Retro is always in fashion. So is its appeal globally. It is no surprise that a wave of retro nostalgia has taken over global social media feeds. Millions of social media users—from ordinary citizens to high-profile politicians, sports stars and celebrities—have replaced their contemporary photographs with faded 1980s-style portraits of themselves in clothes in vogue four decades ago. These images, however, are not retrieved from family photo albums of yore. They are generated in seconds by artificial intelligence.

Mahesh Kushwaha, a researcher of AI, who studies the socio-political impacts of general-purpose technology, argues that technology companies deliberately design products to maximise engagement and sharing because they know such features drive adoption and growth.

“Countries like Nepal make this phenomenon especially visible because of extremely high social media uptake and rapid participation in global digital trends,” he said. “We saw similar patterns during the early adoption of ChatGPT.”

However, the ‘80s trend presents what experts call an illusion of digital convenience that conceals a staggering environmental toll. Behind the simple push of a button on a smartphone lies a vast global network of physical infrastructure. When a user sends a prompt, the request travels via subsea fibre-optic cables to sprawling data centres thousands of miles away. Inside these facilities, thousands of Graphics Processing Units (GPUs) run hyper-intensive mathematical computations, consuming vast volumes of electricity and requiring continuous cooling using fresh water.

International research highlights the severe disparity between generating basic text and creating complex media. A joint study conducted by researchers from Hugging Face and Carnegie Mellon University, titled ‘Power Hungry Processing: Watts Driving the Cost of AI Deployment?, examined the energy usage of 17 state-of-the-art AI models. Lead author and climate advocate at Hugging Face, Alexandra Sasha Luccioni, revealed that energy consumption across models performing similar tasks could vary by up to 46 times depending on the architecture.

People often treat generating an image online as lightly as typing out a quick search query or sending a simple text message, the report reads. However, there is a profound physical disparity. Producing a single image requires hundreds of times more computational energy than processing thousands of words of text.

The study concluded that generating a single AI image consumes an average of 2.9 watt-hours (Wh) of electricity—roughly equivalent to charging a smartphone from zero to 100 percent. Furthermore, doubling the image resolution increases computing energy demands by 1.3 to 4.7 times.

The International Energy Agency (IEA) has issued repeated warnings regarding the trajectory of global grid loads. According to the IEA’s reports, global data centres consumed approximately 415 terawatt-hours (TWh) of electricity in 2024, accounting for nearly 1.5 percent of total global demand. With rapid AI deployment, this figure rose to an estimated 485 TWh in 2025 and is projected to reach 950 TWh by 2030.

But the environmental impact of AI-generated images extends beyond electricity consumption alone.

“Data centres affect the environment through three main pathways: energy use and associated emissions, water consumption for cooling, and the material footprint of the hardware and infrastructure required to run AI systems,” said Kushwaha.

In a study titled ‘Making AI less “Thirsty”: Uncovering and Addressing the secret Water Footprint of AI Models’, a team led by Shaolei Ren, an associate professor at the University of California, Riverside, quantified this hidden consumption. The findings indicate that generating a standard AI image indirectly and directly evaporates roughly 29 millilitres of water—or approximately two tablespoons.

While two tablespoons per image may seem trivial, the collective impact of billions of iterative prompts—where users refine clothing, lighting and facial features—is staggering. Ren’s model projects that by 2027, global AI infrastructure could consume between 4.2 billion and 6.6 billion cubic metres of water annually.

What makes this situation particularly acute is that over two-thirds of the new data centres currently under construction are located in regions already suffering from severe water stress, the study says. This spatial mismatch has already sparked global friction. In Andhra Pradesh, India, Google plans for a $15 billion data centre complex face mounting local opposition over fears of ecological disruption and severe depletion of municipal drinking water supplies. Similar grassroots protests have erupted across the rural United States as tech giants attempt to shift infrastructure away from urban centres.

Tech conglomerates are fully aware of these environmental trade-offs. An internal transparency report published by Google researchers disclosed that a single text query processed by its Gemini model consumes approximately 0.24 Wh of energy and 0.26 millilitres of water. However, the report’s authors conceded that these figures represent a baseline for simple text, acknowledging that images or videos demand orders of magnitude more resource consumption.

The growing energy appetite of digital technologies has also triggered debate within Nepal’s domestic policy space. While the country’s surplus hydroelectricity presents opportunities for hosting localised green data centres, experts have cautioned against committing critical power resources without strict environmental oversight.

There are no official figures regarding how many users participated in the ‘80s trend in Nepal or how many images were generated in total in the country. Even though the servers of AI used by Nepali users are located abroad, the energy they consume and the climate change impacts driven by their carbon emissions are global in nature.

Kushwaha argued that the responsibility does not lie with users alone.

“The discussion should move beyond blaming individual users and focus equally on the design choices, business incentives and environmental accountability of AI companies,” he said.