S4E2 The Environmentalist Debate
AI, the planet, and how to argue both sides
Core reading & links
-
Core reading: The Bottleneck, by E.O. Wilson
-
Chat with Chipp about Environmentalism and AI
-
The Environmentalist Debate lesson
Word
Docs
-
Harkness discussion document
Word
Docs
The big idea: E.O. Wilson's two voices
In his essay The Bottleneck (the core reading above), E.O. Wilson says there are two types of people, two viewpoints, on how the whole "climate change" story will play out. Today's whole debate is really these two voices arguing. Read the essay, and listen for which voice each team is using.
Human nature is to design, to grow, to invent, to keep building machines, and essentially to keep polluting and carry on as "business as usual." Eventually those same inventions will find ways to balance the pollution and the food shortages, and human innovation will be our solution. Humans don't need to change: if we are digging a hole for ourselves, we will simply design a better ladder to climb out.
Humans must realise the Earth is, in fact, a finite source of energy, space and water, and as our population keeps soaring we must adapt and change to the reality of Earth's resources, or we will perish. No amount of engineering or ingenuity can supersede human growth, greed, and the general path towards an unsustainable future. Only by listening to the data from environmental scientists, and making societal changes on a global scale, can we begin to tilt things back towards a sustainable future.
First — what actually happens when you use AI?
Your message leaves your screen and zooms off to a giant building full of computers, called a data centre. There are thousands of them around the world.
Those computers do a huge amount of work to answer you. That uses electricity — far more than your phone or laptop uses by itself.
Most of the world's electricity is still made by burning gas and coal. Burning them puts carbon dioxide (CO₂) into the air.
CO₂ is invisible, but it works like a blanket around the Earth, trapping heat. More CO₂ means a hotter planet. That's what people mean by climate change.
And there's a second cost: water. The computers get so hot they would break, so data centres pour clean, fresh water through them to cool them down. In hot, dry places, that can mean using drinking water to cool computers.
Red team, blue team — and what "red-teaming" is
Today is not about winning, and it is not about picking the answer you already believe. You will be given a side in secret — RED, the economists, or BLUE, the environmentalists — and your job is to argue that side as well as you possibly can. You might be put on the side you don't agree with. That is the whole point.
This is called red-teaming: you argue a position properly, from the inside, so you really understand it and can see where it is strong and where it is weak. It is one of the best ways to think clearly about a hard question, because it stops you from only ever hearing your own side. When the debate is over, we drop the teams, and you say what you actually think.
Below is your kit. Every team has its best lines, the numbers behind them, and — just as important — why each number matters. A number on its own is not an argument. "1.5%" means nothing until someone tells you what it does.
Listen first — the big idea under both sides
Both teams are really arguing about one old idea. In 2002 the scientist E.O. Wilson said every argument about the planet has two voices: the economist and the environmentalist. Here is a short discussion of his essay The Bottleneck as a way in. One warning that fits this whole club: this podcast was made by an AI (NotebookLM) from Wilson's essay. It is a way in, not a source. Treat what it says as a claim to check, and verify anything that matters against Wilson's own words in the sources below.
The Bottleneck — two voices in every argument
You heard these two voices in the podcast above. The economist trusts clever ideas and new technology to get us through. The environmentalist says the Earth is finite and we are already taking more than it can give. Listen for them in the whole debate — "AI will solve climate change" is the economist talking, and "demand is growing faster than clean power" is the environmentalist.
Humans always find a way.
Every big problem we've faced, we've eventually invented our way past it.
Why this countsIt's the bet behind the whole RED case: science and technology will see us through the "bottleneck", as Wilson called it.
You can't outsmart a full planet.
To give everyone on Earth an American lifestyle would take four more Earths. Wilson, 2002
Why this bitesBeing clever doesn't change the simple maths. The planet only has so much, and we're already asking for more than it has.
Advanced HigherGo Deeper if you're Able
Our "caveman" brain. Wilson thought our minds were built long ago to notice danger right in front of us, and to ignore costs that are far away and shared between lots of people. AI's cost is exactly that kind — invisible and far off. That may be the real reason it never feels urgent.
Getting through the bottleneck. Wilson imagined humanity squeezing through a tight "bottleneck" into a future that lives within the planet's means. Is AI a tool that helps us get through — or just another way of borrowing from the future to pay for today?
How much electricity does AI really use?
AI uses only a tiny slice of the world's power.
All the world's data centres together use about 1 to 2 units of electricity out of every 100 — and they get more efficient every year. IEA, 2025
Why this countsThat's a small amount for something this useful. Each year, AI learns to do more with less power.
Worried about carbon? Talk to the cows first.
Farming animals makes about 14.5% of all the world's greenhouse gases — data centres make under 1%. And cows burp methane, a gas about 28 times worse than CO₂. FAO
Why this countsIf you really want to cut warming, AI is a tiny target next to a much bigger one.
One AI building can eat the power of a whole town.
One big AI data centre uses as much electricity in a year as 100,000 homes. The biggest new ones, 20 times that. IEA, 2025
Why this bitesThat's a small city's worth of power, day and night, for one building. Most of it comes from burning gas, which makes CO₂ — the gas heating the planet. And there are thousands of these buildings.
"Only a tiny bit" — but it's growing really fast.
That 1 to 2 units out of every 100 is set to more than double by 2030. IEA, 2025
Why this bitesThe problem isn't how much it uses today, it's how fast it's growing. At the same time, we're trying to move cars and heating onto electricity too. There's only so much clean power, and AI is grabbing more of it.
These data centres pollute like a whole country.
All the world's data centres give off about 180 million tonnes of CO₂ a year (2024) — about as much as a whole country. Just the new gas-powered AI ones could soon out-pollute a country like Morocco. Carbon Brief, 2024
Why this bitesThat's a country's worth of warming added on top of everyone else's, and it climbs every year.
If the world's data centres were a country…
The biggest CO₂ emitters, in millions of tonnes a year (2023). The longer the bar, the more it pollutes.
About 25 more countries fit between Germany and here, so the world's data centres would rank roughly 30th out of ~195 countries — bigger than more than 160 countries on their own, but about 70 times smaller than China. A small slice of a giant number.
Could AI save more than it costs?
Don't use less AI. Use it to fix the problem.
Scientists say AI could stop up to 5.4 billion tonnes of carbon pollution every year by 2035, by helping power, farming and transport work better. Stern et al., 2025
Why this countsThat could be more pollution saved than all the data centres add. Those three areas are nearly half of the world's pollution.
AI is already helping — this isn't just a promise.
AI worked out the shape of 200 million tiny proteins, helped find better batteries and solar panels, and made wind farms about 20% more useful. DeepMind
Why this countsThese already happened. Each one helps us make cleaner energy or waste less — small wins that add up.
"Could" is not "did".
The 5.4 billion tonnes is a guess about the year 2035 — and only if the world actually uses AI that way.
Why this bitesThe pollution AI makes is happening now. The savings are a maybe, years away, and only if everything goes right.
Make it cheaper, and we just use more.
When something gets easier or cheaper, people usually do far more of it, not less. (This is called the rebound effect.)
Why this bitesSo the savings can get eaten up by everyone simply using AI much more than before.
Is AI really drinking the world's water?
Across a whole country, it's a drop in the bucket.
All the data centres in the USA use about 1.7 billion litres of water a day — out of more than 1,200 billion the country uses. About 1 part in 1,000. Mytton, 2021
Why this countsFarms and factories use far, far more. Blaming AI's water can miss the much bigger users right next door.
So build them where the water is — like Scotland.
Newer cooling uses far less water, or air instead, and a centre in a wet, cool place barely touches the supply.
Why this countsThe problem isn't AI, it's putting it in a desert. Build it somewhere cold and rainy and the water worry mostly disappears — which is why Scotland and the Nordics want them.
Put the data centres in space — then no water at all.
Companies like Starcloud and Google are already testing AI computers in orbit. In space, solar panels make up to 8 times more power, and the heat just radiates away into the cold of space, so no water is needed to cool them. Scientific American, 2025
Why this countsIt could remove the water problem completely. The catch: it takes huge amounts of energy and rocket fuel to launch them up there.
One burger out-drinks a whole year of your AI.
Making 1 kg of beef uses about 15,000 litres of water — around 100 bathtubs. That's far more than a whole year of your AI questions. Water Footprint Network
Why this countsIf water is the worry, beef is a much bigger straw in the cup than AI.
AI is thirsty — and it drinks the clean stuff.
An AI uses about a 500 ml bottle of water for every 10 to 50 questions you ask. Training one big AI used about 700,000 litres of fresh water. Li et al., 2025
Why this bitesThe water cools the hot computers, then floats away as steam. It's usually the same clean water we drink.
They build them in exactly the wrong places.
About 2 out of every 3 data centres built since 2022 are in places already short of water — Arizona, Texas, Spain, Chile. Bloomberg, 2025
Why this bitesThere, the data centre fights farms and families for water that's running out. Google had to stop a data centre in Chile because the city couldn't spare the water.
"AI's killing the planet — why are you even in AI Club?"
Someone will say this to you. The honest answer is that it's complicated, and clever people don't agree. Both sides have real facts — and now you have them, so you can think for yourself and explain it to anyone who asks.
You could say: "AI does use a lot of power and water, and that's a real problem. But it's also helping invent cleaner energy. The question is whether the help is bigger than the harm — and that's exactly what we argue about in AI Club."
The whole debate, at a glance
Where these facts come from
Every figure on this page was checked against the real source before it went here. AI tools (this club's bots, and the podcast above, included) can invent convincing-looking facts and citations. Click-check anything before you rely on it — "I verified the source myself" is exactly the habit this club is about.
Carbon & energy
- CoreIEA (2025). Energy and AI. The 100,000-homes comparison, the ~1.5%-of-world-electricity figure, and the 2030 projection. See the executive summary.
- CounterStern, Romani, Pierfederici et al. (2025). Green and Intelligent. npj Climate Action. Where the "5.4 billion tonnes saved" figure comes from.
Water
- CoreLi, Yang, Islam & Ren (2025). Making AI Less "Thirsty". Communications of the ACM. The 500 ml bottle and the 700,000 litres of training water.
- JournalMytton, D. (2021). Data centre water consumption. npj Clean Water 4, 11. The "1 in 1,000" national figure.
- WhereBloomberg (2025). The AI Boom Is Draining Water From the Areas That Need It Most. Why two-thirds of new data centres sit in water-stressed places.
The big idea
- FrameWilson, E.O. (2002). The Bottleneck, Scientific American (from The Future of Life). The economist-versus-environmentalist voices, the "four more Earths" idea, and our "caveman" brain.
Go deeper
- ExplainerClimate Change News (2026). Will AI data centres make or break the energy transition? Plain-language background on both sides.
- DataPew Research (2025). What we know about energy use at US data centers. What is actually measured, and what is not.
- OptimistGoogle DeepMind. Using AI to fight climate change. The defender's own examples — read them as claims, not settled fact.
Key words
- Data centre
- A giant building full of computers that does the work when you use AI or the internet.
- Carbon dioxide (CO₂)
- An invisible gas made by burning fuel. It traps heat and warms the planet.
- Climate change
- The planet slowly heating up because there is more CO₂ in the air.
- Water stress
- When a place doesn't have enough fresh water for everyone who needs it.
- Economist (in this debate)
- The voice that trusts clever ideas and technology to fix problems as we go.
- Environmentalist (in this debate)
- The voice that says the Earth is finite and we must not take more than it can give.
- The bottleneck (Wilson)
- Wilson's name for this century's tight squeeze, when human demand presses hardest on a finite planet and we have to get through without wrecking it.
- Red-teaming
- Arguing a side as well as you can — even one you don't believe — to really understand it.
- Rebound effect
- When something becomes cheaper or easier, people use much more of it, which can cancel out the saving.