AI Data Centres Face 617 Million-Ton E-Waste Warning

The Verge covered a Basel Action Network report warning that AI infrastructure could leave a much larger electronic-waste trail than earlier studies captured.

By 2050, equipment tied to the AI buildout could amount to enough discarded material to fill 23 million 40-foot shipping containers, a line of containers long enough to circle the world six times.

The estimate is higher because BAN counted more than servers and graphics processors.

Its paper includes the electrical, cooling, backup-power and networking systems that support data centres, plus a broader category it calls “AI Waste Contagion” for telecommunications equipment and personal devices that could be replaced sooner as AI advances.

That wider boundary changes the scale of the problem.

BAN expects global e-waste to triple to as much as 211 million metric tons a year by mid-century and assigns about 15% to 20% of that total to AI.

For data centres, the group puts the waste intensity at 70,000 metric tons for each gigawatt of capacity.

It also cites a McKinsey projection that total data centre capacity could reach up to 219GW by 2030.

The headline range is not limited to the end state in 2050.

For AI-related electronic equipment retired between 2025 and 2050, BAN calculates a cumulative 395 million to 617 million metric tons of e-waste.

Spread across those years, that would mean 8.6 million to 13.1 million metric tons of AI-linked equipment retired annually.

The comparison with existing waste systems is part of the report’s warning.

Worldwide e-waste already totals 68.3 million tons a year, and less than a quarter of it is formally collected and recycled.

Much of the rest moves through informal waste channels where burning or burying equipment can expose workers and nearby communities to toxic materials such as lead and chromium.

The US position adds another pressure point because the country has more data centres than any other market.

The United States has not ratified the Basel Convention, the international agreement designed to restrict trade in hazardous waste, and investigations have found US recyclers still ship e-waste abroad.

The World Health Organization has linked informal recycling around dumped electronics to health risks for millions of children forced to work or live nearby.

Earlier AI e-waste estimates were narrower.

BAN argues that studies focused on servers and accelerators miss about 87% of a data centre’s electromechanical infrastructure.

A 2024 study projected AI e-waste of 1.2 million to 5 million tons by 2030, while a February study put AI-server waste at 131,000 to 225,000 tons a year by decade’s end, roughly comparable to all e-waste from a country the size of Denmark.

Jim Puckett, BAN’s founder and chief of strategic direction, framed the issue as a planning failure rather than a distant cleanup problem.

“AI may feel weightless, but every model depends on an enormous amount of highly specialized, cutting edge hardware,” he said.

The report’s largest numbers turn that hardware into the central risk: not only faster data-center growth, but a replacement cycle that could move hundreds of millions of tons of equipment into waste streams by 2050.