Solving AI’s Energy Problem with Kathryn Huff

Nuclear power is often dismissed as dangerous and dirty, but when you measure deaths per terawatt-hour of electricity generated, it is one of the safest energy sources we have—comparable to solar and wind, and orders of magnitude safer than coal or natural gas. That counterintuitive fact, grounded in the sheer energy density of fission, is the starting point for a wide-ranging conversation about whether nuclear energy can solve the growing power demands of AI data centers and other 24/7 industrial loads.

The Safety Math That Changes the Conversation

The public perception of nuclear power is shaped by high-profile disasters like Chernobyl, Three Mile Island, and Fukushima. But Kathryn Huff, a nuclear engineering professor and former Department of Energy official, argues that the right metric for comparing energy sources is deaths per terawatt-hour generated. By that measure, nuclear is near the bottom of the list—slightly below solar and wind, and far below fossil fuels. The reason is straightforward: nuclear plants produce enormous amounts of electricity from a tiny amount of fuel, so even when rare accidents are factored in, the death rate relative to energy output is extremely low. Coal, by contrast, kills people regularly through mining accidents, particulate pollution, and respiratory disease. The takeaway is not that nuclear is risk-free, but that the risks are diffe