Instagram’s latest AI-powered ’80s trend has transformed social media feeds into retro photo albums, with users creating portraits featuring big hair, feathered bangs, bold silhouettes and acid-wash denim.
However, the viral trend has also renewed concerns about the environmental impact of AI-generated images.
Creating an AI image may take only a few seconds, but the process relies on energy-intensive data centres, water-based cooling systems and resource-heavy hardware. Research by Carnegie Mellon University and Hugging Face suggests that generating one AI image can consume between 0.01 and 0.29 kilowatt-hours of electricity.
While that may seem minimal, the impact becomes significant when millions of users generate multiple versions of the same image.
The International Energy Agency has warned that global electricity demand from data centres could more than double by 2030, largely due to the growth of AI. Data centres also require substantial amounts of water to prevent servers from overheating.
Researchers at the University of California, Riverside, have estimated that training GPT-3 required around 700,000 litres of freshwater.
The environmental cost also extends to the production of GPUs and other hardware, which depends on mining, semiconductor manufacturing and the use of critical minerals.
The issue is not necessarily about abandoning AI, but using it more thoughtfully. Generating one image may appear harmless, yet repeated use of AI tools for short-lived social media trends can add to the technology’s overall environmental footprint.