America's electricity grid is facing a demand shock unlike anything in a generation, and the question of how to meet it is reshaping the politics of energy. Data centers are projected to consume roughly 8 percent of U.S. electricity by 2030, up from about 3 percent in 2022, according to Goldman Sachs Research. The International Energy Agency has projected that global data center electricity consumption will more than double by 2030, with the United States accounting for the largest share of that growth. Hyperscale operators need what the grid has struggled to provide: massive quantities of firm, carbon-free, round-the-clock power on timelines measured in years, not decades.
The scramble is already visible. Microsoft has signed a 20-year power purchase agreement to restart a reactor at Three Mile Island. Amazon Web Services acquired a data-center campus co-located with Pennsylvania's Susquehanna nuclear plant. Google has signed agreements with small modular reactor developers and is exploring direct nuclear procurement for its facilities. Meta has issued requests for proposals seeking nuclear capacity. The common thread: the hyperscalers have concluded, after years of renewable-first procurement, that wind and solar — even paired with batteries and natural gas — cannot by themselves deliver the reliability profile that a 200-megawatt AI training cluster requires. They are turning to nuclear because the math of AI load growth has forced them to.
A domestic industry that forgot how to build
The problem is that the American nuclear industry is not, at the moment, in a position to meet that demand quickly. Long construction times and cost overruns have hampered recent U.S. reaction construction. The troubles are not a matter of bad management or a flawed design. They are structural. Until very recently, the United States had not built a reactor from scratch in decades. The specialized supply chain for large nuclear forgings had atrophied. The skilled construction workforce — welders rated for nuclear-grade pipe, quality-assurance inspectors, project engineers experienced in first-of-a-kind nuclear work — had retired or moved on. Utilities grew wary, domestic companies proved reluctant to take on the prime engineering-procurement-construction role, and commercial viability became an open question. The institutional memory of how to deliver a large reactor on schedule and on budget had largely been lost.
Industry observers from across the political spectrum have reached similar conclusions. The Breakthrough Institute, the Nuclear Energy Institute, and analysts at MIT's Energy Initiative have all documented that recent U.S. nuclear construction has suffered from what economists call the "forgetting curve" — costs rising, rather than falling, with each successive project, because no project follows closely enough on the last to preserve learned efficiencies. Rebuilding that capacity is achievable, but it will take years. The data-center buildout cannot wait.
A Korean reactor with an American lineage
The APR1400 reactor has entered this conversation because it offers an unusual combination of familiarity and delivery record. The reactor is designed and built by South Korea but its technological ancestry is American: it is derived from the Combustion Engineering System 80+, a U.S. design that was certified by the Nuclear Regulatory Commission in the 1990s. The APR1400 received NRC design certification in 2019, making it one of the few large-reactor designs cleared for U.S. deployment.
The numbers are striking. Construction costs for an APR1400 run roughly 26 percent below comparable current U.S. options; on a per-kilowatt basis, the gap widens to approximately 47 percent. Construction lead time runs about 9 percent shorter. But the more consequential fact is the track record. Four APR1400 units at the Barakah site in the United Arab Emirates, executed by a Korean consortium under a fixed-price engineering-procurement-construction contract signed in 2009, came online between 2021 and 2024 — with successive units delivered on an accelerating schedule that ENEC reports represents a 40 percent improvement from Unit 1 to Unit 4. Korean firms have signaled a willingness to maintain that same fixed-price discipline for U.S. deployment, directly addressing the utility trust gap that recent domestic projects have opened.
To be sure, there are legitimate questions. The APR1400 is a gigawatt-scale reactor in a moment when much of the American policy conversation has shifted toward small modular reactors; many argue that the future lies in smaller, factory-built units rather than another generation of large plants. And a reactor built in the UAE is not a reactor built in Georgia or Texas; American labor rules, permitting regimes and seismic standards would add cost and time to any U.S. deployment.
The alliance dimension
Energy infrastructure is increasingly treated as national-security infrastructure, and when that is true, the identity of the partner matters. The U.S.–South Korea alliance is more than seventy years old. South Korea's civilian nuclear industry itself began with American technology: the Kori-1 reactor, which came online in 1978, represented an early export of U.S. nuclear capability to a key Asian ally.
What began as seeds of U.S. nuclear technology half a century ago has grown into a world-class Korean industry. Korean firms bring the intact supply chain, the deep bench of construction experience and the workforce depth; a genuine partnership would mean resolving domestic bottlenecks, revitalizing equipment and fuel supply lines, and training a new generation of American nuclear professionals in the process.
The alternative, in the global nuclear export market, is stark. Russia's Rosatom has signed contracts for at least 22 reactor units abroad. China's state nuclear companies have exported several reactors and plan to build 30 overseas by 2030. Analogies to AUKUS, to the CHIPS Act's semiconductor partnerships with Taiwan and South Korea, and to allied cooperation on critical minerals suggest a template: democracies pooling industrial capability to keep strategic technologies in friendly hands.
Whether the APR1400 ultimately gets built on American soil will depend on utility economics, state-level permitting and the appetite of the hyperscalers now driving nuclear procurement. But in a moment when the United States needs power quickly and is racing to catch up on its own, a reactor of American lineage, delivered on time by an ally, is a serious option.
