The Earth's Warning to the AI Revolution

Biju Dharmapalan

bijudharmapalan@gmail.com

AI has become the hallmark technology of this era. With its applications ranging across healthcare and education to agriculture and governance, AI is revolutionising the way societies operate. Finally, the release of generative AI like ChatGPT has ushered in a new dawn, allowing machines to compose videos, analyse data, write, and even draw. AI is experiencing a surge in adoption by governments, companies and individuals. At USD 189 billion in 2023, the global AI market is expected to reach up to USD 5 trillion by 2033. Today, AI is viewed as a catalyst for economic growth and innovation worldwide.

However, there is another side to AI that no one knows of. It's not just software sitting in the clouds—AI is more of a newfound superpower. It is not simply software that is floating in the sky in the cloud; rather, it is a newfound superpower. Each time a chatbot replies, an image is generated, or an automated recommendation is made, there is a lot of physical infrastructure in data centres, high-performance computers, cooling systems, electrical power grids, water supplies, rare minerals and acres of land.

AI is not code; it is electricity, water, carbon, silicon, copper, lithium, and concrete. The United Nations University Institute for Water, Environment and Health (UNU-INWEH) recently released a report showing the environmental impact of the rapidly increasing world of artificial intelligence (AI). In short, the findings should serve as a catalyst for progress among policymakers, technology companies, and users. The report shows that training the advanced AI models demands a tremendous amount of energy.

The calculation is that GPT-4, one of the most powerful language models ever developed, used between 50 and 70 gigawatt-hours of electricity during training. This electricity is the equivalent of that consumed annually by hundreds of thousands of African residents in a part of the continent. The next model is likely to call for more. Next-generation systems may require more than 100 gigawatt-hours to train, which would generate a carbon footprint equal to that of a small town.

But with regard to training, that's not the whole picture. Everyday use results in a greater environmental burden. There are billions of AI requests around the world every day. The report estimates that inference, the primary task of generating a response to a user, accounts for 80-90 per cent of the total energy used by AI. ChatGPT is handling billions of prompts every day. Individual interactions might not seem like much, but when multiplied, they add up to enormous electricity consumption.

Rapid growth in demand is creating a boom in Data Centres. Worldwide, about 448 terawatt-hours of electricity was used by data centres in 2025. Data centres would be among the biggest power consumers in the world. By 2030, their electricity consumption will need to exceed 945 TWh, representing almost three per cent of global electricity use. The climate effects are significant. In many regions across the globe, especially in developing countries, electric power generation is still significantly dominated by fossil fuels. Therefore, the more AI is used, the more carbon will be emitted.

The UNU report estimates that electricity use in the data centres released some 189 million tonnes of carbon dioxide equivalent emissions in 2025. If these trends remain, then these emissions are likely to more than double by 2030. However, carbon is just one aspect of the issue. Importantly, the report discusses not only greenhouse gas footprints but also water and land footprints.

Data centres use a significant amount of water in order to cool the servers, which are used on a 24/7 basis. Estimates of the global water footprint of data centre electricity use are 4.5 trillion litres in 2025. It could grow to 9.3 trillion litres by 2030, meeting the domestic water needs of over a billion people. This figure will be terrifying in a world where water scarcity already exists. Many areas where data centres set up operations experience water stress.

It is possible for communities to end up competing with digital infrastructure for access to a valuable resource. Another one-of-a-kind concern that is ignored is land use. With the expansion in AI technologies, land is needed for data centres, power generation plants, transmission lines and cooling infrastructure. The land consumption associated with the predicted electric load in a data centre by 2030 is projected to exceed 14,500 km2, equivalent to the area of many metro areas.

AI is not just an energy consumer; it also requires water and land and has an increasing material footprint. High-tech chips rely on key minerals like lithium, cobalt and rare earth elements. These materials are commonly mined, but they harm the environment and cause social disruptions, especially in developing countries. When they reach the end of their life cycle, these computers and their components add to the growing e-waste pile. AI may be responsible for generating up to 2.5MT of e-waste annually by 2030, the report predicts. None of this implies that artificial intelligence is inherently dangerous. Instead, AI can be used to create fantastic opportunities for environmental sustainability. It can help with improved climate simulation, better optimisation of renewable energy systems, improved water management, ecosystem monitoring, and better disaster preparedness. AI can grow to be a strong partner on a path with humanity toward combating climate change. The hard part is making sure that the downsides of artificial intelligence are not more destructive than the upsides.

This World Environment Day, with the theme "climate action", is the perfect time to re-imagine technology. Technological improvement cannot be measured solely by economic returns or numerical capability. An environmentally responsible innovation has to be the key defining principle. Governments need to mandate clear reporting on the carbon, water, and land footprints of AI. Technology enterprises have to invest in energy-efficient models, renewable-energy-equipped information centres, and sustainable hardware design. The researchers indicate that they need to focus on "efficient AI" rather than "big AI". Users also play a part by embracing AI responsibly and being aware of the impact of their online actions.

The Earth is sending unmistakable signals through rising temperatures, shrinking water resources, and ecological stress. As AI becomes woven into every aspect of modern life, the signals we send back will matter. If guided wisely, AI can help build a more sustainable future. If left unchecked, it risks becoming another driver of environmental instability.

Biju Dharmapalan

(Dr.Biju Dharmapalan is the Dean -Academic Affairs, Garden City University, Bengaluru and an adjunct faculty at the National Institute of Advanced Studies, Bangalore)