Here’s another reason Chat-GPT is fucked: how AI is driving climate change

Column by: Climate Clinic VUW

Was your last essay a work of Chat-GPT? Or break up text for that matter? Generative AI is inescapable, and it’s also (and arguably more effectively) generating something else you might not have thought of: climate change. 

AI, like the internet generally, doesn’t exist in the ‘cloud’ as we might imagine. It operates out of data centres: massive facilities filled with turbocharged computers which are very real, very tangible, and very hungry for energy and water. 

Because it involves far more complex computing, simple AI queries typically consume 10 times the energy of regular Google searches. Image generation is even worse, about half an iPhone charge per image. Fresh water is also crucial for the cooling systems of these data centres. A ‘short’ conversation with Chat-GPT evaporates half a litre of water, quickly adding up to billions in a world already reckoning with impending water shortages.

Data centres are proliferating so quickly to meet the demand for computing power that they are expected to consume enough energy to sit at 5th place of global energy consumption by country by next year. That means data centres will be using ever so slightly less energy than Japan in its entirely. Or, put another way, 43 times as much energy as little Aotearoa. And half of all this is just AI. 

You mightily naively assume, coming from the land of renewables, that this would at least be clean energy. Wrong again. The majority of data centres are in the US, where fossil fuels still dominate at 80% of generation (with nuclear being another 8%). So AI is powered by dirty, high-emitting energy sources. 

And this is all going to get much worse. It’s estimated that the computing power required for AI is doubling every 100 days, meaning in the next five years it will need to increase by a million times. 

Despite all this, AI might have some uses in mitigating and adapting to climate change. It has proven its worth in designing more efficient buildings and cities, and predicting weather for more sustainable agriculture. But these positives hardly outweigh the negatives, and do nothing to address the underlying issue: boundless growth inevitably ravages limited resources. 

I’m not preaching to avoid AI like the plague—as it develops the reality is we pretty much have to use it or be left behind. But like all tools, use it wisely, and be mindful of the cost of convenience. 

As the founder of Open AI (Chat-GPT’s parent company) said, “AI will probably lead to the end of the world … but in the meantime, there’ll be great companies.” This end just probably won’t be a dramatic revolution of computers turning on humans: it will be their contribution to the very real and current threat of climate change.