emissions of solar.
Nuclear thus appears like it could solve all the problems putting strain on our energy systems. Yet, like all good things, there’s a catch (or actually several).
One may think that this is primarily the problem of nuclear meltdowns. This certainly is what occupies the public’s imagination. Yet while the crises at Fukushima and Chernobyl particularly were without a doubt grave tragedies we should not repeat, nuclear energy has not necessarily had worse impacts than other energy sources.
Considering the number of deaths per Gigawatt of energy produced, coal and oil are roughly 1000 times more lethal than nuclear. This is because the air pollution from fossil fuels harms more people with a far greater certainty. While this damage may not promote the same visceral reaction, it is important to consider.
Indeed, essentially all of nuclear energy’s historic deaths came from the Chernobyl meltdown, which was an anomaly resulting from extreme Soviet mismanagement. It is unlikely that this would happen in the West. No nuclear accident has happened in the US since 1959, and nobody has ever died from such an accident. This is the case even though nuclear reactors produce 18% of US electricity.
A second and more real problem is nuclear waste. Nuclear reactors leave behind a number of radioactive isotopes which must be dealt with. While most nuclear power plants currently hold most waste on-site, this is a mere stopgap. It is widely agreed that nuclear waste should be held in a deep geologic repository, for which there are no clear candidates after the termination of the Yucca mountain project in 2010. This is a problem the federal government will have to solve in the coming years. Yet the system is not going to implode in the medium term. This is not a reason to wholly dismiss nuclear energy as an energy source.
‘
The final problem is probably the most significant. Nuclear reactors are really damn expensive to build.
It is estimated that 60-80% of the total cost of nuclear power plants come from construction. This compares to only about 25% for natural gas plants. Most recent attempts to build nuclear plants have cost more and taken longer than initially expected. The Vogtle power plant in Georgia was expected to be operational in 2016. It only went online last year, after running 17 billion dollars over budget.
There are two reasons for this high cost. The first is that they’re extremely heavily regulated. While regulation of nuclear plants is without a doubt necessary, these regulations in some places overstep the line. The bigger problem is a lack of repetition. There is little standardization in the production of nuclear power plants, with each one being built essentially “from scratch”. The upshot is that nuclear scientists are constantly reinventing the wheel, and not able to harness economies of scale.
Small modular reactors (SMRs) promise to change this. Several startups and larger companies, from Google-backed Kairos Power to legacy incumbent Westinghouse Energy, promise to build compact reactors with repeatable parts that can be produced at scale. This promises to address nuclear energy’s perennial price problem.
There are reasons to take the promise of SMRs with a grain of salt. SMRs have been proposed since the late 1990s and early 2000s, but are yet to bear real fruit. However, this time may be different. Gen-III+ reactors, while slightly more expensive than initially believed, are in early stages of deployment. Big tech may just be the catalyst to make them a success. In the past, SMRs have faced a kind of coordination problem, where no one player wanted to be the guinea pig to fund them while they worked out their initial kinks. Yet the unquenchable energy demands of data centers may cause big tech to step into this role, providing overwhelming demand that allows SMRs to sort out their teething problems.
The revitalization of nuclear energy would be an extremely beneficial step in the US’s path to decarbonization, and provide a more consistent supplement to renewable energy to lead our systems towards net zero.
There are a few things the US can do to help make this dream a reality. The first is to rethink the regulation of nuclear energy – to try to slash wasteful bureaucracy while keeping the essential safety features. The second is to encourage onshore or friendshore production of High-Assay Low-Enriched uranium, a grade of uranium that is used in nearly all nuclear reactors but which presently is only produced commercially by Russia. The final step is to look into solutions for the long-term disposal of nuclear waste.
None of these issues are insurmountable. Nuclear energy has incredible theoretical upside. It may both literally and metaphorically be nuclear reactors' time to shine.
Comments