代理型人工智慧（Agentic AI）的兴起大幅推升了能源消耗压力，其每次查询所消耗的电力远高于传统聊天机器人。随著各大科技公司全力推广能自主处理复杂任务的 AI 代理，若未来普及至日常搜寻规模，预估将新增相当于全美电网半数负载的庞大电力需求，对全球电网架构构成严峻冲击。
面对电力短缺与电网限制，资料中心开发商掀起天然气发电采购热潮，电力公司亦纷纷延后燃煤与燃气电厂的退役计划。摩根士丹利预估资料中心相关排放量将在未来几年翻倍成长，尽管有倡导者寄望 AI 带动各领域减碳效益，但目前眼前的碳排放激增却是迫切且真实的气候隐忧。（关键数字：2030）
The rapid expansion of artificial intelligence has sparked a fierce climate debate over its surging power consumption and carbon emissions. Although data centres currently account for a relatively small share of global emissions and basic query efficiency continues to improve, the International Energy Agency warns that aggregate electricity usage is climbing rapidly and projected to double within five years.
The emergence of agentic AI is vastly intensifying energy demands, as autonomous agents executing complex workflows consume exponentially more electricity per query than standard chatbots. As tech giants accelerate consumer and enterprise adoption, scaling these autonomous systems to mainstream levels could generate monumental energy requirements equivalent to nearly half of current US grid capacity.
To bypass grid bottlenecks, data centre operators are heavily procuring gas-fired power, prompting utilities to delay the retirement of fossil-fuel infrastructure despite ongoing growth in renewables. Analysts estimate that data centre emissions could quadruple by 2030, overshadowing speculative future efficiency gains and presenting immediate environmental risks for the global energy transition.