Apr 2, 2025

Take OpenAI’s latest Ghibli-style image update, for example. In just a few days, it went viral, flooding social media with dreamy, hand-drawn portraits. You’ve probably seen friends, family, or even yourself experimenting with it. Here’s something most people don’t think about—AI-generated images use a surprising amount of energy. Image: X AI is now a part of our everyday lives—creating art, answering questions, and assisting with all sorts of tasks. Its capabilities are incredible, but there’s a side we don’t always talk about: the environmental cost. Take OpenAI’s latest Ghibli-style image update, for example. In just a few days, it went viral, flooding social media with dreamy, hand-drawn portraits. You’ve probably seen friends, family, or even yourself experimenting with it. The demand was so high that OpenAI CEO Sam Altman admitted their GPUs were struggling to keep up. “It’s super fun seeing people love images in ChatGPT, but our GPUs are melting,” Altman wrote on X, as users flooded the platform with AI-generated art. The response was so overwhelming that OpenAI had to limit users to just three images per day to keep their systems from buckling. A few days later, Altman followed up with an exhausted plea: “Can y’all please chill on generating images, this is insane. Our team needs sleep.” But here’s something most people don’t think about—AI-generated images use a surprising amount of energy. According to a new study by researchers at the AI startup Hugging Face and Carnegie Mellon University, one AI image can consume anywhere from 0.01 to 0.29 kWh of electricity. To put that into perspective, charging your smartphone takes about 0.012 kWh. That means creating just one AI image can use as much power as fully charging your phone—or even more. Environmental burden of AI: Carbon, water, and e-waste The Growing popularity of AI comes to bear heavy energy demands. The training and running of generative AI models depend on huge data centres consuming tremendous quantities of electricity and water while leaving a heavy carbon footprint. Researchers at the University of California, Riverside, found that Microsoft used nearly 700,000 litres of freshwater to cool its data centres during GPT-3’s training—enough to produce over 370 BMWs or 320 Teslas. Even everyday AI interactions add up: generating 20 to 50 ChatGPT responses uses as much water as a 500ml bottle. A study by Hugging Face and Carnegie Mellon University estimates that producing a single AI-generated image consumes as much energy as fully charging a smartphone. In contrast, text-based AI is far less resource-intensive, with 1,000 text prompts using only 16 per cent of a smartphone charge. Next comes e-waste: AI models built on microchips manufactured using rare earth elements, which get mined in environmentally damaging ways. For every computer that weighs 2 kg, 800 kg of raw materials are required. AI: A double-edged sword AI isn't just bad for the environment—it can help too. The United Nations uses AI to detect harmful methane gas leaks from oil and gas sites, which helps fight climate change. AI can also improve energy use in power grids, transport, and industries, reducing pollution. But the more AI is used, the bigger its footprint gets. MIT professor Elsa A. Olivetti’s 2024 study, The Climate and Sustainability Implications of Generative AI, explains this problem. AI can both help and harm the planet at the same time. If it keeps expanding without eco-friendly solutions, its impact may become more damaging than its benefits. Can AI be made greener? Tech companies are now looking for ways to make AI more sustainable. Using renewable energy, improving cooling systems in data centres, and creating energy-efficient AI models can help reduce its impact. For now, next time you generate an your 'Ghibli' art AI image—whether it's a cool fantasy portrait or a digital artwork—remember that it comes at a hidden cost to the environment.