For decades, the problem of space debris has proven to be a serious problem with consequences ranging from affecting satellites and communications on Earth, to causing substantial damage to astronauts. However, although in numbers we speak of millions of pieces, in weight they are “barely” 9,500 tons. Very far from the 5 million tons that AI could generate.
A team of scientists from the Chinese Academy of Sciences, in collaboration with Israel’s Reichman University, has estimated the amount of electronic waste that will be generated in the coming years due to the implementation of generative AI applications.
In their study, published in Nature Computational Science, the group attempted to add all circuit boards, batteries and other electronic hardware parts used to power generative AI applications that will no longer be useful in the next decade.
As generative AI applications like ChatGPT have taken the world by storm, An overlooked aspect of their growth is the hardware used to run them. These applications typically run on specialized GPUs connected to high-powered computers. They are usually hosted together in data centers and server farms and there are hundreds of them.
Generative AI applications are resource and energy intensive, and because they have become critical for some users. At the same time, Huge battery reserves are needed to guarantee operation in case of power outages.
Unfortunately, all of this equipment has a useful life. As they age or become obsolete, they are replaced. Old hardware then becomes electronic waste. With this in mind, the authors of the study focused on estimating the total amount of this type of electronic waste that will be generated until the end of this decade.
To make their estimates, they calculated the amount of hardware typically used to run a given application in a standard data center/server farm and the average useful life of each of its components. Next, they identified the number of those data centers. They made educated guesses about the expected demand for such applications and their services in the coming years. Finally, they integrated all their data into a computer model programmed to make these types of estimates.
The model showed that if things continue as they are, the AI industry could produce between 1.2 and 5.0 million tons of e-waste by the end of the decade. It also showed that the annual production of electronic waste will increase from 2,600 tons in 2023 and potentially will reach up to 2.5 million tons per year by the end of the decade.
Taking into account that the space race has been going on for more than 60 years and that in that period it has generated 9,500 tons of space debris, enough to worry us, but that It represents a trifle next to what AI will generate, the issue is worrying.
The authors, led by Peng Wang, point out that Such huge amounts of waste production could be avoided if the industry adopted a circular economy approach in which the hardware would be recycled.