A recent publication highlights a significant increase in electronic waste (e-waste) driven by the proliferation of artificial intelligence. Projections suggest that by 2050, the e-waste produced could fill about 23 million shipping containers, equivalent to circling the globe six times with these containers.
Broadening the Scope of E-Waste Impact
The report, released by the Basel Action Network (BAN), suggests that previous assessments have underestimated the e-waste impact of AI due to overlooking the infrastructure supporting servers in data centres. Jim Puckett, founder and strategic director at BAN, stresses the importance of addressing this impending issue to prevent a potential toxic waste disaster.
A mere 23% of the 68.3 million tons of e-waste generated globally each year is formally recycled, with the remainder often ending up in informal recycling settings. Here, hazardous materials like lead and chromium are released during improper disposal methods, endangering local communities and workers.
US Inaction and Global Consequences
The report highlights that the United States, despite hosting more data centres than any other nation, has not ratified the Basel Convention, meant to control hazardous waste trade. Consequently, e-waste from U.S. recyclers is frequently exported, with much ending up in unsafe “backyard recycling” abroad, posing health risks, especially to children, as noted by the World Health Organization.
Future E-Waste Projections
BAN estimates that global e-waste could reach 211 million metric tons annually by 2050, with AI contributing 15 to 20% to this figure. Unlike prior studies concentrating on servers and GPUs, BAN’s analysis includes additional equipment such as power and cooling systems. It also introduces “AI Waste Contagion,” a variable encompassing various outdated devices and communication infrastructures accelerated by AI advancements.
BAN forecasts 70,000 metric tons of e-waste per gigawatt of data centre capacity, referencing a McKinsey forecast of total data centre capacity expanding to about 219GW by 2030.
Volume and Scope of AI-related E-Waste
Between 2025 and 2050, AI is expected to account for 395 to 617 million metric tons of e-waste, translating to approximately 8.6 to 13.1 million metric tons annually. Previous conservative predictions underestimated impacts by ignoring 87% of data centre infrastructure, according to BAN.
Other studies have presented varying estimates, with a 2024 analysis predicting AI e-waste could rise to between 1.2 million and 5 million tons by 2030. An additional study suggested AI servers could contribute between 131,000 and 225,000 tons of e-waste per year by 2030, comparable to the e-waste output of Denmark.
Reporting based on coverage first published by theverge.com.