The United States is no longer debating whether artificial intelligence will reshape the economy; it already has. While fears of mass unemployment and runaway machines dominate headlines, AI has quietly become core infrastructure across the economy. The real risk is no longer technological failure, but whether the physical systems required to support this transformation can keep up.
After decades of electricity demand growth below 1% annually, U.S. utilities and grid operators are now forecasting sustained growth closer to 2-3%, with higher increases in certain regions. Artificial intelligence, data centers and advanced manufacturing are transforming electricity from a background utility into a binding constraint. A grid built for gradual expansion is now confronting a sharp change in demand.
AI infrastructure illustrates this shift. Modern data centers are industrial-scale facilities requiring continuous, ultra-reliable power, often operating at load factors above 90%. A single hyperscale data center can demand 100-300 megawatts of electricity, roughly equivalent to a mid-sized city.
The Department of Energy estimates indicate U.S. data centers accounted for approximately 4.4% of total electricity use in 2023 and could reach between 6.7% and 12% by 2028, implying a near doubling of demand within the decade. These loads are geographically concentrated and developed far more quickly than new generation and transmission infrastructure, which typically requires 10-15 years to permit and construct.
The constraint is not ambition, but scalable power. Wind and solar contribute meaningfully to the energy mix, but capacity factors of roughly 20-35% limit their ability to support always-on, high-density loads. Natural gas has filled much of the gap, yet fuel-price volatility, emissions pressures and pipeline constraints make it an incomplete long-term solution.
Nuclear energy offers a partial but critical answer. Existing nuclear plants operate at capacity factors above 90% and provide some of the most reliable power on the grid. Public resistance remains strong, shaped less by engineering realities than by decades of cultural imagery. For many Americans, nuclear power is still associated with Chernobyl, Fukushima or caricatures such as Homer Simpson, despite today’s stringent safety standards.
Preserving the existing nuclear fleet is among the fastest ways to stabilize supply. Looking forward, small modular reactors (SMRs) typically designed at 300 megawatts or less, offer a path to deliver firm power near industrial facilities or data centers, reducing transmission strain while improving reliability. Early projects, such as TerraPower’s Natrium reactor in Wyoming, highlight both the promise and cost challenges of first-of-a-kind deployment.
Power generation alone is insufficient. Expanding any energy source requires extensive physical infrastructure, transmission lines, substations, transformers and reactors, all of which depend heavily on copper. A January S&P Global study projects global copper demand reaching roughly 42 million metric tons by 2040, while supply is expected to fall short by nearly 10 million metric tons, a gap of approximately 25% even after increased recycling.
The challenge ahead is not technological limitation, but physical execution. The writing on the wall is clear: America is not producing enough power to sustain its economic transformation, and preserving nuclear capacity, advancing SMR deployment, expanding transmission and addressing material constraints will determine whether that gap widens.
Jackson Holder ‘29 is an economics and Government & Law double major at Lafayette College. He wrote this essay to encourage critical examination of America’s energy infrastructure.