To meet the surging electricity demands of the AI industry, the U.S. government plans to invest $80 billion in restarting nuclear power construction, betting on Westinghouse Electric's AP1000 large reactor technology. This plan, supported by funds from the U.S.-Japan trade agreement, proposes building eight reactors across four sites, aiming to control the cost of each reactor at $10 billion through standardized construction, in an effort to break the long-standing "cost curse" of delays and budget overruns that have plagued U.S. nuclear projects.
To meet the surging electricity demand brought by the boom in the artificial intelligence industry, the U.S. government plans to invest $80 billion to restart the long-stagnant domestic nuclear power sector, placing its bets on Westinghouse Electric's "AP1000" large reactor technology, which has long been criticized for cost overruns.
This ambitious plan will leverage financial support obtained through a trade agreement between the U.S. and Japan, aiming to free the U.S. from the historical shadow of repeated delays and cost overruns in nuclear construction and establish energy dominance in the AI era.
According to The Wall Street Journal, using part of the proceeds from the U.S.-Japan trade agreement, the U.S. government announced in October a partnership plan with Westinghouse Electric and its owners, Brookfield Asset Management and Cameco. The plan proposes building eight AP1000 reactors at four sites, marking the largest scale of nuclear construction in the U.S. in decades.
This strategy marks a significant shift in U.S. nuclear policy: previously, market focus was largely on small modular reactors (SMRs), but now, driven by the massive energy consumption of AI data centers, large reactors with a single-unit capacity of approximately 1,100 megawatts are back in favor.
As the financial backbone of this initiative, Japan has pledged to invest up to $550 billion in critical U.S. infrastructure, with the current nuclear investment stemming from this commitment. Under the agreement’s structure, the U.S. government is not only a financier but will also profit from the project. Once Westinghouse Electric's profits exceed $17.5 billion, the U.S. government will receive 20% of the excess, and Westinghouse Electric could potentially be spun off as a publicly listed company with government ownership.
However, this massive bet faces extremely high execution risks. The last major U.S. nuclear project—the Vogtle nuclear plant in Georgia—ended up costing over $30 billion, more than double the initial budget, and was completed seven years behind schedule. Despite this, the Trump administration is attempting to revive this once investor-deterrent industry by controlling costs at around $10 billion per reactor through government backing and standardized construction.
Betting on AI Energy Consumption and Energy Dominance
As the AI boom breaks the decades-long trend of flat electricity demand in the U.S., power shortages have become a key bottleneck constraining AI development. Interim CEO of Westinghouse Electric, Dan Sumner, stated in an interview that these reactors will help the U.S. win the AI race and achieve the President's "energy dominance" agenda.
Sumner noted that Westinghouse Electric possesses the right technology and that this is a reasonable solution to address U.S. electricity needs. Previously, due to slow growth in electricity demand and cheap natural gas resulting from the shale gas boom, the so-called "nuclear renaissance" of the early 2000s in the U.S. failed. Analyst Hugh Wynne of Sector & Sovereign Research believes that when demand grows slowly, smaller reactors make more sense, but under current conditions, larger reactors better align with market demands.
If this plan is implemented, it will bring unprecedented certainty to Westinghouse Electric and its supply chain. Company executives stated that the proposed reactors would be sufficient to power a medium-sized city or a large AI data center.
Japan's Capital Support and Profit-Sharing Mechanism
The core of the deal lies in leveraging external capital to drive domestic infrastructure construction. This transaction is supported by Japan’s investment initiative in critical U.S. infrastructure. Wyatt Hartley, Co-President of Brookfield Renewable Partners, noted that the pool of funds obtained through trade agreements will benefit both American industry and Japanese manufacturing.
Hartley stated that the final structure not only directly benefits Westinghouse Electric but also provides the American public with an opportunity to participate in various profit-sharing schemes. When Brookfield and Cameco acquired Westinghouse Electric, which had emerged from bankruptcy protection, for approximately $7.88 billion, no valuation was assigned to its new reactor construction business, as the focus was solely on its fuel manufacturing and service operations. Now, with a shift in policy direction, its valuation logic may be redefined.
Challenging the 'Cost Curse' and Standardized Construction
Keeping the 'overnight cost' (excluding financing costs) of a single reactor below $10 billion means overcoming the industry’s long-standing fatal flaw: spiraling costs and schedule delays.
Units 3 and 4 of the Vogtle nuclear power plant are the only AP1000 units completed in the United States. Originally scheduled for completion in 2016 and 2017 with a budget of $14 billion, they eventually came online in 2023 and 2024, with total costs soaring above $30 billion. Severe cost overruns led Toshiba, Westinghouse’s former parent company, to file for bankruptcy in 2017. A similar project in South Carolina was abandoned in 2017 after spending over $9 billion. In contrast, China currently operates four AP1000 units, is constructing its own version, and has another 14 under construction.
Grant Isaac, President and Chief Operating Officer of Cameco, stated that the AP1000 no longer faces risks related to technology, licensing, fuel supply, or regulation; the only remaining risk is project execution. The current goal is to minimize risks through standardization, sequencing, and simplification—essentially repeating the same process over and over again.
Execution Details Remain Unclear
Despite the ambitious vision, uncertainties remain at the implementation level. Jacopo Buongiorno, a professor of nuclear science and engineering at MIT, pointed out that while an $80 billion scale is indeed suitable for accelerating nuclear power development, there is currently a lack of detailed information on site selection, participating companies, timelines, and labor. He believes that the current announcement is encouraging, but without substantive content, it remains a castle in the air.
Under an executive order issued in May this year, the Trump administration proposed a plan to begin construction of 10 large reactors by 2030. In addition to new projects, efforts are underway to restart existing facilities. Currently, Michigan, Pennsylvania, and Iowa are working to bring reactors back online. In South Carolina, utility company Santee Cooper is in talks with Brookfield to sell its partially completed AP1000 project; if the deal goes through, these units may be completed to exclusively power AI data centers.
Sumner believes that the U.S. government's intervention as a "second mover" sends a strong signal to manufacturers, suppliers, and construction companies that orders will follow, which is critical for building confidence and reducing costs.
