Technology giants such as Google, Microsoft or Meta are immersed in a frenetic race to develop more and more tools based on generative artificial intelligence (AI). That competition is leaving its mark on the planet. The training and maintenance of these sophisticated models requires gigantic computing power running 24 hours a day in data centers. That has triggered the energy consumption of these infrastructuresas well as their associated carbon emissions and the water consumption, used to cool the systems. But the environmental footprint of generative AI It doesn’t end here. The equipment used in the data centers in which this technology is cooked must be continually renewed, and this produces a large amount of digital waste, including highly toxic metals, such as lead or chromium. A group of scientists has made calculations and their conclusion is alarming: if measures are not taken to reduce electronic waste associated with generative AI, it will multiply by 1,000 by 2030, reaching between 1.2 and 5 million tons.
The investigation, whose results are published today in the magazine Nature Computational Sciencewants to contribute to finding a way to reduce electronic waste associated with generative AI, in particular large language models (LLM) such as ChatGPT, Copilot, Gemini, Claude or Llama. “I am quite optimistic about the possibility of promoting circular economy strategies among the main actors involved in generative AI,” Peng Wang, a researcher at the Institute of Urban Environment of the Chinese Academy of Sciences, in Xiamen, tells this newspaper. and one of the authors of the study. “However, I have deep concerns about the competition between the rate of expansion of generative AI and the adoption of the circular economy. Given the unprecedented increase in demand for this technology, to win this battle, shock measures should be implemented imminently,” he adds.
Wang and his colleagues took the eight-GPU Nvidia DGX H100 server from 2023 as a reference for their calculations. It is, today, the hardware reference in the AI processing, which needs much more powerful equipment than those used in the rest of the programs that run online. The authors of the study have constructed four future scenarios to estimate the evolution of generative AI, and its associated demand for data processing, between 2020 and 2030. The first considers a limited expansion of the chip industry and related manufacturing. Generative AI. It places the growth at 41%, the same as that experienced between 2022 and 2023, understanding that it is impossible for the pace to be lower because since then improved versions of several of the most used models have been announced or presented.
The next three scenarios are the conservative one (+85%), which is based on the growth figures harvested by Alexa-type voice assistants; the moderate (+115%), inspired by TikTok numbers; and the aggressive one (+136%). For the latter, it has been considered that large language models become “a ubiquitous tool in people’s daily lives,” taking as a reference the growth rate of Facebook, “a platform used literally by everyone.”
According to their calculations, waste generation would go from the 2,600 tons recorded in 2023 to between 0.4 and 2.5 million tons by 2030, that is, up to 1,000 times greater. That volume of waste would be equivalent to discarding between 2,100 and 13,300 million iPhone 15 Pros. Or what is the same, between 0.2 and 1.6 phones per human being that year. The estimate assumes that no measures to reduce digital waste will be implemented during the current decade. In comparative terms, all 2022 waste related to information technology equipment, such as laptops or tablets, was 4.6 million tonnes. It is estimated that by 2030 there could be 43.2 million tons.
The projections of Peng and his colleagues have not taken into account something that, as they themselves highlight, could aggravate the situation: restrictions on the import of semiconductors, considered key products from a geostrategic point of view. That blockage could prevent many countries from benefiting from continued improvements in chip efficiency. And that has its weight: the researchers’ analysis concludes that a year of delay in obtaining the latest generation semiconductors can lead to a 14% increase in the generation of generative AI servers that reach the end of their useful life. In absolute terms, that year of delay would mean an additional 5.7 million tons of waste. Given that data centers dedicated to AI are quite geographically concentrated, digital waste will be concentrated in Europe (14%), East Asia (25%) and North America (58%), although most of it ends up being sent to Africa and Asia.
Are the projections drawn by Wang and his colleagues reasonable? “I think so, but I get the impression that the accumulation of decisions that simplify the model (necessary, because otherwise the analysis would not be viable) adds a lot of uncertainty to the results and conclusions. Therefore, it is important to take these results with great caution,” says Álex Hernández, principal researcher at the Quebec Institute of Artificial Intelligence (MILA), an institution that has among its ranks: Yoshua Bengio, considered one of the godfathers of neural networks. Hernández also thinks that the fact that the four scenarios presented do not include an analysis of their probability, or at least their feasibility, makes the study less criticizable, but also less relevant.
“Although predicting the future development of the hardware is difficult, I consider that the document’s forecast is a reasonable indicator of the electronic waste that generative AI will likely generate,” says Shaolei Ren, associate professor of Electrical and Computer Engineering at the University of California in Riverside (United States). , in statements to the SMC Spain service.
To do
Given this situation, the authors examine several circular economy strategies to try to limit the generation of waste. The most effective, logically, is to increase the useful life of the hardware. The numbers from the team of scientists reveal that 62% of the AI servers that are thrown away each year (3.1 million tons) could be maintained if their useful life were increased by one year, which, according to the authors, is usually of three. “According to my own research, GPUs have a life cycle of between three and five years. That is, data centers dedicated to AI renew all their chips every four years or sooner,” says Ana Valdivia, professor of Artificial Intelligence, Government and Policy at the Internet Institute of the University of Oxford. The engineer, who has not participated in this study, investigates electronic waste.
Reusing some parts of GPU processors (those used in training AI models), such as communication, memory or battery modules, could reduce electronic waste by 42% (2.1 million tons) . Valdivia does not see it clear that this is feasible. “GPUs cannot be inserted into a circular economy because it is very expensive to recycle their components, something that I do not see discussed in the article [de Wang y sus colegas]. 100% of a GPU ends up incinerated or in a landfill,” says the expert.
The equipment used in data centers consists of three main elements, according to the study: circuits with semiconductors, batteries and structural parts. When discarded, toxic materials remain, from lead and chromium to acrylonitrile or polycarbonates, but also precious metals, such as gold, silver, platinum, nickel or palladium. If properly recycled, these materials could be worth between $14 billion and $28 billion, Wang and his team estimate.
However, the study could have been more subtle. “It strikes me that server cooling devices are excluded from the analysis, when they play a fundamental role in this type of device and also represent a large amount of the physical material of the servers,” says Hernández.
