Electronic devices define modern life. Most of us can’t make a coffee, find a train station or relax in the evening without one. The effect this is having on humans is one thing, but the impact on our planet is just as, if not more, concerning. Our discarded electronics don’t just disappear when we’re done with them. E-waste is one of the world’s fastest-growing waste streams: in 2022, the global volume of e-waste could circle the equator in 40-ton trucks. This is a problem. When discarded, this waste is full of toxic metals like lead and mercury. These materials contaminate soil and water, and, when burned, release poisonous compounds linked to severe health issues by the World Health Organisation. A common strategy for recycling companies is to ship waste to developing countries with cheap labour, dumping mountains of waste that leave the handlers at risk.
Generative AI (GAI) is taking over the world—in more ways than one. One new study has shed light on the hidden impact of LLM e-waste, in the hope that we can stem its tide before it’s too late.
Tomorrow’s AI, Today’s Problem: How Toxic GAI E-Waste Could Engulf the Planet
E-waste is inherently complex, costly and dangerous to recycle. It contains a challenging mix of both valuable and hazardous materials that need accuracy and time to extract. The current recycling infrastructure is struggling to cope. A recent UN report warns that the global generation of e-waste is rising five times faster than documented recycling rates. With less than a quarter (22.3 percent) of 2022’s e-waste collected worldwide, the scale of this problem demands urgent action. While the EU aims to address this through its move towards a circular economy by 2050, the escalating, global nature of the e-waste crisis requires a fundamental overhaul of how the world consumes and deals with electronic goods.
E-waste recycling is a complicated issue
Just outside the historic German town of Goslar in Lower Saxony, one of Europe’s largest facilities, Electrocycling, offers an interesting response to this crisis. Every year, the sprawling industrial complex processes up to 80,000 tonnes of discarded electronics. This relies largely on human effort: more than half the site’s employees are dedicated to manually processing the waste. Their most critical job is removing batteries. If these aren’t discarded properly, they pose a significant fire hazard. This makes them one of the biggest challenges in safely managing the stream of incoming e-waste.
“It’s not a dream job,” says Hannes Fröhlich, Electrocycling’s Managing Director. Dismantling these varied and complicated appliances every day with hammers and pliers is fiddly work; “There are more and more devices, they are getting smaller, and they all contain lithium batteries”. These batteries, which are made from rare materials found in the Earth’s core, are often “permanently installed, soldered or glued in place.” There is no easy, quick or efficient way for humans to extract the usable components from these devices. However, Fröhlich believes “we can do better.”
Automating the assembly line
Automation in the form of machines or robots might seem like an obvious solution for these tasks. But the problem has historically been flexibility. While humans can more easily decipher the different tasks associated with dismantling different electronics, every slight change in product design or process requires a time-consuming and therefore costly hardware and software overhaul for a robot.
To overcome the limitations of industrial robots programmed for single, repetitive tasks, Electrocycling collaborated with the EU-funded ReconCycle research initiative to develop self-reconfiguring robots. Supported by state-of-the-art AI, these robots—developed by researchers from Slovenia, Germany, and Italy—can automatically adapt both their hardware and software for various component extraction tasks.
“We wanted to expand robotics [and] introduce robots where there aren’t any yet,” explained Dr Aleš Ude, who heads the Department of Automatics, Biocybernetics and Robotics at the Jožef Stefan Institute and coordinated the ReconCycle research team.
The project, which operated from 2020 to 2024, was able to develop adaptable robots that successfully remove batteries from common household items like smoke detectors and radiator heat meters. Replaced every five to eight years, these items in particular generate significant waste.
How AI-powered robots work
Ude’s team has created a versatile “smart robotic worker”—a complete station that includes the robot, its tools and a computer brain. The key breakthrough is that this system can teach itself to do different jobs without needing to be manually reprogrammed via AI-driven software. For added circularity, its parts can be quickly snapped together and taken apart. It even uses the SoftHand: specialised, soft-grip hands that can handle objects with the delicate touch of a human.
So, are AI-powered robots like those from ReconCycle the definitive end to our e-waste recycling problem? The answer requires a necessary distinction.
Is AI a friend or foe in the fight against e-waste?
As demonstrated by projects like ReconCycle, AI tools can and do have an absolute positive environmental impact. They have the power to significantly improve the precision and efficiency of complex issues like e-waste disassembly, and Dr Ude hopes that these adaptable robotic systems will have applications far beyond recycling. Their flexible, agile handling style could allow them to perform general housekeeping tasks or even support carers in senior homes. “Robotics could be of great help in such areas,” he suggests. This type of solution-oriented AI could well improve our world in an ever-increasing number of ways.
However, AI is not a monolith, and not all AI is good. There is a huge difference between general AI, such as the type used in ReconCycle, and Generative AI (GenAI). This distinction is key to understanding the exploding environmental footprint.
RESET just launched our deepest dive yet into the environmental cost-benefit of AI.
GenAI and other large language models are designed to support consumption. They require vast amounts of computational power, energy and the hardware becomes obsolete rapidly. The specialised hardware (GPUs, servers, memory) GenAI relies on has a very short lifecycle—often just a year or even less. This means GenAI is creating its own deluge of e-waste.
The energy and resource demands of GenAI are also staggering. According to a recent study by the Öko-Institut on behalf of Greenpeace, global greenhouse emissions from AI data centres will rise from around 212 million tonnes (Mt) in 2023 to 355 Mt in 2030, with AI’s annual water usage predicted to be the same as the whole country of Denmark in 2027.
The real challenge, then, is not to toss out all AI. When developed sustainably, positive applications—like ReconCycle’s robots—can offer benefits to some of our planet’s biggest problems. The goal is to ensure that the effort to save our planet’s resources doesn’t inadvertently consume them in the process.
How can we ensure a green digital future?
Growing e-waste, carbon emissions from AI, data centre water usage—is rampant digitalisation compatible with a healthy planet? Our latest project explores how digital tools and services can be developed with sustainability in mind.
