Ctrl+Z won’t undo this: The hidden environmental cost of AI image trends
By Nandini Bhasin, Student of Class XII, DPS, Pune
Viral AI photo trends may look harmless, but every image is powered by energy- and water-hungry data centres.
The 80s retro trend and the Ghibli trend have come and gone, but they have left a mark on the planet that can’t simply be undone with Ctrl+Z.
Retro-style AI portraits, with their big hairstyles, nostalgic backdrops and vintage clothing, flooded social media feeds. The images may look harmless. Alongside the debate over copying original art styles for entertainment, there is a less visible cost: behind every AI-generated picture sits a chain of data centres, and their environmental footprint is becoming hard to ignore.
A warming world, a hungrier grid
The stakes are high. According to recent UN climate assessments, the world is expected to exceed 1.5°C above pre-industrial levels in the coming years. There is a 91% chance that at least one year between 2026 and 2030 will temporarily exceed 1.5°C above the 1850–1900 average, and a 75% chance that the five-year average for 2026–2030 will do so. Against this backdrop, every new source of electricity demand matters, and AI is one of the fastest-growing.
Image generation is particularly energy-intensive. A 2026 United Nations University (UNU) assessment estimated that a typical AI-generated image requires around 1,450 times the energy of a basic text-classification task. Research from Carnegie Mellon University and Hugging Face estimated that generating AI images can use between 0.01 and 0.29 kilowatt-hours of electricity, depending on the model and process used.
The infrastructure behind this is expanding rapidly. In North America, MIT reported that data-centre power requirements nearly doubled, rising from 2,688 megawatts at the end of 2022 to 5,341 megawatts a year later.
The problem of scale
Energy use per image is only part of the concern. The bigger issue is the sheer volume that viral trends produce. Free versions of popular tools typically limit how many images a user can create each day, but when millions of people join a trend at once, the collective total soars. The UNU estimates that more than 15 billion images were generated using text-to-image systems during 2022 and 2023, with roughly 34 million AI images being generated each day in 2023.
Electricity is not the only resource at stake. Data centres also need large amounts of water to keep their systems cool. Projections suggest that data centres could require around 945 terawatt-hours of electricity a year by 2030, with AI a major driver of that growth. According to the UNU’s 2026 assessment, the associated water footprint could reach about 9.3 trillion litres, while the land needed to generate that electricity could exceed 14,500 square kilometres.
Efficiency as the first line of defence
Efficiency measures offer the quickest and cheapest way to address these problems. Beyond cutting emissions, greater efficiency can lower costs for communities living near power plants and ease the strain on electricity grids. Sustainability also depends on asking harder questions: Where is the electricity coming from? Which water sources are being drawn on, and at whose expense?
According to Hongpeng Lei, Chief of the Mitigation Branch at the UN Environment Programme (UNEP), sustainable procurement, which shapes how data centres and their equipment are designed and operated, helps ensure that installations are energy-efficient and climate-friendly. Procurement specifications can set clear expectations for energy efficiency, cooling performance and operational best practice, reducing electricity demand and costs from the outset. They can also promote transparency through common performance metrics.
The UNEP-led United for Efficiency initiative is supporting governments in these areas, including by developing sustainable procurement guidelines for data centres and computer servers. These guidelines help governments, companies and data-centre operators apply such measures consistently and at scale.
In addition, UNEP and the International Energy Agency (IEA) run the Demand-Driven Electricity Networks Initiative (3DEN), supported by the Italian government. The initiative backs smart technologies that make power systems more efficient and help integrate renewable energy into grids, allowing electricity to reach millions more people without increasing emissions. In India, for example, 3DEN supported a consortium that developed a digital twin of electricity distribution grids, which improved reliability and outage management and reduced costs for 20,000 consumers.
The way forward
More solutions will emerge that boost productivity without making the planet pay the price, but only if environmental impact is built into decisions from the start. Despite the challenges, AI data centres can play a valuable role in a country’s development: they drive economic growth, keep national data secure and build digital independence. The goal is not to abandon the technology, but to make sure the next viral trend doesn’t leave the planet paying for it.