Authors: Rasika & Keaobaka Medupe
Plastic is well acknowledged as the common enemy, poisoning our land, water and bodies. What if we tell you AI is the same? It is affecting our land, water and ability to think. As AI seeps deeper into our lives, leaving us with no true alternative, just like plastic, only a small fraction of the population is thinking about the large negative impact AI will have on our society.
Notwithstanding, the applications of AI could be truly revolutionary, helping solve major problems such as early cancer diagnosis and treatment.Nevertheless, we are irrevocably becoming dependent on the simplicity and accessibility of AI for less important aspects of our lives. Planning our holidays, deciding on the best vacuum cleaner to buy and writing our messages are just a few examples of day to day AI-crutches. It is this widespread and unnecessary use, which is rapidly developing into large scale problems.
Convenience always comes at a cost, and once we get accustomed to convenience, there is no way we are letting go of it, no matter the cost. Through this article, we hope to take a moment of your time to think about the energy, water, land and communities that are being exploited and impacted by our frivolous use of AI.
Energy Consumption and Resource Depletion
When we access AI from our devices, we forget that there is an entire physical infrastructure behind these services. Behind every AI interaction are large data centres that require vast amounts of resources such as water, electricity and raw materials. Creating a data centre means using materials such as metals mined or acquired from other sources. Once these machines are set up in the form of data centres, the processing requires a large amount of electricity to keep it constantly functional. To put this intro perspective, the electricity usage of AI is set to consume as much energy as a small country like the Netherlands. Some even predict that AI has the potential to become one of the biggest energy consumers.
Furthermore, the United Nations identifies three types of environmental impacts fromAI: direct, indirect and high-order effects. The direct impacts come from the greenhouse gas emissions resulting from energy and water usage, computations and mining for raw materials. The indirect impacts come from GHG emissions due to the application of AI and machines. Thirdly, the use of AI can worsen the already existing gaps in environmental justice and broaden biases and inequalities.
These trends suggest that as the usage and reliance on AI increases, so too will the resources it requires. For example, it is predicted that the energy demand for AI will increase by 165% by 2030. As humans fight to reduce their energy consumption, AI seems to be making all of that effort futile. Our electricity, no matter where it comes from, renewable or non-renewable sources, has a carbon footprint on the planet. While humans and animals struggle to adapt to climate change, we have introduced AI to worsen the situation.
An additional worrying aspect of our reliance on AI is that data centres, as the backbone of AI infrastructure, are expected to expand by 28% by 2030. AI-related energy consumption could account for 3 to 4% of global electricity use by the end of the decade, with associated carbon emissions potentially doubling in the same period.
Air Pollution and Public Health Risks
The operation of AI data centres often relies on fossil-fuelled power plants and diesel backup generators, both of which emit hazardous pollutants such as nitrogen oxides and fine particulate matter. These emissions are linked to increased rates of respiratory illnesses, cardiovascular conditions, and elevated cancer risks among nearby communities. A recent model projects that, by 2030, the US data centres alone could be responsible for nearly 1,300 premature deaths annually, creating a public health burden valued at more than $20 billion.
Noise Pollution and Community Impact
The expansion of data centres also presents challenges beyond air quality. The constant operation of servers and cooling systems produce significant noise pollution, which disrupt surrounding communities. Prolonged exposure to such noise can contribute to stress, headaches, and sleep disturbances, while also negatively impacting wildlife. Similar to the effects of marine engines and sonar on aquatic ecosystems, noise from data centres can alter land animals’ migration patterns, forcing them to seek new habitats.
Land and Water Usage
While discussions of AI’s environmental footprint often focus on energy consumption and greenhouse gas emissions, its impacts begin much earlier with the land and water required to build and operate its infrastructure.
Most large-scale AI systems are hosted in sprawling data centres, often operated by major cloud providers. These facilities demand vast tracts of land for their construction, leading to habitat destruction, soil degradation, and disruption of local ecosystems. Globally, the United States, Germany, and the United Kingdom host the largest number of data centres, while markets in Latin America are expanding rapidly, with Brazil (representing an estimated value of 1.4% of the global market) and Mexico leading the region in both area and energy capacity.
Water consumption presents an equally pressing challenge. Beyond the water required for hardware manufacturing, the operation of AI systems demands significant volumes to cool servers and prevent overheating. Training a single AI model with the computing capacity of a human brain for one year can consume approximately 126,000 litres of water. Even everyday usage has hidden costs. Studies suggest that generating responses through platforms such as ChatGPT consumes roughly 500 millilitres of water for every 5 to 50 prompts. These demands place AI infrastructure in direct competition with surrounding communities for already scarce water resources.
The human impact of such competition is stark. Residents living near Meta’s data centre in Iowa (Beverly Morris), for example, report low water pressure and sediment-contamination supplies, rendering water unfit for drinking, cooking, or even bathing. Such water insecurity carries severe public health complications: excessive industrial withdrawal of potable water exacerbates the spread of waterborne diseases, increases risks of dehydration, and undermines basic hygiene. Communities already vulnerable to climate stress are therefore burdened with additional health challenges driven by the expansion of AI infrastructure.
In another instance, the company 1778 Rich Pike plans to build a data centre campus on 1000 acres of land. To ensure that the temperature of the centre is maintained, the project would include two water sheds. Additionally, every building is envisioned with its own well. Along with this, the company also plans to build a gas power plant to fulfil the energy demand. Understandably, local residents have voiced concerns about the resource strain such a facility would place on them. Importantly, this phenomenon is not isolated to the US, as Asia too is seeing a large growth in data centre demands, raising similar challenges for local environments and communities.
Innovative solutions are being tested, but they come with their own risks. In China, for instance, an underwater data centre has been constructed in the ocean and connected to a nearby offshore wind farm to reduce energy consumption.While this may appear promising, it introduces new ecological threats. Marine ecosystems, already under stress from rising water temperature, face further disruption from the heat, noise, and physical presence of submerged infrastructure. Thus, even forward-looking approaches to AI infrastructure risk compounding existing environmental challenges.
Risk To Animals
Beyond the environmental and social challenges that AI infrastructure expansion poses, it is important to recognise the risks to wildlife and broader ecosystems. Habitat destruction from large-scale data centre construction displaces species, disrupts migration patterns, and degrades biodiversity. Both marine and land animals face threats as experimental approaches introduce heat, noise, and physical disturbances into already fragile environments.
At the same time, it is important to recognise that AI itself is not inherently detrimental. Advocates point to its potential contributions to environmental sustainability.
AI’s potential for environmental solutions
It is undoubted that AI is increasingly being optimised. Over time, the greenhouse gas emissions, energy requirements, and water consumption associated with AI systems are likely to reduce over time. Some data centres, for example, have already adopted closed water loops to reduce water consumption. Should this progress ease our concerns? The answer remains complex. While the resources required to process a single AI request may decline in the future, the rapid growth of AI adoption and integration ensures that overall demand will continue to rise. This surge in usage is likely to offset efficiency gains, leading to higher aggregate consumption of resources no matter the smoke screen mitigation steps promised to be taken.
Climate change mitigation and adaptation are also some of the most important positive applications of AI. These range from predicting weather patterns, processing data more efficiently, and proposing solutions to adapt to climate change. These tools can provide valuable insights and help societies prepare for the impact of climate change effectively. Nevertheless, AI might be a part of the climate change solution, but it is also a part of the problem. This raises a critical question: is it more effective to channel resources into managing the consequences of AI’s environmental impact, or to limit the scale of the problem at its source? To borrow an analogy, why invest energy into effectively extinguishing a fire, when the more sustainable option may be to prevent it from igniting in the first place?
A Deeper Dilemma
AI’s footprint is already visible in the high energy demands of data centres, the extraction of raw materials, the escalation of e-waste, extensive water consumption, and widespread land alteration. These environmental costs are inextricably linked to social costs, which disproportionately fall on communities in regional and developing areas, those least likely to benefit from AI innovation.
Proponents of AI often highlight its potential to improve efficiency or even support climate change adaptation. Yet these promises cannot obscure the pressing reality: the infrastructure that sustains AI is itself a significant driver of environmental harm. The danger lies in the speed of its expansion, technologies introduced without proper safeguards rarely wait for regulation to catch up, and the damage they cause is often realised only once it is too late.
The urgency, therefore, lies not in dismissing AI altogether, but in confronting its environmental burden head-on. Raising awareness, demanding transparency from technology providers, and establishing robust sustainability frameworks must become immediate priorities. Without such measures, AI risks following the arguably destructive path of plastic which was once celebrated as transformative, now recognised as one of the planet’s most pervasive pollutants. Only by acting decisively now can we prevent AI’s “solutions” from deepening the very crisis they claim to address.