Quantity of electronic waste could increase

Xiamen/Kaiserslautern (dpa) - According to a study, if artificial intelligence (AI) spreads rapidly, up to a thousand times more associated electronic waste could be generated in 2030 than in 2023. However, the amount of waste could be significantly reduced through various measures, writes a group led by Peng Wang from the Chinese Academy of Sciences in Xiamen in the journal "Nature Computational Science". The figures are based on model calculations that assume a switch to newer computer systems every three years.

Large language models are used for AI applications such as ChatGPT. "Large language models require significant computational resources for training, which requires extensive computer hardware and infrastructure," the authors write.

Not just more energy and more_CO2

Studies on sustainability have so far mainly focused on the energy consumption and carbon dioxide emissions of AI models. Wang and his colleagues, on the other hand, wanted to know what quantities of electronic waste would be generated if computationally intensive AI applications were used in more and more areas.

The scientists use a scenario in which they assume the adoption of large language models for everyday use, as can already be seen with some search engines and social platforms. With such a broad AI application, the data centers for training and providing AI models would have to grow very quickly.

As a result, the amount of e-waste from discarded servers and other devices could increase from around 2550 tons in 2023 to up to 2.5 million tons in 2030. In scenarios with less use of AI, the amount of scrap could remain limited to 400,000 to 1.5 million tons in that year.

Great potential for reduction

The researchers also calculated how much different measures could reduce the amount of scrap. The most effective measure would be not to scrap servers and other devices after three years, but to use them for one year longer for simpler AI tasks or for completely different purposes. This would reduce the amount of waste by 62 percent compared to the baseline scenario.

If individual modules of the systems, such as processors and memory, were refurbished and reused, this could save 42%. In addition, improved algorithms would offer a savings potential of 50 percent and more efficient chips 16 percent.

Smartphones and PCs are still to come

Wang's team also refers to the latest "Global E-Waste Monitor". According to this, the amount of waste generated by smaller electronic devices - such as smartphones or personal computers - is expected to total a good 43 million tons by 2030.

The waste generated by AI servers and devices over the years, as calculated by the authors, could amount to a cumulative 5 million tons by 2030 in the base scenario, i.e. just under 12 percent of this amount. Based on the study's most conservative scenario, the cumulative e-waste generated by AI would account for around 3 percent of the e-waste from smaller electronic devices.

More circular economy needed

Christiane Plociennik from the German Research Center for Artificial Intelligence (DFKI) in Kaiserlautern pointed out that there is only a small data basis for the assumptions made by the authors in the baseline scenario. However, even the most conservative scenario with significantly lower waste volumes and the forecast of the "Global E-Waste Monitor" provided important reasons for establishing a circular economy in information technology.

"We need to raise awareness in society that behind a cloud or an AI application are data centers with high resource consumption," emphasized Plociennik. Reusing IT devices is preferable to recycling them.