I use AI every day. I understand it reasonably enough — and perhaps because of that, the environmental impact genuinely scares me.
This isn’t abstract concern. A U.S. congresswoman recently flagged AI data centres contaminating water quality in her state. Not a projection. Not a future risk. Happening now. And it happened fast — faster than most people realise.
Beyond water, electricity bills for ordinary households in the U.S. have risen because consumers are quietly subsidising data centre grid costs. The people paying are not the people benefiting.
AI may feel inevitable. I’m not arguing that. But how we use it — and how society governs that use — deserves a much larger conversation than it’s getting.
Data centres already account for roughly 1–2% of global electricity consumption, a figure expected to double by 2030 according to the International Energy Agency. Training a single large AI model can emit as much carbon as five cars over their entire lifetimes. Water consumption for cooling is in the billions of litres annually — drawn, in some cases, from the same sources supplying local communities.
Most of us have no idea how much our daily AI usage contributes to this. Not because we don’t care — but because nobody has made it simple enough to act on.
(Detailed stats at the end)
I’m putting together a report. Not a theoretical document — a short, actionable checklist built around the things we all do every day.
PowerPoints
Proposals
Emails
Image and video generation
Thinking, analysis works.
Which tasks consume more AI credits. Which impact the environment more. How to optimise — which also saves you money by consuming less credits but if you’re using a free plan, make your credits last longer
The work still gets done. It takes the same amount of time, sometimes less. But understanding how AI actually works, even at a basic level, means you can make better choices without sacrificing output.
Something people can actually use. Not read once and forget.
I’m looking for sponsors at $400–$500 each.
What’s in it for you: being part of something relevant, useful, and ahead of where this conversation is going. The report will be no more than a dozen pages, in a checklist format, built to be shared. Your brand gets visibility in something people will actually pass around — not a PDF that sits in a downloads folder.
As AI’s environmental cost becomes more mainstream — and it will — this is the kind of association that ages well.
Comment below or message me here if interested.
Saurabh Parmar, Fractional CMO linkedin.com/in/saurabhparmar
ps- more detailed stats for those interested:
Data centres are estimated to consume roughly 1.5% of global electricity today (about 415 TWh per year), which fits the “around 1–2%” range you quoted.
The International Energy Agency projects that data‑centre electricity use will grow to about 945 TWh by 2030—roughly double the 2024 level—so the “expected to double by 2030” line is consistent with IEA analysis.
A widely cited 2019 study from the University of Massachusetts Amherst found that training certain large‑scale NLP models can emit more than 626,000 pounds of CO₂ equivalent, which is roughly comparable to the lifetime emissions of five average cars (including manufacture).
This figure is an upper‑bound, model‑specific example, not a strict rule for every AI model; emissions vary strongly with model size, hardware, electricity grid mix, and how many times the model is trained.
Recent analyses estimate that U.S. data centres alone used on the order of 17 billion gallons (about 64 billion litres) of water directly for cooling in 2023, with projections that this could double or even quadruple by around 2028.
Some large facilities draw water from local supplies that also serve communities, and in certain regions, data‑centre withdrawals can represent a significant share of local water availability
Comments
Nothing yet. Say the first thing.
Sign in to join the conversation.