How your AI usage is impacting the environment
In 1921, Czech playwright Karel Capek released a science fiction play titled Rossum’s Universal Robots, familiarising audiences with ‘artificial people’, which were nothing but robots. Following this, computer scientist Edmund Callis Berkley’s book, Giant Brains, or Machines That Think (1949), made compelling statements comparing human brains to different models of computers.
A century later, the same computer brains have taken over and we call them Artificial Intelligence (AI). What may now seem like a relatively new technological advancement, was sowed much earlier, with the term being coined in 1956 by John McCarthy, a prominent computer scientist and cognitive scientist, also known as the father of AI.
Research papers and debates have had scientists proving that human lives can be revolutionised in the future. And it did. From the first driverless car manufactured by Ernst Dickmann and his team at Bundeswehr University of Munich in 1986, Deep Blue (a super computer) beating the world chess champion, Garry Kasparov (1997), to NASA landing rovers on the moon and virtual assistants like Siri and Alexa coming into the market, the changes have been endless.
But now, it is time to consider an oft-overlooked fact. While AI has been helping us daily, it is also causing serious damage to the environment.
“Most people associate AI with futuristic software, but the reality is that it runs on physical infrastructure,” informs Rijin Reji, freelancing AI-based projects. Harsh Kothari, CEO of Webstrake, an AI-driven SaaS company delivering innovative digital solutions for businesses, adds, “Every time, a large AI model is trained or even runs frequent queries on it, we are tapping into massive server farms that consume a lot of electricity, water for cooling the machines to keep them from overheating. It adds up quickly when you are operating at a global scale.”
Resources in use
To run an AI model, electricity, water, and minerals are used. Massive data centres are powered by electricity, then cooled by water, and are built using rare minerals like silicon for chips, copper and gold for electronic circuits, lithium, cobalt, nickel for batteries, and rare earth materials like neodymium, dysprosium for magnets. These minerals are sourced from mines at environmental and human costs. “All of these minerals go into hardware, which eventually runs the AI models,” says Rahul Roy, AI engineer at Qualcomm R&D.
Over time, as models evolve quickly, older chips and servers get replaced. “Even though AI feels digital and clean on the surface, it has a very real and growing environmental footprint underneath,” Harsh points out.
Hence, there is the accumulation of e-waste once the hardware is outdated. Harsh explains, “The major cloud providers recycle, maybe, 70% of materials at certified facilities, recovering copper, gold, and some cobalt. The remaining 30% ends up in e-waste streams that get exported to developing regions, where informal recycling operations break down boards by hand or use acid baths. This pollutes local soil and water.”
So, are AI engineers and developers responsible for this environmental degradation? According to Rijin, AI developers do not actively think about the environmental impact right now. Their main focus is usually on accuracy, speed, and performance. He says, “Environmental sustainability often takes a backseat because it’s not built into the development culture yet.”
Meanwhile, companies are not required to disclose how much energy or water their AI systems use. They focus more on business metrics — speed, accuracy, and ROI. Rahul laments, “The general public and policymakers are not aware of the effects of AI on the environment. They can’t connect the dots about how generating one image or asking one question can affect the environment.”
However, there’s also a push toward efficiency. “Companies are now looking at using low-power chips, better parallelism, and optimised training techniques,” states Rijin.
Solutions ahead
AI engineers and developers are looking for the blind spot and are trying to find solutions to solve the environmental puzzle.
With AI, a general rule of thumb is: the bigger the model, the better the performance. Hence, the next-generation accelerators like NVIDIA’s Blackwell series and Google’s TPU v5 deliver up to twice the performance per watt compared to the previous generation. “That improvement can cut the electricity needed for a given training run by 30 to 40 percent,” shares Harsh.
At the same time, software techniques like model quantisation help reduce computation without sacrificing accuracy. “When you combine those hardware and software gains, the energy needed per training job has dropped, but only partially. Model sizes and dataset scales continue to grow, so total consumption remains high,” adds the entrepreneur.
While amends are being made for the future, the present is still in crisis. G Sundarrajan, an environmental activist with Poovulagin Nanbargal, says, “AI has an impact on climate and global warming. People might say it is not direct. But the power consumption results in carbon emissions. We are talking about mitigating carbon emissions, and when you set up a data centre as a backbone, you need more power, which means you build more thermal power plants and more carbon emissions and more destruction.”
As a solution to this rippling effect, Sowmya Dutta, trustee of Movement for Advancing Understanding of Sustainability and Mutuality (MAUSAM), suggests the government intervene. He mentions, “The government should take recognition, and demand that the number of searches are brought down to a manageable energy level. There could be better algorithms to do this.”
He adds that a policy should ensure that the energy utilised and released should be “non-carbon, at least non-impactful waste.” Even before the government could step in, service providers are increasingly switching to renewables, air-cooling, and certified circular-economy recycling.
While AI contributes to environmental challenges, it’s also part of the solution by improving efficiency and reducing its impact. In Europe, some enterprise clients now ask about energy use and compliance with green standards, a step towards a less-destructive future.
“AI is being used to optimise cooling systems in data centres, predict energy loads, and even design more efficient chips. There’s a kind of circular benefit here — AI can help monitor and reduce its own impact, if we design it with that goal in mind,” concludes Rijin.
While you might be mindful of the water usage at home, consciously switch off lights and fans when not in use, and try to reduce your carbon footprint, you might still be contributing to environmental damage. From this World Environment Day, make it a point to pause before you ask the bots to write your assignments, help you get on an Instagram trend, or aid in coping with personal struggles, and reflect on the harm you might be causing.