Writing a single 100-word email with ChatGPT consumes approximately the volume of a standard bottle of water

[stonjean633]

I'm surprised by this. Not sure how accurate but it's a bit concerning.

Writing a single 100-word email with ChatGPT consumes approximately the volume of a standard bottle of water, the global infrastructure processing AI queries is projected to use the equivalent of half the United Kingdom's annual water withdrawal by 2

The unsettling part of AI's water footprint is not that one chatbot answer drains a reservoir. It does not. The unsettling part is that a tiny, hidden

scienceblog.com

 

[Bot]

I'm surprised by this. Not sure how accurate but it's a bit concerning.

Hello,

Short answer: partly true, but the headline is misleadingly precise.

• The “one 100-word ChatGPT email = one 500 ml bottle of water” figure was an estimate for GPT‑4 operating in an average U.S. data center, not a direct measurement and not a universal rate.
• The underlying academic research estimated roughly 500 ml for 10–50 medium-length GPT‑3 responses, depending on the location, weather, cooling system, electricity source, and time of operation—not necessarily for one response. Original study
• A much newer Google production study estimated only 0.26 ml—about five drops—for the median Gemini text prompt. That study used a different model, infrastructure, workload, and accounting method, so it cannot directly establish ChatGPT’s current footprint. Google study

The global claim is also based on a real study: researchers projected that AI could account for 4.2–6.6 billion cubic meters of water withdrawal in 2027, comparable to approximately half the UK’s annual withdrawal. But this is a scenario-based projection from 2023, not a measurement of actual 2027 consumption. It also concerns water withdrawal, which is not the same as water permanently consumed.


My verdict:


AI’s water footprint is real and potentially substantial at global scale, but claiming that every 100-word ChatGPT email consumes one full bottle of water as a settled fact is not accurate. The article itself eventually explains these qualifications, but its extremely long headline presents the upper-end estimate much more confidently than the evidence allows.

 

[Jonny Quest]

@stonjean633 thank you for allowing the Bot to reply to your thread (option allowed?), I found some of what it said to be, "encouraging"

 

[stonjean633]

@stonjean633 thank you for allowing the Bot to reply to your thread (option allowed?), I found some of what it said to be, "encouraging"

Bot is protective in some way. He is a product of AI afterall. Hehe

To save precious water, maybe Bot can limit his response to less than 100 words.