The Price Behind Your AI 80s Photo!

In the last two-three days, if you opened Instagram, you’d have suddenly seen people posting AI-generated 80s style images. A hairstyle that is from the 1980s, makeup, background, clothing, everything is changed in those images, to make it look like it is straight from that decade, including film grain and faded colors. These images look like what the earlier generation would have kept in their drawers or albums. Even opposing political parties in India took potshots at each other using the trend, from their official pages.

This feels fun, it’s almost like time travel. But there’s something easy to forget when an image appears on your screen in a few seconds: there is a lot happening behind that button. Your photograph has to be processed by powerful computers in a data centre somewhere. Those computers use electricity, generate heat and need to be cooled. And at the same time, you’ve handed an AI system something far more personal than a prompt: your face.

The environmental cost of one AI image isn’t enormous. A UN assessment states that an image generation can require about 1500 times the energy of a text-based AI prompt. It also has a water footprint of about 29 ML. It might not seem like a lot, but that’s where the problem is. An average AI image prompt takes several iterations to get it right. You generate an image and you feel it doesn’t look right, you change the prompt and generate another image, then you tweak the instruction further and generate another one and it can go on till you are satisfied with the output. There are millions of people doing the trend and complex instructions can consume more, it all adds up.

Multiply that behaviour across millions of users and the tiny cost of each image starts adding up. The United Nations University estimates that data centres could consume around 945 terawatt-hours of electricity every year by 2030, more than double their estimated 2025 consumption. It also estimates that the water footprint associated with AI could reach trillions of litres.

But the more interesting part of this trend isn’t actually the electricity. It’s what we’re uploading.
Think about the photograph you’re giving the AI. It might be a selfie, but it could also be a picture of your parents, your children, your friends or your entire family. There could be a house in the background, a street sign, a workplace, a document sitting on a table or dozens of other little clues about who you are and where you live. An image can reveal considerably more than the person uploading it may realise.

And what happens to that image after you upload it depends on the AI service you’re using. This is where things get complicated, because there isn’t one universal rule for AI. OpenAI, for example, says content from consumer ChatGPT services may be used to improve its models unless users opt out, while Temporary Chats aren’t used for training. Google has its own settings and policies around Gemini activity and uploaded content, including circumstances in which data can be used to improve its services and may be reviewed by humans.

That doesn’t mean an AI company is automatically “stealing” your photographs. That’s an oversimplification. The real question is what you have agreed to when you press upload. Can the company retain the image? Can it use the content to improve its systems? How long is it stored? Can humans review it? What happens when you delete the conversation? The answers can be different for different products, accounts and settings, and most people participating in a viral trend probably aren’t stopping to read the privacy policy before uploading their family album.

There’s another uncomfortable point here: the photograph might not belong to you alone. Your grandmother’s face, your child’s face or your friend’s face can end up inside the image you upload. They may have never agreed to have their photograph processed by an AI system. And while you can change a password after a data breach, you can’t exactly change your face.

That is what makes the 80s trend more interesting than just another internet filter. We tend to think of digital things as almost weightless. You upload a photo, tap a button and a new one appears. But underneath that simple interaction is a physical infrastructure consuming electricity, water, chips and hardware. And alongside that physical cost is a data exchange that is becoming increasingly intimate: we are giving AI systems our faces, our voices, our documents and our memories.

None of this means you shouldn’t make the photograph. The point isn’t that one AI selfie is an environmental disaster, or that every AI company is quietly building a giant database of everyone’s faces. The point is much simpler: AI doesn’t make the costs disappear just because the experience feels effortless.

That 80s photograph might look like a memory from another era. But creating it requires resources in the real world, and, perhaps more importantly, it requires you to decide how much of yourself you’re comfortable handing over to a machine.