Do you know 10 billion iPhones a year thrown into a large landfill? This is not a virtual image, yet unfortunately artificial intelligence has something to do with it. Here I am. In practice, the rapid expansion of AI is driving a global technological transformation, but its growing computing needs have been underestimated in the current enthusiasm. In fact, due to obsolete infrastructures that need generational replacement, between now and 2030 a quantity of electronic waste could be generated equal to the number of iPhones quantified at the beginning of this article. A quantity that is almost impossible to visualize.
The study published on Nature and conducted by researchers from the University of Cambridge and the Chinese Academy of Sciences is now pushing the sector to seek innovative solutions to limit environmental impact. It is practically expected that the AI industry, starting from the 2.6 thousand tons of electronic waste generated in 2023, could produce between 0.4 and 2.5 million tons of e-waste per year by 2030, with an increase of almost a thousand times. And such a mass of waste also leads to the complexities of recycling advanced electronic components, rare and expensive materials found in chips and batteries and which must be extracted, reused or properly disposed of to avoid further environmental damage.
Apart from this, another crucial aspect in the growth of the AI sector is energy consumption. Whenever an AI model is trained or applied to large data, it requires enormous amounts of energy. Increasing energy demand from infrastructure has led to pressure on current energy production systems, with existing technologies often inadequate to support an exponentially growing industry. To meet this challenge, some of the world’s largest technology companies, such as Google and Amazon, are exploring radical and sustainable solutions. Recently, both companies announced plans to purchase mini nuclear reactors, compact devices that produce clean, continuous nuclear power. These facilities, which are expected to come into operation in the coming years, represent perhaps the only opportunity for big tech to power their infrastructures autonomously and sustainably, reducing dependence on traditional electricity grids and helping to reduce associated CO2 emissions to energy of fossil origin.
Returning to the issue of waste, the study published on Nature he further explains that server and data center components, designed to operate at full capacity for years, will inevitably end up becoming obsolete and creating digital waste. And that it is crucial to start predicting and addressing the risks associated with rapid expansion now. To evaluate the environmental impact of e-waste, researchers have in fact modeled different scenarios, including low, medium and high growth rates of the AI sector: each of these estimates the computing resources necessary to support the expected growth and quantifies the duration of server before becoming obsolete. The results show that – in all scenarios – the e-waste problem is destined to grow, requiring a restructuring of the production and recycling cycle to avoid a sustainability crisis.
The answer, at this point, is in short the circular economy and therefore the reconditioning of servers and the design of modular equipment, which allow hardware updates without the need for complete replacements. But we certainly need to act quickly, before those 10 billion iPhones thrown away every year are just the beginning of an unsustainable burden for the world of the future.