Commentary by Cy McGeady

Published October 4, 2024

The Power Line is a newsletter series that reviews key stories in the electric power sector-ranging from grid reliability, transmission issues, and emergent technology and business models. The Power Line will highlight the connections and relevance between the sector's technical, local, and often esoteric stories to high-level public policy conversations on climate and energy taking place in Washington, D.C.

Artificial intelligence (AI) is changing the game for nuclear power, and potentially, the direction of the U.S. electric power sector.

Last week, Constellation, a major power plant operator, announced that it will restart an 835 megawatt (MW) nuclear reactor at Three Mile Island in Pennsylvania with Microsoft as the sole contracted buyer of the unit's power. The unit was decommissioned five years ago for economic reasons, and its planned restart in 2028 is the first time a reactor has been restarted to serve datacenter demand. Another reactor restart is underway at the Palisades nuclear plant in Michigan, and reports indicate that NextEra is considering a restart at the Duane Arnold nuclear plant in Iowa to serve datacenter demand growth. These restarts follow Amazon's purchase of a datacenter powered by a nearby operating nuclear plant.

The speed and scale of the turn to nuclear by large tech firms is remarkable. Big tech firms have been the most ambitious, innovative, and sophisticated buyers in the power sector for decades. This shift indicates that leadership is heading in a new direction.

There isn't much more capacity to be gained from restarts. The real question now is whether industry can move beyond restarts and build a new reactor, only this will mark the true start of a U.S. nuclear renaissance.

The Problem Statement

By some estimates, the United States needs to find power to serve 50 gigawatts (GW) or more of datacenter demand by 2030. Another indication of the scale of the energy demand is the reported proposal by OpenAI to build multiple 5 GW datacenter campuses across the nation.

The surge in electricity demand is real; now a new class of anchor buyers has emerged and the creation of multibillion-dollar AI infrastructure funds and pledges by banks indicate that capital is available for nuclear projects. New nuclear construction is expensive, but the large tech firms are clearly willing to pay a premium for clean-firm power supply. The Department of Energy (DOE) estimates that a newly built reactor could be delivered at a $100 per megawatt-hour (MWh) price point, which is comparable to the price Microsoft is reportedly paying in the Three Mile Island deal. All of this indicates that the market may be inclined to take on new nuclear builds independent of further government support. It should be noted that considerable federal support already exists for reactor construction including a 40 percent investment tax credit and access to low-cost DOE Loan Programs Office financing.

However, the real challenge is not the price tag, but risk. The recently completed Vogtle reactor projects came in $17 billion over budget. Neither tech firms nor utilities are interested in taking on such risks to their balance sheets. Given the scale of these projects, cost overruns can bankrupt project developers. Likewise, after the Vogtle cost overruns, even pronuclear state regulators have no interest in shifting this cost risk onto ratepayers.

Given the scale of energy volumes required and the vast strategic value of securing U.S. leadership in both AI and nuclear technology, the current equilibrium arguably represents a market failure. Importantly, this outcome is at least partially a function of poor regulatory policy on the part of the Nuclear Regulatory Commission (NRC), which makes construction and engineering work very costly. All of this plausibly establishes a case for federal government intervention to bridge the risk gap and restart the U.S. nuclear industry.

What Would a Federal Policy Response Look Like?

Let's assume this equilibrium holds, the market fails to deliver new nuclear investment, and federal leadership decides that intervention to relaunch the nuclear power sector is a strategic necessity. What's the policy model?

The federal government would need to set a target of 5-10 new nuclear financial investment decisions (FIDs) within the next 24-36 months. A fixed orderbook creates investment certainty and the compressed timeline is required to provide utilities responding to urgent load interconnection requests with a viable nuclear pathway. An ambitious target like this requires utilizing existing assets, investments, and technology.

On technology, the best option is to start with AP1000 reactor technology used at the Vogtle 3 and 4 projects. Leveraging the expensively won design experience, labor force training and supply investments will lead to faster, lower risk, and lower cost project delivery. For example, one study found that the next AP1000 reactor could achieve 25 percent cost reductions or more. It's clear that tech firms are exploringinvestments in small modular and other advanced reactor technologies, but within the time constraint set above none of these designs are ready for commercial scale.

On siting, the initial homework has already been done. The DOE has found that 54 operating and retired nuclear power plant sites have room to host one or more AP1000 units, representing roughly 60 GW of potential capacity. Detailed analysis by the Idaho National Lab has identified and ranked dozens of sites suitable for near-term deployment, leveraging already initiated or completed NRC licensing, site acquisition and preparation, and existing transmission infrastructure. Crucially, research shows communities that already host nuclear power plants are far more likely to support additional nuclear expansion, which greatly de-risks permitting processes versus greenfield nuclear deployment. Congress should direct the NRC to institute a fast-tracked licensing process for new nuclear projects at these types of sites.

The trickiest question is on contracting. Who will bear the financial risks of delays and cost overruns? This calls for innovative public-private partnership models. The federal government should position itself to take on risk and insulate state ratepayers; this position can be further leveraged to bring state policymakers on board by ensuring state ratepayer access to portions of new reactor capacity on a risk-free basis. Whatever the model, the incentives must be properly aligned to avoid moral hazard, risk will need to be spread across parties, including the project sponsors (likely utilities), construction firms, and the tech companies who benefit from access to new clean firm power. There is work to do on the details here, but assuming the federal funds can be found, there is a path forward on risk-sharing models.

Importantly, this program leaves large volumes of electricity demand to be filled by other resource classes, including renewables. Keep in mind that beyond meeting datacenter demand, the U.S. power sector must also meet growing demand from electric vehicles, semiconductor fabs, and other reshored industries. It also must backfill the 90+ GW of aging coal-fired power plants that will retire in the coming decades. This building program would tail off just as new reactor technologies are breaking through the demonstration phase and ready to achieve commercial scale. These new designs will benefit from the broad sectoral momentum, including supply chain and labor force expansion initiated by this building program.

The Need for a Comprehensive Federal Strategy

Developments in AI and tech firms' power procurement are moving fast, and the market may yet deliver new nuclear investment decisions. Regardless, developments in AI only reinforce the case for a broad national electricity demand growth strategy. The best-case scenario for nuclear construction would see new capacity additions hitting the grid in the early 2030s; until then the power sector will be under considerable duress.

Crucially, a rapidly growing electricity sector needs more high-voltage transmission even if the resource mix tilts more toward nuclear than it appeared five years ago. For example, the Tennessee Valley Authority built its 500-kilovolt backbone transmission network across the southeast United States to extract maximum value from its coal and nuclear generation fleet. Any ambitious federal program on scaling nuclear must be coherently matched with strong federal policy action on transmission and permitting reform to deliver maximum strategic value to the nation.

The commercial developments at Three Mile Island indicate the potential for an era of significant investment, innovation, and expansion in the electric power sector. This result is crucial if the United States is to retain its strategic energy, economic, and technology advantages in the decades to come. It's time for federal policymakers to match the moment and deliver a policy environment that enables rather than constrains this potential.

Cy McGeady is a fellow with the Energy Security and Climate Change Program at the Center for Strategic and International Studies in Washington, D.C.

Programs & Projects