The AI data center e-waste problem is huge — and getting bigger

A report published by the Basel Action Network expands previous calculations of AI-related electronic waste by including the full array of equipment that supports data centers, not just servers and accelerators. Taking power distribution, cooling and backup systems, networking gear and a broad category BAN calls “AI Waste Contagion” into account, the group finds AI’s hardware footprint could be far larger than earlier studies indicated. BAN quantifies the problem by estimating about 70,000 metric tons of e-waste per gigawatt of data-center capacity and cites a McKinsey projection that global capacity could reach as high as 219 gigawatts by 2030. Using that basis, the group concludes equipment retired because of AI between 2025 and 2050 would amount to roughly 395–617 million metric tons, or 8.6–13.1 million metric tons annually — far above prior published ranges. The report’s authors say earlier estimates were conservative because they focused mainly on servers and accelerators and omitted approximately 87 percent of a data center’s electro-mechanical infrastructure. BAN also highlights a spillover effect in which AI-driven advances could shorten the lifespans of telecommunications hardware and personal devices, increasing the volume of discarded electronics. Environmental and human-health concerns underlie BAN’s warning. Global formal collection and recycling currently capture less than a quarter of the roughly 68.3 million metric tons of e-waste generated each year. Much of the remainder moves into informal recycling channels where burning and burial can release toxic metals; investigations have shown U.S. recyclers still export e-waste abroad, and the United States has not ratified the Basel Convention that restricts hazardous-waste trade. BAN’s founder, Jim Puckett, urged companies and governments to plan for the coming surge or risk turning the AI buildout into a much larger toxic-waste crisis.