[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"tag-電力":3},{"tag":4,"articles":10},{"id":5,"name":6,"slug":6,"article_count":7,"description_zh":8,"description_en":9},"797c3c24-583d-4fe3-b187-1e1a9b158042","電力",3,"電力已成 AI 基礎設施的核心瓶頸：大型資料中心、GPU 叢集與訓練任務不只吃晶片，也直接拉高用電、併網與自建發電需求。這個主題關注的是算力擴張背後的能源成本、電網壓力與政策博弈。","Power has become a hard constraint on AI infrastructure. As data centers scale for GPU clusters and training runs, the debate shifts to grid capacity, on-site generation, water use, and the policy tradeoffs that determine where compute can actually grow.",[]]