What happens when the world’s most transformative technology becomes its most energy-intensive? Global data center electricity consumption reached 460–490 TWh in 2025 — a 17% year-over-year surge that dwarfs overall global electricity demand growth of 3%. According to the International Energy Agency, AI-focused facilities drove a 50% increase in that figure, and consumption is projected to double to 945 TWh by 2030 — equivalent to Japan’s current electricity demand. This documentary examines the full scientific and economic dimensions of the AI Data Center Energy Crisis. We analyze the carbon cost of AI model training, the thermodynamic constraints of large-scale GPU computation, the cascading stress on regional power grids in Northern Virginia and Dublin, and the trillion-liter annual water burden placed on cooling infrastructure in water-stressed geographies. We investigate the race for structural solutions: the revival of nuclear fission at Three Mile Island, corporate small modular reactor procurement pipelines that grew from 25 GW to 45 GW in 18 months, and the Jevons Paradox — the counterintuitive mechanism by which unprecedented gains in per-task AI energy efficiency are outpaced by exponential growth in total AI usage. The decisions made by governments, engineers, and technology companies in the next five years will determine whether artificial intelligence fulfills its extraordinary scientific promise — or becomes constrained by the very infrastructure required to run it.