Catherine Arrow
Sep 1

An AI Glass Half Empty - Managing Hyperscale Harm

When we talk about ethics in AI, the conversation too often stalls at bias, transparency, or governance codes. These are important, but they are only fragments of the whole picture. Ethics in AI is broad, deep and continuous. It is not a sideline. It is the very heart of practice.

Consider this. Every question asked of ChatGPT uses a nip of electricity and a sip of water. Alone, trivial. Together, immense. At 2.5 billion questions a day, the electricity demand equals that of 30,000 households and the water use of thousands of families. In Ireland, data centres already consume more electricity than all urban homes combined. In Virginia, water taken for cooling has risen by two-thirds in just four years. In Memphis, residents now live beside an AI data centre built in under three weeks, its 35 gas turbines altering local air and water balance and, it seems, the health and wellbeing of local citizens.

This is not simply infrastructure. It is impact. It is what I call data draining. The drain falls not on platforms but on people – on families competing for water, on communities carrying health costs from pollution, on localities where the lights may flicker because a hyperscale has arrived.

As practitioners, we must provoke - prompt even - this conversation with our leaders and colleagues. We must constantly remind organisations that ethics is not a single thread but a fabric that stretches across environment, community and governance. It is not enough to say “our model is fair.” We must ask where is it housed, what resources does it consume and who bears the societal cost?

Right now, out in the field, we are being bamboozled by the major providers. Most fudge or quietly sidestep questions about the resources required for AI to function. To cut through the noise, I have developed the AI Impact Tool below. The tool gives you a clear sense of how the nip of power and the drop of water used by a single prompt magnify at scale. Remember, OpenAI itself admitted in mid-2025 that ChatGPT handles around 2.5 billion prompts every day – that’s 1.7 million every minute. Stop and think about that. What looks like a tiny drop from one prompt becomes a flood. The planet is already drowning in data.

Environmental impact and societal consequences are not matters of optional corporate responsibility. It is the heart of governance and our licence to operate. The ethics of AI must now be understood as a continuous discipline, embedded in every stage of use, from design to deployment to reporting. And it falls to practitioners to make this plain - ethics in AI is not an appendix. It is a foundation.

My calculator below gives you space to explore the simple question: What’s Your AI Footprint? It allows people to see their AI use in practical terms: electricity, water and everyday equivalents – kettles boiled, homes powered, showers taken. It is a way to make the invisible visible and to decide whether our use of AI is responsible or whether we are adding to the drain.

So I invite you to try it. Run your daily use through the tool - probably a drop in the ocean for you - but then consider the prompts and use across all aspects of your organisation. Add the numbers and see what comes out. Do those numbers represent the sum of your values? Share the results with your teams. Start the conversation in your organisation. Because ethics in AI is not a line in a policy – it is the practice of making choices with eyes wide open, recognising the real cost of convenience and deciding whether we soak the Earth in data for short-term gain.

For the record: working on the coding, calculator and verification with AI support from ChatGPT and Manus, my 45-query session had a footprint of about 0.0135 kWh of electricity and 3–27 millilitres of water. That’s the same as keeping a 5-watt LED on for a couple of hours and less than a teaspoon of water. A nip of power, a sip of water. Almost nothing alone — but scaled to billions of queries a day, it becomes the daily demand of a small city.


References:

Axios, “ChatGPT receives 2.5 billion prompts a day” (21 July 2025).
https://www.axios.com/newsletters/axios-am-d51f7833-bc42-4dd0-bce6-352302479043
The Verge, “OpenAI confirms ChatGPT now handles 2.5B prompts daily” – Rob Friedlander confirmation (21 July 2025).
https://www.theverge.com/news/710867/openai-chatgpt-daily-prompts-2-billion
Associated Press, “Ireland’s data centres now use more electricity than urban households” (7 February 2025).
https://apnews.com/article/6c0d63cbda3df740cd9bf2829ad62058
Financial Times, “US tech groups’ water consumption soars in ‘data centre alley’” (3 February 2025).
https://www.ft.com/content/1d468bd2-6712-4cdd-ac71-21e0ace2d048
Time, “Elon Musk’s New AI Data Center Raises Alarms Over Pollution” (15 July 2025).
https://time.com/7021709/elon-musk-xai-grok-memphis/
What's your AI impact? - PR Knowledge Hub
PR Knowledge Hub - Learning to Break Boundaries

What's your AI impact?

Check out what's trickling down the digital drain every time you engage with AI

Learn more at learn.prknowledgehub.com

Understanding Your Digital Environmental Impact

We know that AI chews through electricity and chugs down water to keep things cool. This dashboard will allow you to see how your use of AI through the massive data centres relates to other uses of our planet's resources.

Remember - it may only be half a kettle boiled at home but major platforms such as ChatGPT deal with 2.5 billion prompts per day as disclosed by OpenAI in mid-2025 - so it all adds up.

Enter your digital activities below to see their environmental impact in everyday terms.

Your Digital Activities

Your Environmental Impact

0.00
kWh Energy Used
💧
0.00
Litres Water Used
🏠
0.00 min
Home Powered
🏘️
0.00 min
High-Use Home Powered
💡
0.00
LED Bulbs (8 hours)
0.00
Kettles Boiled
🚗
0.00
Kilometers Driving
🌾
0.00
Postage Stamps of Land
🛁
0.00 mL
Water Used
🚿
0.00 mL
Water Used
🥤
0.00
Glasses of Water