A single fallen power line in Loudoun County, Virginia, knocked out a major data center for hours, exposing a critical vulnerability in AI's rapid expansion. While headlines focus on chip stocks and model breakthroughs, the electricity grid is quietly becoming the industry's biggest bottleneck. Data center power demand is projected to surge from 2% to 8% of global consumption by 2030, straining infrastructure already under pressure. This episode unpacks the energy crisis hiding in plain sight: why chip efficiency gains aren't enough, how hyperscalers are scrambling to secure nuclear and renewable sources, and why the next AI trade might be in utilities rather than semiconductors. Lucas and Luna explore solutions from underwater data centers to small modular reactors, and ask whether AI's future depends as much on voltage as on algorithms. With the SOXX semiconductor index hovering around 527, the market is betting on demand—but can the grid deliver?