SADAF SULEMAN

The latest 1980s AI photo trend has turned social media into a gallery of retro portraits. But behind every digitally transformed image is a physical infrastructure consuming electricity, generating heat and relying on water for cooling.

The 1980s have returned to social media.

Not through old family albums or photographs stored in dusty boxes, but through artificial intelligence.

Over the past few days, social media feeds have been filled with AI-generated portraits that imagine users as they might have looked in the 1980s with retro hairstyles, vintage clothing, dramatic lighting and grainy film effects. The trend has spread across Instagram and other platforms, drawing participation from ordinary users, influencers, celebrities and public figures. (The Indian Express)

The process is deceptively simple. Upload a photograph, enter a prompt and wait a few seconds for an AI-generated version to appear.

But while the image may appear almost weightless on a phone screen, the process behind it is not.

Every AI-generated image depends on physical computing infrastructure, powerful servers housed in data centres, consuming electricity and generating heat. Depending on how those facilities are designed, cooling that infrastructure can also require significant quantities of water.

The rapid spread of the 80s trend raises a question that extends beyond nostalgia: what happens when millions of digital experiments, repeated and regenerated across social media, add to the growing demand for energy-intensive AI infrastructure?

A viral trend but no reliable headcount

There is currently no independently verified figure showing exactly how many people have participated in the 80s AI photo trend.

That is an important distinction.

Social media trends are often described as involving “millions” of users before reliable platform-level data is available. Reporting over the past few days confirms that the 80s AI trend has spread widely across social media and has been adopted by celebrities, politicians and ordinary users. But no credible global figure has yet established precisely how many images have been generated or how many people have participated.

For that reason, assigning a number to the trend would be speculation. But the absence of a precise headcount does not make the environmental question insignificant.

The larger story lies beneath the viral photographs: the rapidly expanding infrastructure required to support artificial intelligence.

The infrastructure behind an AI image

For most users, artificial intelligence appears to exist entirely inside a screen.

A photograph is uploaded. A prompt is typed. Seconds later, an entirely new image appears. Behind that process, however, are data centres filled with computing equipment.

UNESCO warns that generative AI is becoming increasingly resource-intensive, requiring enormous computing power, substantial amounts of electricity and fresh, potable water for cooling. The organisation estimates that AI could consume between 4.2 and 6.6 billion cubic metres of water annually by 2027, an amount greater than Denmark’s annual water use, according to the figures presented on its green digital transformation initiative. (UNESCO)

UNESCO also estimates that training a single large language model can require around 50 gigawatt-hours of electricity, while one billion daily ChatGPT queries could require approximately 124 gigawatt-hours of electricity. These figures are estimates and should not be interpreted as the footprint of an individual prompt or image, which can vary considerably. (UNESCO)

That distinction is crucial.

There is no single amount of water that can accurately be assigned to every AI-generated 80s portrait.

The environmental footprint of an individual request depends on the AI model being used, the complexity of the task, the hardware running it, the location of the data centre, the cooling technology, use freshwater for cooling, and the electricity source.

The concern, therefore, is not necessarily one photograph.

It is scale.

One image is not the story. Repetition is.

Viral AI trends are rarely limited to one request per user.

A person may generate an image and dislike the result. They try again.

They change the hairstyle. They replace the background. They adjust the clothing. They generate multiple versions before choosing one to post.

The same pattern is then repeated by thousands,  potentially millions  of other users.

Social media rewards participation, and participation encourages repetition.

This is where an apparently harmless digital trend becomes part of a much larger environmental question.

The International Energy Agency estimates that data centres consumed approximately 415 terawatt-hours of electricity in 2024, representing around 1.5 per cent of global electricity consumption. Their electricity consumption has grown by around 12pc annually since 2017, more than four times faster than overall global electricity consumption. (IEA)

The IEA projects that global electricity consumption by data centres could more than double to approximately 945 terawatt-hours by 2030. Artificial intelligence is identified as the most important driver of this increase, alongside other expanding digital services. (IEA)

The 80s AI photo trend is only a tiny part of this global growth. But it illustrates a broader shift: artificial intelligence is moving rapidly from specialised applications into everyday life including entertainment.

The climate question behind the screen

Using more electricity does not automatically mean producing more carbon emissions.

The climate impact of AI depends significantly on how the electricity powering data centres is generated.

According to the IEA, renewable sources currently supply around 27pc of the electricity consumed by data centres globally, while coal accounts for about 30pc, although the energy mix varies significantly from one region to another. (IEA)

The IEA projects that renewables will meet nearly half of the additional electricity demand created by data-centre growth in the coming years. However, natural gas and coal are also expected to remain part of the energy mix used to meet rapidly rising demand. (IEA)

A more accurate concern is this: as AI use expands, it increases demand for electricity and data-centre infrastructure. The climate consequences depend partly on whether that additional demand is met through renewable or fossil-fuel-based energy.

At the same time, water remains a separate environmental concern.

Data centres can create local pressure on water resources, particularly where facilities depend on freshwater cooling and are located in water-stressed regions. UNESCO has specifically called for greater transparency around the energy, carbon and water footprints of AI and has argued that raising awareness of these environmental costs can encourage more eco-conscious use of the technology. (UNESCO)

The illusion of a weightless digital world

The popularity of AI image trends exposes a contradiction in the way people think about technology.

Digital activities often feel immaterial.

Unlike driving a car, switching on an air conditioner or buying a physical product, generating an AI image leaves no immediately visible trace.

There is no smoke.

No fuel tank.

No physical waste left on the user’s desk.

The environmental infrastructure is somewhere else.

Behind the image are servers.

Behind the servers is electricity.

Behind the electricity is an energy system.

And, depending on the data centre, behind the cooling system may be water.

The user sees only the final portrait.

The physical resources required to create it remain invisible.

Beyond the 80s trend

The 80s trend will eventually disappear.

Another AI trend will replace it.

Before retro portraits, social media saw waves of AI-generated Studio Ghibli-style images and digital figurines. The next trend may involve another decade, another art style or increasingly sophisticated AI-generated video.

The trend itself is not the central environmental problem.

Nor is the problem an individual choosing to generate a nostalgic portrait.

One AI image is not responsible for the climate crisis.

The larger question is what happens when artificial intelligence becomes embedded in everyday behaviour, when millions of people repeatedly use increasingly powerful systems for entertainment, creativity and communication.

UNESCO’s response to this challenge is not to argue against AI, but for more sustainable AI: smaller and more efficient models, greater transparency around energy and water consumption, and greater awareness among users of the environmental costs behind digital technologies. UNESCO says that relatively small changes in how large language models are built and used can substantially reduce energy consumption without necessarily compromising performance. (UNESCO)

That may be where the conversation needs to move.

Not towards blaming people for following the latest social-media trend.

But towards asking whether the technology industry and increasingly, its users are paying enough attention to the resources required to make such trends possible.

The 1980s may have returned to our screens.

But the environmental questions raised by artificial intelligence are firmly about the future.

As AI becomes a routine part of everyday life, its most significant environmental cost may not be visible in the image we generate, but in the water, electricity and infrastructure required to generate millions more like it.