Artificial intelligence is being hailed as a breakthrough in the fight against climate change. It forecasts floods before they strike, balances energy grids and promises a cleaner future. Supporters say it could cut billions of tonnes of emissions. Yet behind the optimism lies a contradiction. The same technology that promises climate solutions is consuming vast amounts of electricity and water. Global data center growth now threatens to rival the energy use of major economies. Some experts warn AI could soon consume as much electricity as Japan. The question is no longer whether AI can help save the planet. It is whether we can manage its environmental cost before it becomes part of the problem.

AI as Climate Ally

AI offers powerful tools for cutting greenhouse gases. A study by the Grantham Research Institute suggests that AI in energy, transport and agriculture could reduce global emissions by 3.2 to 5.4 billion tonnes a year by 2035. The World Economic Forum predicts similar gains. It says AI could enable annual climate benefits of 3 to 6 gigatonnes of CO₂ equivalent by 2035. In practice, AI models already play a role in disaster planning. They help predict floods and droughts, allowing faster evacuations and better resource planning. In farming, AI systems manage irrigation and reduce waste. In energy, they balance supply and demand, making renewable sources more reliable.

The Hidden Cost of AI: Power, Water and Waste

The benefits come at a price. AI systems rely on massive computing power, which in turn demands electricity and water on an industrial scale. The International Energy Agency warns that global electricity use by data centers is expected to double by 2030, reaching 945 terawatt hours. That is close to the annual energy consumption of Japan. Generative AI, which powers chatbots and image tools, is particularly demanding. Researchers at MIT report that training a large AI model can use seven to eight times more energy than typical computing workloads. Water is another concern. Data centres use huge volumes for cooling. Some facilities consume millions of gallons a day, enough for a medium-sized city. Studies suggest AI operations could require up to 6.6 billion cubic metres of water each year by 2027. That is more than half of the UK’s annual use. Communities in drought-hit regions have already protested against new data center projects. In Latin America and parts of the US, activists argue that AI-driven water consumption worsens existing shortages. Hardware production adds another layer of impact. Chips for AI models require rare minerals. As demand soars, so does mining, which brings carbon emissions and ecological damage. Rapid hardware turnover also increases electronic waste.

The Debate Over Green Claims

Tech companies say they are improving efficiency, but not everyone is convinced. Google recently reported that its Gemini AI model uses 0.24 watt-hours of energy per text prompt and 0.26 millilitres of water, claiming a 33-fold improvement over last year. Independent experts dispute the figures. They say the estimates leave out indirect energy and water use, as well as the heavy costs of training models. They also argue that emissions accounting often uses optimistic assumptions. This lack of standardised reporting makes it hard for policymakers and the public to judge AI’s true footprint.

Towards Sustainable AI

Solutions exist, but they require commitment. Researchers are designing carbon-aware algorithms that shift workloads to times and places where renewable energy is available. Some AI models are being scaled down to reduce power use without losing accuracy. Early tests show emissions falling by up to 17 per cent. Data centers are also changing. Facilities in colder regions or near seawater can use natural cooling instead of freshwater. Google has built one such center in Finland. Campaigners say transparency must come first. They want companies to report energy and water consumption in detail, including the lifecycle emissions of hardware. Others suggest carbon tariffs for data centres and stricter efficiency standards.  AI has the power to accelerate climate action. It can help predict disasters, manage energy and reduce waste on a global scale. Yet without strong regulation and accountability, its own environmental cost may outweigh the benefits. The challenge is clear. Can we deploy AI to fight climate change without letting it become another driver of the crisis?