The rapid expansion of artificial intelligence and the construction of thousands of massive datacenters could create an unprecedented surge in electronic waste over the coming decades. According to a report from the Basel Action Network (BAN), AI-related infrastructure could help put the world on course to triple its annual generation of toxic e-waste within the next 25 years. The discarded electronics associated with AI over this period could fill enough shipping containers to circle the Earth six times.
The problem is particularly concerning because electronic waste is already one of the world’s fastest-growing waste streams, while only about one-fifth is properly managed. Computers, servers and other electronic equipment can contain hazardous materials such as lead, mercury, cadmium and persistent chemicals. At the same time, they contain valuable metals that could potentially be recovered but are often difficult or expensive to recycle.
AI datacenters could dramatically increase this waste stream because they require enormous amounts of hardware and supporting infrastructure. The BAN report says 4,871 new datacenters are in the construction pipeline in the United States, although not all are specifically intended for AI. Worldwide, analysts project that roughly 100 gigawatts of additional computing capacity could be built by 2030, with further growth expected afterward.
Building this infrastructure requires far more than computer chips. Datacenters contain large racks of servers, networking switches, copper cables, cooling equipment, power-distribution systems and energy-storage technology. Server racks can weigh around 1,360 kilograms each, while individual switches can weigh up to 30 kilograms. Hundreds of kilograms of copper cable may also be required to connect each rack.
One of the biggest challenges is the short lifespan of this equipment. The report projects that much of the hardware will not be designed for easy repair or reuse and could be completely replaced every two to five years. Rapid improvements in AI could also encourage faster replacement of personal computers, smartphones and telecommunications equipment as increasingly sophisticated AI systems demand greater computing capabilities.
Taking this broader range of equipment into consideration, BAN estimates that AI-driven hardware retirement could generate as much as 13 million metric tonnes of electronic waste annually by 2030. Poor management of this material could create serious environmental consequences. Electronics sent to landfills can release hazardous substances into soil and water, while discarded equipment exported overseas may end up in countries without adequate recycling infrastructure.
The environmental footprint of AI datacenters extends beyond waste. Their enormous electricity requirements have also raised concerns about air pollution, energy consumption and public health. Environmental and health experts have warned that pollution associated with generating electricity for datacenters could contribute to significant health impacts in the United States.
Meanwhile, investment in AI infrastructure continues to accelerate. A 2025 McKinsey & Company estimate suggested nearly $7 trillion in global capital spending could be required over five years to support the planned expansion of artificial intelligence.
The BAN report highlights the need to consider the entire lifecycle of AI infrastructure. Designing computers and servers for longer lifespans, repair and reuse, while improving recycling and recovering valuable materials, could help prevent electronic waste from becoming one of the major environmental consequences of the global AI boom.
https://www.theguardian.com/world/2026/sep/16/datacenters-pollution-electronics