Artificial intelligence, better known as AI, has been in the spotlight recently, mainly because of its potentially damaging effect on people’s lives and jobs, especially in the creative industries. However, many fail to notice the damage caused by its environmental impacts.

When a computer or machine is given the ability to perform tasks usually done by humans – mainly tasks relating to creativity and problem-solving – it is known as artificial intelligence. At first glance, this seems like a harmless yet useful tool – leading to its rapid growth in popularity. As of August 2025, OpenAI’s chatbot ChatGPT has around 800 million weekly active users, quadruple the 200 million users in August 2024.

AI can be used to improve productivity, make fewer mistakes, and overall is a cost-effective method to enhance your work.

However, there seems to be a problem… or two.

Water consumption

Data centres produce high levels of heat due to the immense computational power required to process millions of requests per minute. To prevent overheating, they rely on cooling systems that often use large volumes of water. While some of this water can be reused, a large portion is lost during evaporation.

Water is also lost during electricity generation to power AI systems, though the amount of water depends on the energy sources.

On average, training a single AI model, such as GPT-3, requires around 700,000 litres of water. Even a single conversation with ChatGPT uses up to 500 ml of water. With just over 122 million daily users, ChatGPT alone accounts for roughly 61 million litres of water usage per day.

Greenhouse gases

Burning fossil fuels to create energy releases large amounts of greenhouse gases, which everyone knows is detrimental to the planet. And AI is very greedy when it comes to energy use: for context, a single ChatGPT query consumes roughly ten times the electricity of a standard Google search.

Sam Altman, CEO of OpenAI, recently revealed that adding extra words like “please” and “thank you” to AI conversations costs millions of dollars in extra energy costs.

According to the MIT Technology Review, training one large AI model can produce more than 626,000 pounds of carbon dioxide equivalent – that’s nearly five times the lifetime emissions of an average American car.

Electronic waste

AI data centres also produce electronic waste (e-waste), which contains toxic substances such as mercury and lead. Around 62 million tonnes of e-waste were produced worldwide in 2022, according to the UN’s Global E-Waste Monitor 2024, and the amount of e-waste is growing five times faster than e-waste recycling programmes. When it comes to generative AI, a recent study concluded that the potential increase in e-waste could be as high as five million tonnes per year.

Using AI to combat climate change

There are ongoing efforts to make AI more environmentally sustainable, for example by tracking energy consumption and switching to renewable energy sources.

Some companies are experimenting with new cooling methods. Tech giant Microsoft, for example, is testing a fluid that boils at 10°C, making it significantly more energy-efficient than air conditioners and other cooling techniques currently used in data centres worldwide.

AI can also be used to find solutions to the climate crisis, such as designing lighter materials to reduce fuel consumption in aviation. AI satellite imagery can help detect floods, deforestation and methane leaks and even illegal fishing.

However, AI also brings a plethora of ethical concerns – such as the spread of misinformation on search engines and using copyrighted creative work to train AI models – which need to be addressed separately.

Despite the possibility that AI may be environmentally friendly in the future, this may not be achieved fast enough. While achieving such a goal seems daunting, every step toward sustainable AI development counts. Scientists warn that 2030 – just five years away – is probably the point of no return, when global warming exceeds 1.5°C and its effects become irreversible.