AI’s Green Dilemma: Can Intelligence Be Sustainable?

It’s a strange paradox—every tool that makes life easier often hides a complex mechanism behind it. Whether it’s a sleek gadget, a machine, or a helpful person, simplicity on the surface is powered by intricate systems or tireless human effort. The same is true for generative AI.

Each time you type a query into an AI assistant like ChatGPT, the answer may seem instant and seamless. But that response requires a surprisingly heavy backend—one that demands far more energy than a simple Google search. In fact, according to the International Energy Agency (IEA), a single AI request consumes 10 times the electricity of a typical search engine query. In Ireland, a rising hub for tech infrastructure, data centres may account for 35% of national electricity consumption by 2026, driven largely by AI’s proliferation.

Since 2012, the number of data centres has grown from 500,000 to over 8 million, and with that comes a rising environmental toll.

The Two Faces of AI

AI’s potential to address climate issues is well-known. It’s used to optimize energy use, design low-emission infrastructure, and even predict the impacts of climate change. Yet, ironically, the same systems meant to help the environment also contribute significantly to its degradation.

In 2019, a University of Massachusetts Amherst study shocked the public by showing that training a single large language model could emit around 300,000 kg of CO₂—equivalent to 125 round-trip flights between New York and Beijing.

But that’s just the beginning.

Data Centres: Energy and Water Hogs

While data centres today account for 0.5% of global combustion emissions, that number could nearly triple to 1.4>#/b### in the coming decade under high-growth AI scenarios. This increase isn't just about energy—it’s about water, too.

AI's infrastructure demands massive water resources for cooling. In fact, up to 9 litres of water can evaporate for every kilowatt-hour of energy used in data centre cooling. And if electricity is sourced from thermoelectric power plants, the indirect water footprint increases even more—around 43.8 litres per kWh withdrawn just for power generation.

Meanwhile, 1.1 billion people globally lack access to safe water, and 2.7 billion face water scarcity at least one month a year. The race to scale AI could worsen this crisis if sustainability isn’t prioritized.

Hidden Footprints: Mining and E-Waste

The environmental impact of AI also extends to supply chains and electronic waste. Manufacturing a single 2 kg computer requires 800 kg of raw materials, and the chips that power AI rely on rare earth elements often mined unsustainably.

Moreover, the disposal of data centre hardware creates e-waste containing hazardous substances like mercury and lead, further harming ecosystems and human health.

A Balancing Act: AI as a Solution

Despite its environmental costs, AI can be a powerful force for good if deployed responsibly:

Detecting methane leaks in oil and gas operations.
Improving efficiency in fossil fuel power plants and cement production.
Optimizing transportation and reducing vehicle emissions.
Smart building management, cutting energy use by 10%.
If widely adopted, AI could reduce emissions by 1,400 Mt of CO₂ annually by 2035—3 to 4 times the emissions from data centres in that same scenario. But that promise hinges on the right policies, infrastructure, and incentives.

The Path Forward: Making AI Sustainable

To manage AI’s environmental impact, the UN Environment Programme (UNEP) recommends:

Standardized measurement tools for tracking AI's environmental costs.
Mandatory disclosures on AI-related energy and resource use.
Algorithmic efficiency improvements and water recycling practices.
Greening data centres with renewable energy and carbon offsets.
Integrating AI strategies into broader environmental policies.
However, these changes require global momentum that is currently lacking. Barriers such as limited digital infrastructure, regulatory gaps, and social resistance may delay or dilute AI’s climate-positive potential.

Whether AI becomes a climate ally or adversary will depend on the decisions we make today.

References:

1. https://www.forbes.com/sites/cindygordon/2024/02/25/ai-is-accelerating-the-loss-of-our-scarcest-natural-resource-water/
2. https://www.iea.org/reports/energy-and-ai/ai-and-climate-change
3. https://www.nature.com/articles/s42256-020-0219-9
4. https://hbr.org/2024/07/the-uneven-distribution-of-ais-environmental-impacts
5. https://www.unep.org/news-and-stories/story/ai-has-environmental-problem-heres-what-world-can-do-about
6. https://news.mit.edu/2025/explained-generative-ai-environmental-impact-0117
7. https://www.scientificamerican.com/article/ais-climate-impact-goes-beyond-its-emissions/
8. https://unu.edu/ehs/series/artificial-intelligence-help-or-harm-climate#:~:text=Additionally%2C%20data%20centres%20and%20their,negative%20environmental%20impact%20of%20AI.
9. https://www.weforum.org/stories/2025/01/artificial-intelligence-climate-transition-drive-growth/#:~:text=AI%20also%20generates%20emissions%20through,to%20accelerate%20low%2Dcarbon%20technologies.
   
   Cover Image Source: Nasscom Community