How Will AI Impact E-Waste?

AI is revolutionising industries across the planet, but it is also causing a number of significant issues, mostly in the form of e-waste.  It’s development has highlighted the importance of responsible use and disposal, and by proactively monitoring and optimising resources while utilising scalable solutions, you can do your part in ensuring a more sustainable future.

In 2022, humans generated 62 billion kg of e-waste across the planet, with expectations for it to reach nearly 80 billion kg by 2030. This makes for grim reading already, but the uptake of AI could exacerbate the issue even further.

The Rise of AI

AI is spreading like wildfire. It is borderline impossible to not encounter it with AI suggestions on search engines and websites adopting AI chatbots. The beginning of this widespread adoption came with OpenAI releasing ChatGPT in November 2022 and reaching 100 million monthly users within two months.

The growth of AI does not look to be slowing.  The global AI market size was valued over $224 billion in 2024 and is predicted to grow by 32.5% from 2025 to 2030, taking it to $1.236 trillion.

AI growth is synonymous with an increase in energy usage. For example, it is estimated that ChatGPT uses 500ml of water per 300 prompts. When you consider that ChatGPT receives 1 billion messages per day, those numbers add up quickly. It is difficult to find an exact number on energy usage due to the different ways that machine learning models can be designed and configured. What we do know is energy usage is high, and the most energy intensive aspect comes with model training and deep learning.

Constant Upgrades Lead to E-waste

We are witnessing the rapid development of AI, and nobody wants to be left behind. Phones laptops, TVs, and even ovens are now available with AI grafted on. As much as anything else, AI is a powerful selling tool, and consumers want the latest and greatest. 

The unfortunate reality is that an individual willing to upgrade this year for the promise of AI might do the same again next year as AI continues to progress. What does the consumer do with their old device? They may trade it in, it may sit in a cupboard or garage indefinitely or it may end up in landfill, despite being in perfect working order.

The Main Culprit

The biggest issue lies within data centres. Data centres have taken the brunt of AI growth through near constant expansions and upgrades.

According to Forbes, North America data centres could multiply sixfold by 2027, and hyperscale data centres will continue to double in capacity every four years. Rapid advancements are causing hardware to become outdated faster, increasing the likelihood of it becoming e-waste.

Government Technology estimate that by 2030, AI will be responsible for 2.6 million metric tons of e-waste per year. Rapid obsolescence is completely unsustainable yet shows no signs of slowing as mass adoption of AI continues.

The main reason data centres are changing their design, infrastructure and equipment can be tracked to GPUs. A GPU’s ability to handle multiple tasks at once was quickly identified as the perfect tool for training AI, and they have become a vital part of AI infrastructure ever since.

Before the AI boom, data centres generally used CPUs, meaning less power and cooling. AI has left data centres scrambling to accommodate the surge in energy usage. GPU integration means everything about data centres has to be reevaluated, and in some cases, redesigned.

But upgrading to GPUs is not enough. AI is still progressing, which forces hardware to keep up. The result is, data centres are finding themselves in an endless cycle of upgrades, removing operational equipment.

Just 22% of e-waste is properly collected and recycled, which does not bode well for discarded equipment. Processors are also especially difficult to recycle as they have trace amounts of so many different metals. 

Is AI All It’s Cracked Up to Be?

Whatever your stance on AI, it’s here to stay. We just don’t know to what extent. It is widely accepted that AI is going to revolutionise many aspects of our lives. Its immense potential was quickly identified, but it is not beneficial to every single aspect of our lives. 

AI has also started to run into limitations. The OpenAI cofounder Ilya Sutskever recently announced that they had hit peak data, meaning, amazingly, that they have used up all the useful data on the internet to train their AI. This does present a significant roadblock in AI progression, as businesses will now need to rely on new techniques such as synthetic data.

AI and the S-Curve

The S-curve is a representation of the progress over time. It suggests that progress starts slowly, before experiencing rapid progression and growth, and eventually reaching a plateau. The signs of an approaching plateau in AI after such rapid growth suggests that we are at this final point of the S-curve to some, while others argue that we are only at the beginning.

It is easy to get swept up in the hype and excitement around a new technology. The excitement comes from a truly revolutionary technological milestone, but can AI continue to make advancements at the rapid rate we have seen for the past few years? 

 It may feel u  nprecedented, but we have seen this occur before in technology. Mobile apps were subject to rapid innovation upon the introduction of app stores. This brought a perceived requirement for everyone to have their own app. Today, however, innovation in this space is rare, and having your own app is not the necessity it once seemed to be. Will we see a similar trend form with AI?