America’s nuclear paradox: Yesterday’s reactors are being resurrected to power the future. For example: At Palisades, Michigan, engineers rush to repair steam generators before an October 2025 restart, while at Three Mile Island—where a 1979 meltdown once buried the country’s postwar nuclear ambitions—Microsoft has committed billions for carbon-free electricity to feed its voracious artificial-intelligence operations. Yet regulatory hurdles and technical challenges threaten to undermine this flowering nuclear renaissance, much less the building of new reactors, before it properly begins. That would be a missed opportunity that America, with its surging electricity needs and clean energy aspirations, can ill afford.

This latest stage in America’s information revolution is consuming electricity at a startling rate. A recent Richmond Fed research essay, “Has Nuclear Energy’s Time Come?,” reveals the scale of this power-hungry transformation. Virginia’s data centers already devour electricity equivalent to that of two million households—with demand set to quadruple within 15 years. The situation appears even more acute when examining AI’s contribution. Lawrence Berkeley National Laboratory forecasts suggest data centers could claim 12 percent of America’s total energy consumption by 2028, nearly tripling from 4.4 percent in 2023.

Against this backdrop, nuclear power’s reliability makes it a great fit to meet the growing energy challenge. Nuclear plants operate at full capacity 93 percent of the time versus 36 percent for wind and 25 percent for solar. This disparity helps explain why 20 nations, America included, have committed to tripling nuclear capacity by mid-century.

But formidable obstacles threaten this atomic vision, the Richmond Fed explains. The nuclear workforce has collapsed, with engineering programs declining from nearly 100 in 1978 to barely 30 today. Regulatory processes remain painfully slow, while upfront construction costs deter investment. Every recent US nuclear project has faced crushing delays and budget overruns. 

In a recent podcast chat with me, economist Eli Dourado explains that nuclear power is expensive primarily due to lack of iteration and scaling. If regulatory barriers were reduced, the industry could develop many small test reactors to quickly validate different designs empirically and generally simplify reactor designs. Once optimized designs are developed, they could then be replicated efficiently. Additionally, a flourishing nuclear industry would establish more robust supply chains and expand the pool of nuclear expertise.

If we can get the regulation right, then plenty of reason for optimism, as the bank notes. Small modular reactors may offer standardized factory construction at lower costs. Advanced large reactor designs with safer fuel cycles could overcome historical concerns by the public. Meanwhile, Commonwealth Fusion Systems is developing America’s first commercial fusion plant in Virginia, though commercial viability remains uncertain.

The question isn’t whether nuclear power works. It safely provides 20 percent of America’s electricity and half its carbon-free power today. Rather, the bank concludes, it’s whether the industry can overcome decades of fragmentation and regulatory paralysis quickly enough to seize this fleeting opportunity. That, before utilities commit to easier but higher-emission alternatives like natural gas to meet exploding energy demand in an increasingly AI-powered economy.