The arrival of OpenAI's Sora system, capable of generating realistic videos from simple text instructions, marks a technological milestone. It's a marvel that opens up countless creative possibilities, but it has a catch: its impact on energy consumption, the water used to cool data centers, and our global sustainability.
When we generate a video with AI, we're asking for much more than text or even images: we're demanding the recreation of movement, light, sound, texture, and so on. This translates into enormous computing power. A recent study on "Sora-type" video generators concludes that these models have a significantly higher carbon footprint than AI models that only work with text.
To give some approximate figures:
. Training a video model requires thousands of chips running for months and consuming gigawatt-hours of power.
. Using these models—even though each individual use consumes much less—is multiplied by billions of times per day. And that “less” becomes “a lot.”
Furthermore, data centers require enormous volumes of water for cooling. For example, one estimate says that generating a 60-second video could consume “as much as charging 1,800 mobile phones” or keeping twenty 1,000W heaters running for an hour.
In other words: the training phase is brutal, and the usage phase also has its hidden cost, especially if we consider the millions of times it is used every day.
It's true that technology is improving. Manufacturers design more efficient chips, data centers optimize their cooling, and AI architectures are refined. But here's the paradox: demand is growing faster than efficiency. Because when you multiply an operation millions of times a day, the impact starts to be serious.
The conclusion is clear: the "free" for the user hides a real price for the planet's energy system.
Against this backdrop, the European regulation, the Artificial Intelligence Act emerged, which came into force on August 1, 2024. This law requires developers to document the computing and energy needs of their AI systems. This is good news, but the road will be long: full implementation will take until 2026-2027.
Transparency is key: without public data, we can't know how much water, electricity, or carbon footprint a service we use almost without thinking is generating.
Because AI that generates videos is no longer just for movies or advertising: it's in apps, entertainment, education, and social media. Every video created, edited, or posted with an AI "button" has an infrastructure cost. And that cost is paid by the electrical system and the environment.
Furthermore, if use grows unchecked and the energy source remains fossil fuels, we are increasing emissions and consuming resources that could be allocated to other urgent priorities (such as health, education, and climate change).
What can we do as responsible users?
. It's not about stopping using AI video: it's about using it judiciously.
. Choose moderate resolutions, avoid generating unnecessary things, and be aware of the "easy button."
. Demand transparency from platforms: How much energy do they consume? What cooling system do they use?
. Support policies and standards that guarantee the efficiency and sustainable use of AI.
AI-generated video is an impressive technological leap. But it also confronts us with a reality: the energy and water consumption of this technology is greater than we think. While we celebrate advances, we must not forget that our sustainability is above any technological toy.
The promise of AI-generated video is real; efficiency is progressing, but uncontrolled mass use can take its toll. And now that European regulations are beginning to demand accountability, it's important to observe, ask questions, and decide responsibly.
Because AI not only transforms the content we see, but also the energy we consume. And that really matters.
You can watch the SORA presentation video:
EN 
ES