AI Doesn’t Just Consume a Lot of Energy. It Also Generates a Massive Amount of E-Waste

  • Many companies are disposing of functioning hardware.

  • Spending on artificial intelligence has increased eightfold from 2022 to 2023.

Some of the world’s largest companies are investing billions of dollars in artificial intelligence. This field is poised to advance and integrate AI technology into various sectors, including transportation, education, and medicine. As such, the business opportunities in AI are immense.

However, to maintain a competitive edge, there’s a cost that goes beyond mere economics. The data centers that train and operate AI language models consume significant amounts of energy and, depending on their cooling systems, also require a considerable amount of water. Additionally, they generate a large volume of electronic waste.

AI and Electronic Waste

A recent study conducted by leading researchers from Chinese and UK universities suggests that the total e-waste associated with AI could range from 1.2 million to 5 million tons by 2030. The paper highlights the current trend and urges various stakeholders in the industry to manage waste proactively.

Data from the United Nations Global E-Waste Monitor indicates that the amount of e-waste is increasing five times faster than the documented recycling efforts. In 2019, approximately 53.6 million tons of e-waste were generated, while this figure rose to 62 million tons by 2022.

Experts examine the situation from several perspectives. From an investment standpoint, spending on AI increased eightfold between 2022 and 2023, surpassing $25 billion. Companies have allocated a significant portion of this investment specifically to build and equip AI data centers.

A key aspect of AI data centers is that component replacement is driven not only by the lifespan of the components but also by their performance. Many companies are eager to incorporate the latest GPUs from Nvidia and other manufacturers to enhance the computing capabilities of their data centers, even when older components are still operational.

Early component replacement isn’t a new concept in the enterprise sector. Historically, the industry has preferred to destroy functional hard disks rather than recycle them, motivated by concerns that sensitive data could be compromised. However, initiatives are underway to mitigate this practice.

Researchers propose reusing replaced components. While they acknowledge that these components may not be suitable for high-demand tasks, they envision them being utilized in data centers to support less intensive operations, such as website hosting and data backups. Additionally, they suggest donating these components to educational institutions.

Image | Hans Ripa | Meta AI

Related | France Has a Striking Plan to Recycle Some of Its Nuclear Waste: Turn It Into Forks, Doorknobs, and Saucepans

See all comments on https://www.xatakaon.com

SEE 0 Comment

Cover of Xataka On