How Much Electricity Does AI Use? The Energy Cost of AI Video, Images and Music

Artificial intelligence is becoming increasingly visible in entertainment, from generated images and music to dubbing, visual effects and rapidly improving AI video. What audiences do not see is the electricity required by the powerful computers working behind every finished image, voice or clip.

That does not mean creating one AI picture or asking one question carries an enormous environmental cost. The more important issue is scale, because different types of artificial intelligence use very different amounts of computing power while millions of people and businesses are beginning to use the technology.

How much electricity do data centres actually use?

The International Energy Agency (IEA), an organisation that provides governments with research and advice on energy, estimates that data centres worldwide consumed around 485 terawatt-hours of electricity in 2025.

A terawatt-hour (TWh) is a unit used for measuring very large amounts of electricity, with one TWh equal to one billion kilowatt-hours. In simple terms, it is the type of measurement used for the electricity consumption of countries and major industries rather than individual homes.

In its latest update, published in April 2026, the IEA projects that global data-centre electricity consumption could rise to around 950 TWh by 2030. That would be roughly double the 2025 level and represent around 3% of electricity demand worldwide.

There is an important catch when reading those figures. They cover all data centres, not artificial intelligence alone, because the same infrastructure also supports streaming, websites, cloud storage, online services and countless business systems.

How much of the growth is being driven by AI?

Artificial intelligence is still becoming a major source of additional demand. The IEA says electricity consumption from data centres focused on AI rose by around 50% during 2025 and is projected to triple between 2025 and 2030.

This is happening while the technology itself is becoming more efficient. The IEA says the electricity needed for an individual AI task has been falling rapidly because both computer hardware and software have improved.

That might sound contradictory, but there is a simple explanation. If each individual AI request becomes cheaper to process while vastly more people begin using AI and the tasks themselves become more ambitious, total electricity demand can still rise.

Does every AI prompt use lots of electricity?

No, and this is where some of the discussion around AI energy use can become misleading. A simple text request can require dramatically less computing than some of the more demanding forms of generative AI now appearing.

A query simply means a request made to an AI system. According to the IEA's April 2026 assessment, simple text queries have become much more energy efficient, while newer applications such as video generation and advanced reasoning can require hundreds or even thousands of times more energy per query.

That figure should not be read as saying that every AI video always consumes exactly a thousand times more electricity than every piece of text. It shows how wide the difference between different kinds of AI task can become.

Why can AI video use so much more energy?

Video is particularly important for the entertainment industry because generating moving pictures is a much bigger computing challenge than producing a few lines of text.

A video consists of a sequence of frames. An AI system may need to create large amounts of visual information while keeping characters, faces, objects, backgrounds and movement consistent from one frame to the next.

Length and quality also matter. Generating a few seconds of relatively simple video is different from producing a longer, higher-resolution scene containing detailed characters and complex movement.

The United Nations University (UNU), the academic and research arm of the United Nations system, says AI's environmental footprint can vary according to factors including the model being used, output length and whether the content being produced is text, imagery or video.

That means there is no honest universal figure for the electricity required to make “an AI video”. The answer depends heavily on what is being created and how it is being created.

Where do AI images fit in?

Generating an image also involves more complex processing than many simple text tasks, but once again there is no single electricity figure that applies to every AI picture.

Resolution, model choice, the number of attempts made and other settings can change the computing involved. A creator generating one finished image is also very different from a service producing enormous numbers of images every day.

This distinction matters because environmental discussions can easily focus on the cost of one prompt while overlooking the bigger issue of how frequently these tools are being used.

What about AI-generated music?

Music and synthetic voices are another growing part of AI entertainment, but there is currently less reliable public evidence allowing us to give readers one meaningful figure for the electricity needed to produce a typical AI-generated song.

Audio generation involves producing information over time rather than creating one still result. The computing required can therefore vary according to factors including the system, length of the audio, quality and complexity of what is being created.

For that reason, claims that one AI song uses a fixed amount of electricity should be treated cautiously. Until there is better standardised reporting from technology companies, giving a precise universal figure would suggest a level of certainty that does not currently exist.

Why this matters to film and television

This is no longer simply a technology-sector question. The British Film Institute (BFI) and CoSTAR reported in June 2025 that generative AI was already being used across the UK screen industries, including animation, visual dubbing, visual effects and other parts of production.

Their report also identified high energy consumption and associated carbon emissions as one of the issues facing the industry. It called for data-led sustainability guidance and greater transparency around the environmental impact of AI tools used by creative businesses.

That does not mean the BFI is arguing against using artificial intelligence. Its report also highlights possible benefits, including faster production workflows, new creative opportunities and making some forms of content creation more accessible.

The question is therefore how those benefits develop alongside the computing infrastructure required to deliver them.

Is AI going to keep using more electricity?

Nobody can predict the exact figure with certainty because AI is changing extremely quickly. More efficient chips and software could reduce the electricity needed for individual tasks, while the rapid growth of AI video and other demanding applications could push overall consumption upwards.

The IEA's current central projection is that total data-centre electricity use will roughly double between 2025 and 2030, with AI-focused facilities growing considerably faster. Those forecasts will need updating as both the technology and the number of people using it change.

For entertainment, the biggest question may not be how much electricity one image, song or video consumes. It is what happens when AI-generated entertainment moves from something relatively experimental to something created on a huge scale every day.

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 explainers, industry coverage and stories following the growth of AI-created entertainment.

Sources: International Energy Agency, April 2026; United Nations University, June 2026; British Film Institute and CoSTAR, June 2025. The latest figures and original reports were used throughout.