The AI data center e-waste problem is huge — and getting bigger
The AI data center e-waste problem is huge — and getting bigger
AI 数据中心电子垃圾问题巨大，且正日益严重
E-waste from the AI boom has been vastly underestimated, a new report warns. By 2050, it could become enough trash to fill 23 million shipping containers — roughly enough 40-foot containers to circle the world six times if lined up in a row. 一份新的报告警告称，人工智能（AI）热潮产生的电子垃圾被严重低估了。到 2050 年，这些垃圾的总量足以填满 2300 万个集装箱——如果将这些 40 英尺长的集装箱首尾相连，足以绕地球六圈。
It’s a significantly higher estimate of AI’s e-waste than previous studies have found because the authors of the new report factor in all the infrastructure needed to support servers in data centers. The broader scope provides a clearer picture of junk AI leaves behind, according to the nonprofit Basel Action Network (BAN), which published the paper today. 这一对 AI 电子垃圾的估算远高于此前的研究，因为该报告的作者将数据中心支持服务器所需的所有基础设施都纳入了考量。发布该报告的非营利组织“巴塞尔行动网络”（BAN）表示，这一更广泛的统计范围更清晰地呈现了 AI 所遗留的垃圾问题。
”AI may feel weightless, but every model depends on an enormous amount of highly specialized, cutting edge hardware,” Jim Puckett, founder and chief of strategic direction at BAN, says in a press release. “If companies and governments do not begin planning for this new waste tsunami, today’s AI buildout could become an even more cataclysmic toxic waste crisis than we are already experiencing.” “AI 可能让人感觉轻盈无形，但每一个模型都依赖于海量高度专业化、尖端的硬件，”BAN 创始人兼战略方向负责人 Jim Puckett 在新闻稿中表示，“如果企业和政府不开始为这场新的垃圾海啸做规划，当下的 AI 建设可能会演变成一场比我们目前所经历的更为灾难性的有毒废物危机。”
Less than a quarter of the 68.3 million tons of e-waste created each year worldwide is formally collected and recycled. Most e-waste is ultimately dumped into shadowy “informal” waste collection where burning or burying equipment can expose workers and the local environment to toxic materials like lead and chromium. 全球每年产生的 6830 万吨电子垃圾中，只有不到四分之一被正式收集和回收。大多数电子垃圾最终流入了隐秘的“非正规”回收渠道，通过焚烧或掩埋设备，使工人和当地环境暴露在铅、铬等有毒物质中。
The US, with far more data centers than any other country, has yet to ratify the Basel Convention established to limit the international trade of hazardous wastes. US recyclers still ship e-waste abroad, investigations have found, where it often slips into “backyard recycling.” Millions of children forced to work or live nearby face health threats from that informal recycling, according to the World Health Organization. 美国拥有的数据中心数量远超其他任何国家，但至今尚未批准旨在限制危险废物国际贸易的《巴塞尔公约》。调查发现，美国的回收商仍在将电子垃圾运往国外，这些垃圾往往流入了“后院回收”作坊。据世界卫生组织称，数百万被迫在附近工作或生活的儿童正面临这种非正规回收带来的健康威胁。
BAN predicts e-waste generated globally will triple in amount to up to 211 million metric tons a year by 2050. It attributes about 15 to 20 percent of that garbage to AI. Unlike previous AI e-waste estimates that focused on servers and GPUs, BAN expands its count to include other categories of equipment: power supply and distribution, cooling systems, backup power, and networking equipment. BAN 预测，到 2050 年，全球产生的电子垃圾总量将增加两倍，达到每年 2.11 亿公吨。其中约 15% 到 20% 的垃圾归因于 AI。与以往仅关注服务器和 GPU 的 AI 电子垃圾估算不同，BAN 将统计范围扩大到了其他类别的设备：电力供应与分配、冷却系统、备用电源和网络设备。
Notably, BAN also counts what it calls “AI Waste Contagion” — a wide-ranging category that incorporates everything from telecommunications infrastructure to personal devices likely to become obsolete and replaced earlier as a result of advances in AI. 值得注意的是，BAN 还统计了其所谓的“AI 垃圾传染”（AI Waste Contagion）——这是一个涵盖广泛的类别，包括从电信基础设施到因 AI 进步而可能提前过时并被更换的个人设备等一切物品。
Counting it all up, BAN expects 70,000 metric tons of e-waste per gigawatt of data center capacity. It cites a McKinsey projection of total data center capacity reaching up to 219GW by 2030. 综合计算，BAN 预计每吉瓦（GW）数据中心容量将产生 7 万公吨电子垃圾。报告引用了麦肯锡的一项预测，即到 2030 年，全球数据中心总容量将达到 219GW。
Considering AI-related electronic equipment retired between 2025 and 2050, BAN concludes that’ll add up to between 395 to 617 million metric tons of e-waste by the middle of the century. It breaks down to about 8.6 million to 13.1 million metric tons of equipment retired each year as a result of AI. 考虑到 2025 年至 2050 年间退役的 AI 相关电子设备，BAN 得出结论：到本世纪中叶，这将累计产生 3.95 亿至 6.17 亿公吨的电子垃圾。这意味着每年因 AI 而退役的设备量约为 860 万至 1310 万公吨。
Previous estimates of AI’s e-waste have been more conservative. But by focusing on servers and accelerators, past studies miss about 87 percent of a data center’s electro-mechanical infrastructure, according to BAN. 此前对 AI 电子垃圾的估算较为保守。但 BAN 指出，由于过去的研究仅关注服务器和加速器，它们忽略了数据中心约 87% 的机电基础设施。
A separate study published in 2024 predicted that e-waste from AI could reach between 1.2 million and 5 million tons by 2030. Another study published in February found that AI servers could create between 131,000 and 225,000 tons of e-waste annually by the end of the decade — an amount that’s still comparable to all the e-waste produced by a country the size of Denmark. 2024 年发表的另一项研究预测，到 2030 年，AI 产生的电子垃圾可能达到 120 万至 500 万吨。今年 2 月发表的另一项研究发现，到本十年末，AI 服务器每年可能产生 13.1 万至 22.5 万吨电子垃圾——这一数量相当于丹麦这样一个国家产生的全部电子垃圾。