OPINION:
When I served at the CIA, analysts were often asked to prognosticate about a foreign adversary’s plans. Cognizant of the challenge of evaluating so many different dynamic variables, my CIA colleagues rightly hedged by highlighting levels of confidence for their analytical judgments.
It is one thing to analyze a particular historical event but entirely something else to predict the future, especially when we are dealing with imperfect or incomplete information.
Today, energy demand forecasts must be confounding U.S. government analysts. Utilities routinely overshoot, regulators discount, and, within a decade or less, the projections are rendered ineffective.
Yet something different is happening: The forecasts are being rapidly revised upward.
It is all about physics.
Artificial intelligence is relentlessly hungry for electricity. Training a single frontier AI model can consume as much power as a small city uses in a month. Running that model at scale, continuously, across millions of users results in a permanent, compounding load.
Every major technology company has, out of necessity, also become an energy company. The challenge for our national security is that America’s energy grid is not up to the task.
There is no comparison to cryptocurrency, which migrates to wherever electricity is cheapest, without regard for jurisdiction, latency or physical security.
AI workloads are the opposite.
Sensitive models running under government contracts that handle protected financial data and health records must be under U.S. jurisdiction. Frontier inference requires proximity to users to manage latency.
The data centers that house all this AI require physical security, which no offshore facility can guarantee.
The massive AI workloads require continuous, uninterrupted power. Data centers cannot run on an intermittent supply. Currently lacking storage solutions at the necessary scale, solar and wind, for all their cost and environmental advantages, cannot provide the around-the-clock baseload that an AI facility requires.
Natural gas can provide the continuity, but with an increasingly untenable carbon footprint and price volatility, long-term reliance on gas would be a gamble. That is why we should turn to nuclear power to augment our energy capacity.
The problem is that after a long pause in building nuclear reactors after the Three Mile Island nuclear accident in 1979, the U.S. has largely lost the ability to build nuclear reactors on time and on budget.
Vogtle Units 3 and 4, the first new American reactors in a generation, eventually came online, but only after the project ran years behind schedule and billions of dollars over budget, because the project-execution infrastructure had atrophied.
The specialized workforce, supply chains and institutional knowledge that come only from the regular construction of nuclear power plants have severely eroded.
Rebuilding those capabilities to boost our nuclear industry, essentially from scratch, is at best likely to take a decade. Yet the AI infrastructure build-out is happening now, and with it, the rising demand for reliable energy.
That is why the more sensible path would be a partnership with one of our closest allies.
Consider South Korea, which has maintained continuous reactor construction for more than 40 years. South Korea’s engineering workforce is highly developed, supply chains are intact, and state-backed engineering, procurement and construction firms have delivered reactors on schedule in conditions of real adversity, including the Barakah plant in the United Arab Emirates. It is the Arab world’s first commercial nuclear power plant, built on time and on budget in a desert.
A close U.S.-South Korea nuclear partnership would be a force multiplier, accelerating the rebuilding of U.S. nuclear capability through shared projects, workforce training and supply chain integration. Fortunately, the vehicle for this fruitful partnership already exists.
The U.S. Nuclear Regulatory Commission has already certified the APR1400 reactor design. With a demonstrated construction record across multiple sites, the APR1400 traces its lineage directly to American reactor technology. South Korean EPC firms can offer fixed-price contracts, thereby spurring growth and ensuring market stability.
Every APR1400 reactor built in the U.S. would result in domestic construction jobs and training for our nuclear labor force while adding much-needed depth to a domestic supply chain on the hook to support the nuclear industry over the horizon.
Time is of the essence. Any delay in building out our nuclear capacity risks lagging our AI infrastructure, while China and Russia push forward to dominate AI.
The Trump administration should expect bipartisan support from Congress as it charts the most effective, efficient and powerful energy strategy, on which our national security so deeply relies.
• Daniel N. Hoffman is a retired clandestine services officer and former chief of station with the Central Intelligence Agency. His combined 30 years of government service included high-level overseas and domestic positions at the CIA. He has been a Fox News contributor since May 2018. He can be reached at danielhoffman@yahoo.com. All statements of fact, opinion or analysis expressed are those of the author and do not reflect the official positions or endorsement of the CIA or any other U.S. government agency.
Follow the author
Please read our comment policy before commenting.