A Ghibli-style portrait, an AI action figure, a vintage photograph or a highly edited version of an ordinary selfie can take only a few seconds to create. For the person sharing it, it may feel like just another social media trend. But behind that quick result is a network of data centres, servers, cooling systems and electricity that makes generative artificial intelligence possible.
The environmental cost of AI is easy to overlook because the process happens on a screen. In 2025, social media was flooded with Ghibli-style images, boxed action figures, images in a red saree, vintage portraits and 3D avatars. More recently, AI tools have continued to fuel image trends, including nostalgic and retro-style photographs. While the trend changes, the infrastructure supporting it in the background remains the same. The issue is not a single person generating one image. The larger environmental concern comes from the scale at which these tools are now being used. When millions of people repeatedly generate images, edit photographs, create videos or ask AI systems to perform increasingly complex tasks, the demand on the infrastructure behind them also increases.
The electricity behind the screen
AI systems operate through data centres containing large numbers of servers. These machines require electricity not only to process requests but also to store information and keep the equipment running at the required temperatures.
The International Energy Agency estimates that data centres consumed around 415 terawatt-hours of electricity in 2024, about 1.5 per cent of global electricity consumption. It expects global data-centre electricity use to roughly double to around 945 terawatt-hours by 2030, with AI one of the main drivers of that growth. While simple text generation is considerably less energy-intensive, video generation, reasoning and agentic AI can use hundreds or even thousands of times more energy per task.
It is not only about carbon
Electricity use is only one part of AI’s environmental footprint. Data centres also require water, particularly for cooling, while expanding digital infrastructure requires land, construction materials and hardware. A 2026 report by the United Nations University Institute for Water, Environment and Health estimates that data centres could have an associated water footprint of about 9.3 trillion litres a year by 2030, along with a land footprint exceeding 14,500 square kilometres. The scale of environmental pressures becomes easier to understand when viewed against the wider climate events being witnessed around the world. The recent floods in Nepal have brought renewed attention to the growing risks faced by climate-vulnerable regions. The United Nations has described the disaster as an example of the losses increasingly faced by countries vulnerable to climate change.
The hardware has a footprint too
AI also depends on specialised computer chips, servers and other equipment. Manufacturing this hardware requires raw materials, energy and water. As technology develops and companies expand their computing capacity, older equipment eventually has to be replaced. The United Nations University estimates that AI infrastructure could generate up to 2.5 million tonnes of electronic waste annually by 2030. That adds another layer to the environmental cost, beyond the electricity used each time an AI tool is opened.
The growing popularity of generative AI therefore places environmental costs across several stages, from producing chips and building data centres to generating electricity, cooling servers and eventually disposing of electronic equipment.
A trend that keeps moving
The popularity of AI-generated content shows how quickly a digital trend can spread. The Ghibli wave was followed by action figures, vintage portraits and 3D figurines, while newer tools continue to introduce different ways of transforming photographs and creating content. Each trend may be temporary, but the infrastructure supporting the technology is expanding. At the same time, AI is becoming more efficient. The IEA says energy use per AI task has fallen significantly as hardware and software improve. However, the same improvements can make AI cheaper and easier to use, encouraging people to use it more often.
For the user, the process remains simple: upload a photograph, type a prompt, and wait for the result. Behind that convenience, however, is a physical system consuming electricity, water, land, and hardware. As AI becomes an ordinary part of social media and everyday digital life, the adverse impact it has on the environment will become harder to ignore over time.
