NEW YORK: Researchers forecasting a 1000-fold increase in e-waste from AI computer servers by 2030 called for recycling strategies to reduce the environmental impact.
A recent study published in Nature Computational Science predicts a staggering increase in electronic waste (e-waste) associated with generative AI, forecasting a potential rise to 1.2 to 5.0 million metric tons by 2030—about 1000 times the amount produced in 2023.
Asaf Tzachor, a sustainable development expert at Reichman University and co-author of the study, noted that the e-waste generated by large language models could soar to 2.5 million tons annually if no waste reduction strategies are adopted.
The study underscores the urgent need for effective recycling and sustainability practices, suggesting that implementing strategies to prolong, reuse, and recycle AI hardware could reduce e-waste by 16% to 86%. The study highlighted the global nature of the e-waste crisis and the importance of cross-border management to mitigate its impact.
E-waste encompasses discarded electronic devices like computers, smartphones, and servers, making up 70% of global toxic waste, yet only 12.5% is recycled. The study stressed that improper disposal can release hazardous materials such as lead and mercury, posing significant risks to ecosystems and human health.
The researchers used a model to estimate the e-waste generated by data centers supporting generative AI. Their calculations indicated that the figures might be conservative due to the rapidly evolving AI landscape, with geopolitical factors and frequent hardware upgrades further exacerbating e-waste production.
