How AI uses our drinking water - BBC World Service — Transcript
Full transcript
- 0:001/15 of a teaspoon. That's how much water the average single
- 0:05interaction with ChatGPT uses, according to Sam Altman, the
- 0:09boss of OpenAI. So if you type, can you help me solve this
- 0:13maths problem? That's a drop. Or can I put lime instead of
- 0:18lemon in this recipe? That's a drop. Or why is the sky blue?
- 0:22Help me write this email. Help me improve my website code.
- 0:27Mr. Altman claims there are 1 billion messages sent to
- 0:30ChatGPT every day, and ChatGPT is just one AI bot. Chuck in
- 0:36Gemini, DeepSeek, Claude and others. It's clear that the AI
- 0:40revolution is a thirsty one. Striking though it is, some
- 0:46experts are more than a little sceptical of Sam Altman's
- 0:49estimate on water usage.
- 0:52At this point, there was just not enough information for me
- 0:55to agree with or trust the number. Their number was perhaps
- 1:01referring to some tiny models. We're considering a medium
- 1:06sized large language model that's the size of GPT-3.
- 1:10Basically, if you write an email or ask some questions, if
- 1:14you have 10 to 50 queries, you're going to be consuming
- 1:17roughly 500ml of water.
- 1:19This calculation includes water used in cooling and
- 1:22electricity generation. The BBC asked OpenAI for more
- 1:26details about Sam Altman's estimate, but the company
- 1:29declined. Either way, it's clear AI uses a lot of water. But
- 1:34why? Every time you send a prompt to an AI, it has to run
- 1:39complex calculations to understand and respond. This work is
- 1:44done by the most powerful and specialised computer chips in
- 1:47the world, housed inside enormous data centres. Even before
- 1:52users can send prompts, the training process for the models
- 1:55uses the chips to carry out intense work. And all that extra
- 2:01power means the hardware can overheat and become damaged if
- 2:04not cooled properly. Most data centres use air cooling
- 2:08systems, which was fine until I came along.
- 2:12But now, because these data centres and the infrastructure
- 2:16that's going in is so much more energy intensive, there are
- 2:21liquid cooling approaches that are now being implemented.
- 2:24For liquid cooling the water must be clean to prevent
- 2:28bacteria growing or clogs and corrosion in the system, which
- 2:31means using mostly drinking water. Here's how the most
- 2:34common liquid cooling process works. It begins by piping
- 2:38coolant over the processing chips within the servers. This
- 2:42cooling liquid absorbs the heat and takes it away from the
- 2:45electrics to a heat exchange unit. Water is used to reduce
- 2:49the temperature of the coolant. The coolant then recirculates back to cool the servers.
- 2:54Meanwhile, the now hot water is piped
- 2:56to cooling towers, where a combination of fans and water vapour dissipate the heat,
- 3:01cooling the water. Some of the water
- 3:04evaporates in that process, while the rest is recirculated through the cooling process
- 3:09several times before being
- 3:10discharged back into the nearby water source. Overall, up to 80% of the water evaporates.
- 3:17What it means is that this type of water is gone, and that
- 3:20we are extracting water from a water circuit that is
- 3:24necessary for irrigation, for human consumption and hygiene.
- 3:30Communities around the world concerned about data centres
- 3:33putting stress on water sources and electricity grids are
- 3:36pushing back. Protests have been held in Spain, India,
- 3:40Chile, Uruguay and parts of the US. And it's not just the
- 3:44operations within the data centre that need water. Generating the electricity to run
- 3:49them requires a lot of water too, because power
- 3:51plants like coal, gas and nuclear heat water to create steam,
- 3:56which drives a turbine. The International Energy Agency
- 4:00has said electricity demand for AI optimised data centres is
- 4:04expected to increase by 400% by 2030 to 300 terawatt hours.
- 4:11That's roughly the electricity consumption of the whole of the
- 4:14UK for a year. And aside from electricity,
- 4:18water is also needed when manufacturing the semiconductor chips used to run AI.
- 4:24So water is both used directly and indirectly in the whole
- 4:27supply and creation chain of AI technologies. It is used for
- 4:32the refination of the critical raw materials that are needed
- 4:37to create the hardware of AI.
- 4:39Getting accurate figures on how much water it takes to build
- 4:42AI systems and run them is difficult. Google, Meta and
- 4:47Microsoft release annual figures showing that their data
- 4:50centres use billions of litres of water every year from
- 4:53local sources, but none of them indicate how much of it is
- 4:57due to AI. Most tech giants recognise the impact it's
- 5:00having. Many, including Google, Microsoft and Meta, have
- 5:04pledged to be water neutral by 2030.
- 5:10S5 We hope that can happen,
- 5:12there is a long way to go to get to those kind of numbers. Part of what
- 5:18we hope to see is, across the industry,
- 5:21a range of innovations that allow us to maybe minimise the use of water.
- 5:25Companies are trialing, for example, ways to cool data
- 5:28centres without evaporating any water at all. And to use the
- 5:32heat that's generated to warm homes. There are also
- 5:35experiments to move data centres away from communities
- 5:38entirely under the sea, to the Arctic, or even off the
- 5:42planet.
- 5:43Can we actually put capacity out in space? It's very, very
- 5:48early stage. So what, you know, we at NTT are looking at is,
- 5:52can we launch satellites that can at least do some more
- 5:56backup-oriented or other oriented tasks?
- 6:00Though skeptics point to the many hurdles that need to be
- 6:02overcome, there is optimism, too, about a more sustainable
- 6:06future.
- 6:07Let's remember that that GenAI capability is still very,
- 6:10very young. It's moved exponentially fast, but as an
- 6:14industry and as a use, it is still young. Ideally, we can
- 6:18learn together as as a society and as a world society, how
- 6:23do we minimise against the use of water and energy? Because
- 6:26this is all, you know, a world resource when we
- 6:29talk about water.
About this transcript
This page contains the full transcript of How AI uses our drinking water - BBC World Service by BBC World Service, generated from the public captions YouTube serves with the video. The transcript has 986 words across 101 segments, with the original timestamps preserved so you can click any line to jump to that moment in the embedded player.
What you can do with it
Use the transcript to take notes, quote the speaker, build a study guide, generate a summary with ChatGPT or Claude via the YouTube Summary tool, or export it as a timed subtitle file with YouTube to SRT. You can also re-open it in the transcriber to choose a caption track available on the video.
Free YouTube transcript tool
YouTube2Text is a free YouTube transcript generator — no signup, no daily limit. Paste any YouTube link and get the full transcript instantly, with timestamps, click-to-jump, caption language selection, AI prompts for ChatGPT, Claude, and Gemini, and exports to TXT, SRT, VTT, or Markdown.