The hardware that powers chatbots could increase e-waste a thousandfold by the end of the decade, researchers warn.

See also: Boost your security with the 42 MDR Unit

Waste generated by physical materials supporting the development and operations of generative artificial intelligence could amount to more than 10 billion iPhones per year, a study by researchers from the University of Cambridge and the Chinese Academy of Sciences.

The study does not accurately forecast e-waste from AI servers and associated equipment, such as GPUs, CPUs, storage, internet communication modules, and power systems. Instead, it serves as a guideline for the industry to limit the negative impact of the rapid expansion of technology. The study did not take into account auxiliary machinery such as cooling and communication units.

Researchers studied computing requirements and product lifelines in low, medium, and high AI growth scenarios. Computing devices typically have a lifespan of two to five years, after which they are replaced with updated versions.

Based on current growth rates, e-waste could increase by 3 to 12 percent by the end of this decade. “Our results indicate a potential for rapid growth of e-waste, from 2.6 thousand tonnes per year in 2023 to approximately 0.4 to 2.5 million tonnes in 2030,” the researchers said. The study takes 2023 as a starting point, calculating e-waste before and after the public launch of ChatGPT.

Geopolitical restrictions on semiconductor imports worsen the e-waste problem since the same products are manufactured in multiple countries when they could be limited to specific geographic areas for simpler management.

The e-waste estimated by Gen AI is only a small fraction of the 60 million metric tons. product on a global scale.

Researchers advise recycling end-of-life servers and reusing components such as communications and power. Using faster high-end GPUs, capable of doing the work of two low-end GPUs, could also reduce e-waste due to the longer lifespan of more expensive chips and lower material profile .

One of the main reasons businesses dispose of e-waste rather than recycle computers is cost. E-waste can contain metals including copper, gold, silver, aluminum and rare earths, but proper handling is costly. Data security is also a concern: there’s nothing more breach-proof than destroying the equipment.

However, researchers estimate that taking mitigation measures could reduce e-waste by 16 to 86 percent. The reduction range reflects projected adoption, they said. If each GPU were reused in a low-cost inference server after it could no longer be used for AI, this act of reuse would amount to a significant reduction. But if only one in ten people are reused, the impact diminishes. Reducing electronic waste AI is a choice and not inevitable.