How Bad Is AI for the Environment? The Real Cost Behind AI Entertainment

Artificial intelligence is becoming a bigger part of entertainment, from generated images, music and video to dubbing, visual effects and new tools for film and television production. As those uses grow, there is another question worth asking: what does all this new technology actually cost the environment?

Creating something with AI can look effortless, but the technology relies on physical infrastructure most people never see. Behind an AI-generated video, image or voice are powerful computer servers, specialist chips, data centres, cooling systems and the electricity needed to keep them running.

AI is digital, but the machines behind it are real

Artificial intelligence, usually shortened to AI, describes computer systems capable of tasks such as recognising patterns, understanding language or generating new content. Generative AI matters especially to entertainment because it can create text, images, audio and increasingly sophisticated video.

Much of that work happens inside data centres, large facilities filled with servers and networking equipment. Those servers use electricity and produce heat, while the hardware itself must be manufactured, transported and eventually replaced.

How much electricity do data centres use?

The International Energy Agency (IEA), which advises governments on energy, estimates that data centres worldwide consumed around 485 terawatt-hours of electricity in 2025. Its April 2026 update projects that figure could reach about 950 terawatt-hours by 2030, around 3% of global electricity demand.

A terawatt-hour (TWh) is a measure of very large-scale electricity use, with one TWh equal to one billion kilowatt-hours. Importantly, those figures cover all data centres, not AI alone, because the same infrastructure also supports streaming, websites, cloud storage and other digital services.

AI is still an important source of new demand. The IEA says electricity use at AI-focused data centres increased by 50% during 2025 and could triple between 2025 and 2030.

Why entertainment matters

Different AI tasks can require very different amounts of computing. Producing a simple text response is not the same as generating a high-quality image or creating moving video, which matters as entertainment becomes a bigger part of generative AI.

The IEA says energy use for individual AI tasks has been falling as hardware and software become more efficient. However, advanced reasoning and video generation can require hundreds or even thousands of times more energy per query than simple text generation.

That does not mean one creator making an AI image or short video has a huge environmental impact. The bigger issue is scale, as millions of people, platforms and entertainment companies generate more content.

Why does AI use water?

Computer servers generate heat, so data centres need cooling systems. Some of those systems use water, while water can also be involved in electricity production and computer-chip manufacturing.

England's Environment Agency says large data centres can consume millions of litres of water each day, although the amount varies considerably depending on the facility, its cooling technology and local conditions. There is therefore no single water figure that applies to every data centre or every AI prompt.

That distinction matters because some eye-catching claims online give the impression that every AI request has exactly the same environmental cost. In reality, the model being used, the type of content being produced and the infrastructure processing it can all make a difference.

What about carbon emissions?

Electricity use does not automatically tell us how much carbon is produced. A data centre supplied mainly by lower-carbon electricity can have a very different emissions footprint from one relying more heavily on coal or gas.

The IEA's April 2026 update projects that emissions associated with data centres could roughly double to around 350 million tonnes by 2035. It estimates this would still represent around 2% of worldwide electricity-sector emissions, which helps put the figure into context.

The date matters because an earlier IEA report published in 2025 contained a lower central forecast. The April 2026 update is the organisation's newer assessment, so keeping the date attached avoids mixing two different projections.

The footprint goes beyond electricity

The United Nations University, the academic and research arm of the United Nations system, says AI's environmental impact should not be measured through carbon alone. Its research also considers water, land and the infrastructure needed to support growing demand.

There is also the hardware itself. Specialist chips and servers need raw materials, manufacturing and transportation, while old equipment can eventually contribute to electronic waste.

These impacts are difficult to combine into one reliable number. That is why claims about the total environmental cost of AI need context rather than one dramatic statistic, particularly while the technology and the infrastructure supporting it are developing so quickly.

Can AI also help the environment?

There is another side to the argument because AI can also be used to tackle environmental and energy problems. The IEA says it could help electricity networks forecast renewable generation, identify equipment problems and make better use of existing grid capacity.

AI can also analyse large amounts of environmental data and potentially support areas such as weather forecasting and energy efficiency. Those benefits do not automatically cancel out the resources needed to operate AI, but they show why the technology cannot simply be labelled environmentally good or bad.

For the entertainment industry, there may eventually be similar questions about whether AI can reduce some physical production demands while increasing computing requirements elsewhere. That balance will become increasingly important as studios, platforms and independent creators make greater use of the technology.

So how bad is AI for the environment?

AI has a genuine environmental footprint, but there is no credible single figure that tells us how damaging every use is. The answer depends on the task, hardware, data centre, cooling system, electricity supply and the scale at which the technology is used.

For entertainment audiences, that distinction will become increasingly important as generated video, images, music and creative tools become more common. Understanding what sits behind the finished content gives us a clearer picture of both the possibilities of AI entertainment and the resources required to create it.

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 about AI-created entertainment.

Sources: International Energy Agency, April 2026; United Nations University; Environment Agency; British Film Institute.