AI’s Thirst for Water
Artificial intelligence systems like ChatGPT don’t just run on electricity, they also consume water. This happens at the data centre level, where servers are cooled using vast amounts of water to prevent overheating during complex computations. Studies show that generating a simple AI response can use enough water to fill a glass, and training large models like GPT-3 consumed over 700,000 litres of fresh water. Every prompt, no matter how small, adds to this hidden footprint.
When “Please” and “Thank You” Cost Millions
This quiet environmental cost became public when OpenAI CEO Sam Altman jokingly admitted that users typing “please” and “thank you” into ChatGPT was costing the company “tens of millions of dollars well spent — you never know.” While said in jest, the comment highlights a real issue: even polite words require the system to process more tokens, using more energy and water in the background. Multiply that by millions of users, and the environmental cost becomes significant.
Who Bears the Responsibility?
One school of thought argues that users themselves should be mindful of their digital footprint. Just as individuals are encouraged to reduce plastic use or turn off lights to conserve energy, they could (in theory) simplify their AI interactions to minimize excess computational demand.
Yet others counter that this burden should not fall on individuals. After all, AI models are designed to mimic natural human conversation. If anyone should fix the problem, it’s the companies that create and maintain these systems, by investing in greener data centres and more efficient model design.
Why It Matters for Africa
For regions like Africa, where water scarcity and power limitations are already development challenges, the environmental cost of AI adoption cannot be ignored. Without careful planning, countries risk importing technologies that silently strain their resources.
The Takeaway
This raises an important question for all of us engaging with AI technologies: Should the burden of reducing AI’s environmental footprint rest with the individual user; or does the greater responsibility lie with the companies and developers shaping these systems at scale?
As AI becomes a more integral part of life and business, this debate will only grow more relevant.
Leave a reply