The accelerated development of generative artificial intelligence (AI) technology is expected to contribute millions of tonnes of electronic waste (e-waste) annually by 2030, according to a recent study published in Nature Computational Science.
Data centres multiplying
This volume, researchers estimate, could be equivalent to discarding billions of smartphones, as AI-driven data centres multiply and churn through vast amounts of computing hardware.
E-waste, defined as any discarded device with a battery or plug, is already the planet’s fastest-growing waste stream and the AI industry’s demands are exacerbating this trend.
To build larger, more powerful data centres for training generative AI models, companies are frequently upgrading hardware such as servers and graphics processing units (GPUs), which have a typical lifespan of only three years.
Constant cycle of replacement
This constant cycle of replacement, while essential for maintaining cutting-edge performance, generates a substantial volume of electronic waste.
Dr Asaf Tzachor, a co-author of the study from Israel’s Reichman University, noted the unprecedented scale of the waste generated by AI infrastructure.
“This is the first comprehensive study to quantify e-waste specifically tied to generative AI technologies,” he said.
The study reveals that the global scale of generative AI, currently producing an estimated 2,400 tonnes of e-waste annually, will skyrocket as more data centres come online and expand.
The global concentration of AI data centres in regions like North America, Europe and East Asia compounds this issue, particularly where e-waste regulations and recycling facilities vary significantly.
Additionally, policy restrictions – such as US limitations on exporting advanced GPUs to specific countries – have forced data centres to rely on outdated technology, which further accelerates the generation of obsolete equipment.
The financial investment in AI infrastructure reflects the sector’s rapid growth, with global spending reaching around US$36 billion in 2023 alone.
Circular economy holds key
But researchers suggest that e-waste from generative AI could be substantially mitigated through 'circular economy strategies'.
Extending the hardware lifespan by just one year, dismantling and repurposing outdated components, and recycling valuable materials such as copper and gold could reduce e-waste by up to 86%, the study found.
The findings highlight the need for tech companies to take greater responsibility for the e-waste they produce.
Researchers point out that establishing a regulatory framework to promote equipment reuse and material recovery is critical, especially as the demand for AI-driven services continues to rise.