The viral ChatGPT '80s photo trend has turned social media feeds into a stream of retro portraits, with users asking artificial intelligence to reimagine their present-day photographs as if they were taken decades ago.

The trend has become popular for its nostalgic appeal. People are experimenting with vintage hairstyles, old-school clothing, grainy studio photography, warm lighting and classic family-album aesthetics.

For many users, creating the image takes little more than uploading a photograph and entering a prompt.

But behind the seemingly simple digital exercise is a much larger technological system.

Every AI-generated image requires computing power, and that computing power consumes electricity.

Data centres running AI systems also need cooling, while water and other natural resources are involved at different stages of the technology's infrastructure.

As the '80s trend continues to spread, it offers a useful example of a wider question surrounding generative AI: what happens to the environment when billions of small AI requests are made around the world?

Why generating an AI photo requires so much computing

An AI-generated photograph may appear on a smartphone within seconds, but creating it involves considerably more computing than simply storing or displaying an ordinary photograph.

When a user asks an AI system to transform a selfie, the request is processed by specialised computer hardware in a data centre.

The system analyses the instructions, interprets the original image and generates a new picture based on patterns it has learned during training.

The amount of electricity involved is not identical for every image.

It can change depending on the AI model, the hardware being used, the image resolution and the complexity of the generation process.

Creating several versions of an image also means running the system repeatedly.

Research published in 2025 that examined 17 different AI image-generation models found substantial differences in their energy requirements.

Some models were found to use many times more energy than others, demonstrating why it is difficult to assign one universal energy figure to every AI-generated picture.

A 2026 assessment by the United Nations University Institute for Water, Environment and Health provides an indication of the scale involved.

It estimates that a typical AI-generated image can use enough electricity to keep a 10-watt LED bulb running for around 17 minutes.

The comparison is not meant to suggest that generating one photograph is equivalent to leaving a household light switched on for a significant period.

Rather, it illustrates that image generation requires a measurable amount of computational energy. The bigger issue is repetition and scale.

A person participating in the viral '80s trend might generate several images before finding one they like.

They may change the outfit, background, facial expression or lighting and generate the image again.

Multiply that behaviour across millions of users and the demand placed on AI infrastructure grows rapidly.

The same UN assessment estimates that most of the energy consumed by AI systems comes during their everyday use, known as inference, rather than during the initial training of the models.

That distinction is important as generative AI becomes part of everyday internet use. AI is no longer restricted to occasional research or specialist applications.

People are using it to create photographs, videos, illustrations, advertisements, presentations and other forms of digital content.

Water is another part of AI's environmental footprint

Electricity is not the only resource involved in generating AI images. Water is also connected to the infrastructure that supports artificial intelligence.

Powerful computer processors produce considerable heat while handling AI workloads. Data centres therefore require cooling systems to prevent equipment from overheating. Depending on the facility and its cooling technology, water can form part of that process.

There can also be an indirect water footprint associated with producing the electricity used by data centres and manufacturing the computer chips and other hardware required to run AI systems.

The UN University assessment estimates that the electricity-related water footprint of a typical AI-generated image is around 29 millilitres.

That is roughly equivalent to a few tablespoons of water.

Again, this is an estimate rather than a fixed amount that applies to every image.

The actual figure can vary significantly according to the location of the data centre, its cooling technology, the source of its electricity and the efficiency of the AI hardware and model.

At the infrastructure level, however, the numbers become much more significant.

The UN University has projected that data centres supporting AI could require around 945 terawatt-hours of electricity a year by 2030 if current growth continues.

This is why researchers increasingly look at AI's environmental impact beyond individual prompts. One AI portrait has a relatively small footprint.

A global ecosystem generating enormous volumes of AI content is a very different proposition.

The environmental question surrounding the '80s trend, therefore, is not whether one person's retro selfie is going to cause significant environmental damage.

It is about what happens when millions of people repeatedly use energy-intensive systems for increasingly routine tasks.

There is also an issue of transparency. Precise calculations are difficult because AI companies do not always disclose detailed information about the electricity and water consumed by individual models and data centres.

The environmental impact of AI will ultimately depend not only on how frequently people use these tools but also on how efficiently the technology develops.

More efficient models, improved computing hardware, cleaner sources of electricity and cooling systems that use less water could all reduce the footprint associated with AI.

For users enjoying the viral '80s trend, the environmental cost of generating a nostalgic portrait is unlikely to be the first thing on their minds.

But the trend highlights a growing reality of the AI era: digital content may exist on a screen, but producing it still requires physical infrastructure and real-world resources.

The retro photograph may be designed to look like it came from the 1980s.

The technology behind it, however, represents one of the fastest-growing areas of modern computing, and its environmental consequences are becoming increasingly difficult to ignore.