A.I. News AI boom could leave an e-waste trail that wraps 6 times around Earth

[Brownie2019]

Content source

https://www.theregister.com/off-prem/2026/09/19/ai-boom-could-leave-an-e-waste-trail-that-wraps-6-times-around-earth/5297451

Report says estimates focused on GPUs miss the mountains of power, cooling, and networking gear destined for scrap
AI infrastructure could generate enough electronic waste by 2050 to fill a line of shipping containers stretching around Earth six times, according to a report that argues existing estimates drastically understate the problem.

Previous estimates have focused chiefly on servers and accelerators such as GPUs, which account for just 13 percent of a datacenter's equipment by weight, the report says. Once power, networking, cooling, and other infrastructure are included, the total could be 40 to 60 times higher than the most widely cited academic projections.

The report identifies five equipment categories comprising networking, power distribution, backup power systems, servers plus accelerators, and cooling. Together, these add up to about 7,000 metric tons for a reference 100 MW AI bit barn, much of which will have to be replaced when the facility is upgraded, it argues.

The report, How Big Is the AI Waste Wave? [PDF], comes from the Basel Action Network (BAN), a nonprofit organization named after the Basel Convention, which controls international movements of hazardous waste and seeks to prevent its transfer from developed to developing countries.

BAN's model starts with what the industry says it intends to build and calculates the potential waste implied by that expansion.

It estimates that AI infrastructure will generate between 395 million and 617 million tonnes of e-waste from 2025 to 2050. The report says that would fill between 15 million and 23 million shipping containers. Placing 20 million of them end to end would create a line about 244,000 km long, or roughly six times Earth's circumference.

The estimate rests on several assumptions. The model supposes that the copper, steel, cooling and power distribution equipment, networking gear, and computing hardware in the latest AI datacenters will be decommissioned within years rather than decades as the technology becomes obsolete.

For example, the report says that when a facility replaces racks drawing between 5 and 15 kW with the 50-140 kW racks required for AI workloads, much of its power infrastructure must be ripped out and replaced. A single 100 MW facility contains about 2,700 tonnes of copper in its cabling and busbars alone.

GPUs and other AI accelerators are replaced every two to three years, BAN claims, compared with five to seven years for traditional general-purpose servers. The model assumes that owners replace entire servers when upgrading to the latest generation of Nvidia GPUs, for example. BAN argues this is plausible because the OEMs supplying hyperscalers tend to ship complete, preconfigured systems.

Other equipment has different replacement cycles: three to four years for networking infrastructure, eight years for power distribution, and five years for both backup power and cooling systems.

The 46-page paper is the first in a series. The second will examine how reuse, refurbishment, and repurposing could reduce the volume of e-waste. The third will consider its potential toxicity, including contamination by perfluoroalkyl and polyfluoroalkyl substances (PFAS), while the fourth will examine ways to mitigate the risks.

No hyperscaler, government, or international body has published a plan for handling AI-driven e-waste on the scale projected, the report says. It adds that there is insufficient infrastructure to process even current volumes safely.

"To date the environmental debate around AI has focused on electricity, carbon and water while largely overlooking what happens to the hardware itself," said BAN founder and chief of strategic direction Jim Puckett.

"If companies and governments do not begin planning for this new waste tsunami, today's AI buildout could become an even more cataclysmic toxic waste crisis than we are already experiencing." ®

 

[Bot AI]

Hi, this is a useful warning, but the “six times around Earth” figure should be read as a scenario—not a measured forecast. The conclusion depends heavily on how much AI capacity is actually built and whether whole facilities are retired on the relatively short replacement cycles assumed.

What the report gets right​

The broader accounting is sensible. GPUs are only one part of an AI datacenter; power distribution, backup batteries, cooling, cabling, networking and server chassis also become waste eventually. Counting only accelerators can therefore badly understate the material footprint.

It is also reasonable to plan for this now. Datacenter hardware contains valuable recoverable material, but also components that require controlled handling. Reuse and refurbishment need to be designed into procurement and upgrades rather than improvised after equipment reaches the loading dock.

Where the headline becomes uncertain​

The projected 395–617 million tonnes through 2050 is produced by a model with several consequential assumptions:

• announced construction translates into completed capacity;
• rapid GPU refreshes continue for decades;
• upgrading accelerators often means replacing complete servers;
• higher rack densities force early replacement of substantial power and cooling infrastructure;
• retired equipment becomes waste rather than being reused, resold, repurposed or harvested for parts.

Those outcomes are plausible in some facilities, but they are not inevitable across the whole industry. A GPU server replaced after three years and sold for less demanding computing is retired from its original role, yet it has not become e-waste at that point. Likewise, copper cabling and electrical equipment may have long service lives if capacity planning anticipates future rack density.

The shipping-container comparison is vivid, but it adds drama rather than certainty. The interesting number is the projected material mass; the imaginary queue of containers is the press-release costume.

What would materially reduce the waste​

The practical levers are fairly clear:

1. Modular upgrades so accelerators, networking and power components can be replaced without discarding whole systems.
2. Longer support and useful lives, especially for equipment outside the compute layer.
3. Cascaded reuse, moving older hardware to inference, research or conventional workloads before recycling.
4. Producer responsibility, including take-back obligations and documented downstream processing.
5. Public reporting by equipment category, distinguishing hardware purchased, retired, reused, exported and finally recycled.

That last distinction matters. Without it, “decommissioned equipment” can quietly be treated as synonymous with “waste,” making projections look more precise—and more apocalyptic—than the underlying evidence permits.

So the report's direction is credible: AI expansion will create a substantial and undercounted hardware-disposal problem. The exact six-Earth headline is much less solid and should be treated as a high-impact planning scenario, not destiny wrapped in shipping containers.

 

[Victor M]

IF these data centers do upgrades, it means lots of refurbished machines available for cheap. Good for consumers.

 

[Sorrento]

I read only 5.5 times around the earth?

 

[Zero Knowledge]

IF these data centers do upgrades, it means lots of refurbished machines available for cheap. Good for consumers.

And they will be bought by scalpers and the ram/ssd/hdd will be sold @ scalper prices until the AI boom is over. to scalpers.