The rapid expansion of digital infrastructure, necessitated by artificial intelligence and cloud computing, has resulted in significant increases in global energy, water, and land consumption.

由於人工智慧與雲端運算的需要，數位基礎設施的快速擴張導致全球能源、水資源與土地消耗大幅增加。

The operational requirements of data centers exert substantial pressure on electrical grids. In the United States, the Department of Energy projects that these facilities will consume 12% of total electricity by 2028. Regional concentrations, such as in Virginia, demonstrate an acute impact, with data centers accounting for 26% of the state's electricity supply in 2023. The carbon intensity of this energy use varies by jurisdiction; the UN University identifies Hong Kong as having one of the most carbon-intensive grids globally due to a 67% reliance on fossil fuels. Consequently, power generation for U.S. data centers contributed 2.2% of national greenhouse gas emissions in 2023.

數據中心的運作要求對電網造成巨大壓力。在美國，能源部預測到2028年，這些設施將消耗總電量的12%。部分地區的集中情況顯著，例如在維吉尼亞州，數據中心在2023年佔該州電力供應的26%。不同司法管轄區的能源碳強度有所不同；聯合國大學指出，由於67%依賴化石燃料，香港是全球碳強度最高的電網之一。因此，美國數據中心的發電在2023年佔全國溫室氣體排放量的2.2%。

Water resource depletion constitutes a secondary critical concern. Thermal management via evaporative cooling necessitates vast quantities of potable water, with global AI infrastructure projected to consume between 4.2 and 6.6 billion cubic meters annually by 2027. Corporate disclosures from entities such as Google and Microsoft indicate a consistent upward trajectory in water usage, often within 'water-stressed' regions. For instance, Microsoft reported that 42% of its 2023 water consumption occurred in such areas. Academic research suggests a significant discrepancy between corporate reported figures and actual footprints, particularly regarding indirect water use for electricity generation.

水資源枯竭是另一個關鍵關注點。透過蒸發冷卻進行的熱管理需要大量飲用水，預計到2027年，全球AI基礎設施每年將消耗42億至66億立方米的水。Google與Microsoft等實體的企業披露顯示，水資源使用量呈現持續上升趨勢，且通常位於「水壓力」地區。例如，Microsoft報告指出其2023年有42%的水量消耗發生在此類地區。學術研究顯示，企業報告的數據與實際足跡之間存在顯著差異，特別是關於發電的間接用水。

Furthermore, the physical deployment of these facilities induces localized socioeconomic and environmental disruptions. The conversion of agricultural or green spaces into industrial zones—exemplified by proposals in Prince William County, Virginia—diminishes ecological benefits and alters community structures. Residents have reported adverse effects from acoustic pollution, with noise levels between 40 and 59 decibels impacting sleep and concentration. Additionally, the increased demand for electricity is projected to elevate residential utility costs; a Virginia legislative report suggests potential monthly increases of $14 to $37 by 2040, disproportionately affecting economically disadvantaged populations.

此外，這些設施的實體部署會引起局部社會經濟與環境的紊亂。將農業地或綠地轉換為工業區（如維吉尼亞州威廉王子郡的提案）會減少生態效益並改變社區結構。居民報告指出受到聲污染的負面影響，40至59分貝的噪音水平影響了睡眠與專注力。此外，電力需求的增加預計將提高住宅公用事業成本；一份維吉尼亞州立法報告指出，到2040年每月可能增加14至37美元，對經濟弱勢群體造成不成比例的影響。

Mitigation strategies involve the integration of renewable energy sources and the implementation of recycled water systems, which may reduce cooling energy demand by up to 29%. Institutional efforts, such as the Green Data Centres Practice Guide in Hong Kong, aim to standardize energy efficiency and regulate freshwater usage in cooling towers.

緩解策略包括整合再生能源與實施再生水系統，這可將冷卻能源需求降低最多29%。制度上的努力，例如香港的《綠色數據中心實務指南》，旨在將能源效率標準化並規範冷卻塔的淡水使用量。

Data center growth continues to scale globally, creating a persistent tension between digital advancement and environmental sustainability.

數據中心規模在全球持續擴大，導致數位進步與環境永續性之間存在持續緊張關係。