Islamabad: A few words can now turn into a photograph, a portrait or an imaginary world within seconds.

For millions of internet users, that has made AI image generation feel almost weightless. A picture can be created, rejected and regenerated with little more than another tap.

But somewhere behind that tap, machines are working.

The rapid spread of generative AI is increasing demand for the data centres that train and operate these systems. The International Energy Agency (IEA) said in April that global electricity consumption by data centres rose 17 per cent in 2025, while electricity use by AI focused data centres increased by 50 per cent.

It expects total data centre electricity consumption to roughly double from 485 terawatt hours in 2025 to about 950 terawatt hours by 2030.

AI image generation is only one part of that expanding demand. But producing images can be considerably more computationally demanding than simpler digital tasks, and the amount of energy required varies significantly between systems.

A 2025 study examining 17 image generation models found a difference of up to 46 times in energy consumption between models. The researchers also found that increasing image resolution could substantially increase energy use in some systems.

That makes the environmental question less straightforward than asking how much electricity is used to create one picture.

The bigger issue is scale.

Generate an image for a trend. Generate another because the first one is not quite right. Change the background. Alter the lighting. Try a different style.

Each request can feel insignificant.

Millions of requests are not.

The physical infrastructure behind those requests is becoming significant enough to attract the attention of energy planners. The IEA says AI focused data centres are expanding rapidly and projects their electricity consumption to triple between 2025 and 2030. At the same time, advances in hardware and software are making individual AI tasks more efficient.

That creates a peculiar tension.

AI is becoming more efficient at the same time as people are using it more.

The result is that improving the efficiency of individual tasks does not necessarily mean overall resource consumption will fall.

There is also more than electricity involved. Data centres require cooling, and the wider technology supply chain depends on water and raw materials for electricity generation, semiconductor production and hardware manufacturing.

Yet most users never encounter any of this.

The electricity does not appear beside the prompt. The cooling system is somewhere else. The servers are somewhere else. The manufacturing footprint is somewhere else.

What arrives on the screen is simply an image.

That invisibility may be one of the defining environmental questions of the AI era.

The issue is not that every AI generated picture represents an unacceptable environmental cost. The IEA notes that AI itself could also help reduce emissions in areas such as energy systems and industry.

The question is what happens when a powerful technology makes creation so effortless that people stop thinking about the resources behind it.

The internet has spent years making digital consumption feel intangible.

AI may be forcing us to remember that the digital world still runs on very physical things.

And the next image may take only seconds to appear.

The infrastructure behind it does not.