AI Growth Faces Challenges from Environmental and Resource Concerns
人工智慧的發展面臨環境與資源問題的挑戰
更新於: 2026年6月14日 上午03:30
While artificial intelligence is often seen as a weightless, digital phenomenon, it is in reality a physically intensive industry.
儘管人工智慧常被視為一種無形且數位的現象，但實際上它是一個極度耗費實體的產業。
As AI models scale, they create a significant environmental footprint that challenges global sustainability goals.
隨著人工智慧模型的規模擴大，它們產生了巨大的環境足跡，對全球永續發展目標構成挑戰。
The energy-climate paradox is central to this issue; while training large models draws headlines, the constant daily usage—or inference phase—accounts for up to 90% of total energy demand.
能源與氣候的矛盾是此問題的核心；儘管訓練大型模型佔據了新聞版面，但持續的日常使用卻佔了總能源需求的九成之多。
This reliance on data centers often keeps the industry tethered to fossil-fuel grids.
這種對資料中心的依賴，往往使產業受制於依賴化石燃料的電網。
Data centers require vast amounts of fresh water for cooling, often competing with agricultural and municipal needs.
資料中心需要大量淡水來進行冷卻，這往往會與農業和市政需求產生競爭。
Furthermore, the hardware powering AI necessitates intensive mining for rare earth minerals, leading to deforestation and water pollution.
此外，驅動人工智慧的硬體需要對稀有礦物進行密集開採，進而導致森林砍伐與水污染。
This cycle is exacerbated by rapid obsolescence, as high-end GPUs are replaced every few years, creating a massive electronic waste problem.
這種惡性循環因硬體快速淘汰而加劇，因為高階繪圖處理器（GPU）每隔幾年就會被替換，造成了龐大的電子廢棄物問題。
Experts argue that to achieve net-zero goals, we must transition to 'efficiency by design' and implement global transparency in reporting.
專家主張，為了實現淨零排放目標，我們必須轉向「設計效率化」，並落實全球報告的透明度。
