Why Does AI Use Water? The Hidden Cost Behind AI Images and Video

Artificial intelligence is becoming part of the entertainment business, from AI-generated images and video to music, dubbing and visual effects. The finished content can appear on a screen within seconds, but the computers creating it produce heat, and keeping those machines running safely can involve something viewers rarely associate with digital entertainment: water.

That is why water use has become an important part of the environmental debate around AI. The reality is more complicated than some of the dramatic figures shared online, because there is no single amount of water used by every AI prompt, image or video.

Why does AI need water?

Artificial intelligence, usually shortened to AI, runs on powerful computer servers housed inside data centres. When those processors are working, they generate heat, so data centres need cooling systems to stop equipment overheating.

Some facilities rely heavily on air cooling, some use water and others use a combination of technologies. This means two data centres carrying out similar computing work can have very different water demands depending on their design and location.

England’s Environment Agency says some large data centres can consume millions of litres of water each day, with demand potentially increasing during warmer weather. That figure should not be treated as the standard for every data centre, because water use varies considerably according to the size of the facility, the cooling system being used and local conditions.

Water use goes beyond cooling

The water connected to AI does not only come from cooling servers. Researchers generally look at direct water use at data centres as well as indirect water use connected with electricity generation and the manufacture of the specialist computer chips that power AI systems.

Chip manufacturing can require extremely clean water during production, while some forms of electricity generation also consume water. This means the wider water footprint of an AI system can extend far beyond the building where the final request is processed.

For entertainment, this becomes particularly relevant because different AI tasks require very different amounts of computing. Generating a few lines of text is not the same as creating a detailed image or producing several seconds of high-quality moving video.

How much water does an AI image or video use?

Research published in June 2026 by the United Nations University Institute for Water, Environment and Health, known as UNU-INWEH, examined how the environmental footprint of AI can change according to factors including the model being used, output length, resolution and whether the system is producing text, images or video.

Using its methodology, researchers estimated an electricity-associated water footprint of around 29 millilitres for a typical AI-generated image, while a high-complexity AI video could reach around 4.1 litres. These figures are estimates rather than universal measurements that apply to every image or video produced.

The phrase “electricity-associated water footprint” is important. It includes water connected with producing the electricity needed for the computing, so it should not be interpreted as meaning that several litres of water are physically poured into a cooling system each time someone generates a video.

Different models, data centres, electricity supplies, cooling systems and local climates can all change the result. This is why simple claims that every AI prompt uses one fixed quantity of water can give a misleading impression.

Why do different studies produce different figures?

Water use can also be measured in different ways. Some studies examine water withdrawn from a source, while others concentrate on water that is effectively consumed and not immediately returned, such as water lost through evaporation.

Those differences matter when large global figures are compared. Two studies can both be credible while appearing to produce very different answers because they are measuring different parts of the same environmental footprint.

UNU-INWEH estimated that projected global data-centre electricity demand in 2030 could be associated with a water footprint of around 9.3 trillion litres. That should not be described as AI water use alone, because the underlying data-centre figures also include streaming, cloud computing, websites, business services and many other digital activities.

What could this mean for the UK?

The issue is particularly relevant in Britain because both AI use and the demand for data-centre infrastructure are growing. The Environment Agency has warned that significant numbers of new facilities could be developed before some of the country’s planned major new water resources become available.

Water availability is therefore becoming part of the conversation around where future data centres should be built. The Environment Agency has said the issue needs to be considered much earlier in the planning process rather than after sites have already been selected.

There is also a wider planning gap. Data centres are not yet fully factored into national water-demand planning, meaning their additional demand is not currently accounted for in the same way as some established sectors.

Industry bodies including Water UK have highlighted this as an issue as the number and size of data centres increase. Better information about future demand should make it easier for planners and water companies to understand where new facilities can be supported without creating unnecessary pressure on local supplies.

Can data centres use less water?

There are several ways the sector can reduce its water demands. Closed-loop cooling systems can reuse water rather than constantly replacing it, while some facilities may be able to use recycled or non-drinking water instead of water suitable for household supplies.

The best solution can vary according to location and climate, so there is no single cooling system that will be right for every data centre. Greater transparency would also help, because researchers and policymakers still do not have complete information about how much water every facility uses.

Why this matters to entertainment

The British Film Institute, known as the BFI, identified environmental sustainability as one of the issues surrounding the growing use of generative AI across film, television, visual effects and other parts of screen production in its June 2025 research.

The BFI also recognised the potential benefits of AI, including new creative tools, faster workflows and opportunities for smaller productions. The environmental discussion is therefore not about arguing that creators should stop using AI, but about understanding the infrastructure supporting those tools as their use becomes more widespread.

An individual creator making one image or short video represents a very different question from millions of people and entertainment businesses generating content every day. The bigger environmental issue is what happens when relatively small individual demands are multiplied across an expanding global network of data centres.

AI-generated entertainment can look entirely digital, but the systems creating it rely on electricity, hardware and sometimes significant amounts of water. As AI moves further into film, television, music and online storytelling, understanding that hidden physical footprint will become increasingly important.

For more on how artificial intelligence is changing film, television, music and digital storytelling, visit SoapLandTV’s AI & Entertainment page. You can also explore our latest AI explainers, industry coverage and stories following the growth of AI-created entertainment.

Sources: Environment Agency water-resource planning, United Nations University research from June 2026, Water UK evidence on national demand planning and British Film Institute research from June 2025.