Name: Does using AI impact the environment? - BBC World Service
Uploaded: 2026-01-07T04:42:51.000Z
Duration: 8 min 59 s
Description: Thousands of YouTube videos with English-Chinese subtitles! Now you can learn to understand native speakers, expand your vocabulary, and improve your pronunciation...

Gerund

To-infinitive

This episode isn't going to record itself.

For many of us, AI apps have become such a big part of our lives.

It's fun to get that instant response to whatever you want to ask in the moment.

But there's a lot of research into the more harmful effects of using AI too.

How problematic is it for the environment when you reach to ask your favorite chat bot about anything and everything?

That's what we're getting into in today's episode.

This is what's in the World from the BBC World Service.

And I'm joined by Laura Kress, who covers tech for the BBC.

So, I mean, I've often heard and been told AI brings environmental issues.

Break it down for us a little bit, though.

Do we know how much energy AI is actually using right now?

It isn't very clear exactly how much energy AI systems use, but

So that's the type of AI used by chatbots such as ChatGPT or Gemini consumes a lot more energy than, for example, a web browser.

And some estimate that more than 33 times as much.

So it's hard to pinpoint the exact number, but

we can very clearly say is that AI is consuming more energy than these other sources.

Well, it's not, for example, your smartphone, it's not your laptop that's using all this energy.

It's giant data centers and they are powering these AI services.

And what they are essentially is giant computers in sheds.

And because these services are so data hungry and they're going 24 7, they're using a lot of energy.

So these massive, massive data centers, and they are huge.

They can be as large as multiple football pitches.

In some cases, these big companies such as Google or Amazon, they are powering these services.

And that's where all this energy is going.

I've got that picture of the giant computers in the sheds there.

The us it's got the most data centers by quite a large margin, actually.

It's got nearly half of the world's facilities.

But to give you an idea of the rest, there's about 9,000 in total, and

There's some in Dublin, there's some in London, Asia as well.

There's data centers popping up now in India, in South Africa.

So whilst the US has a large number of data centers, they are now being distributed around the world.

And scaling it back a little bit more, do we know where that energy that AI uses comes from in the first place?

It is a mixture, essentially, because it really depends where those data centers are based.

If a data center is based in an area where that grid is being powered by fossil, fossil fuels, then that is going to be using fossil fuels a lot more.

However, if it's in an area where more renewables are powering that electricity grid, then that will be different.

So it really depends where the data centers are based.

And I was reading that the International Energy Agency has estimated that the electricity consumption of the big data centers will actually double between now and 2030.

Something like it could be equal to the energy consumption of Japan to date.

And so that is causing organisations to ask, you know, where is this power going to come from?

You know, if an area says we'll buy power from the district over, that district is also planning to buy power from that district.

So it is going to cause a lot of complications.

These data centers, of course, they require masses of electricity from the national grids to keep running.

And when I've been looking into this, I've seen plenty of areas that are concerned that their electricity grids might reach full capacity.

And then what do they do, for example, with new housing developments that are trying to get connections to the grid and they simply can't.

Okay, so we've spoken about some big stats there.

I'm wondering about the pressure being put on the big tech companies because of all this.

Are they open about how much energy they use now?

Now, tech firms, they don't really have to disclose data for what their operations are.

We don't have full access to their operations.

It is worth noting some of them have started to give a bit more data.

Google, for example, in August they released this report and they explained how much energy its Gemini apps was using for each query.

So one that falls in the kind of middle of the range of energy demand.

They said it was the equivalent of running a standard microwave for about one second.

So we are starting to get a bit more drip of information, but they're still not very forthcoming.

So let's move on to the water side of this issue as well.

Is it true that every time I use CHAT GPT I use one pint of water to cool the system, is that right?

Now, this was actually a line that's been used previously and it more related to older AI models, chat GPT3, and

then it was roughly one pint of water, 500 milliliters of water for every five to 50 user queries.

One pint or 500 milliliters every single query.

Now, more recently, Sam Altman, he's the head of OpenAI that makes ChatGPT, he said now he thinks it's about 1/15 of a teaspoon of water for a single query.

But icra, it's important to note that one query to a chatbot, that's only a small part of the water that's being used.

And why do people say that AI is using a lot of water?

So it depends where the data centers are based.

But if you imagine icra, if you put a laptop on your lap, it's going to feel very hot.

And that's essentially the same when you look at these data centers, they get incredibly hot.

So the water is basically just used to cool them down.

And it's a continuous process, so it needs to keep going.

So there's so much water needed to be used.

It's been estimated that for about each kilowatt hour of energy a data center uses, it needs 2 liters of water for cooling.

And obviously we know not everyone in the world has access to clean drinking water.

So is there an argument about the fairness in all of this that comes in as well?

Subtitles ListPlay Video

Does using AI impact the environment? - BBC World Service

episode

essentially

absolutely

equivalent