Taking Stock of AI 

We need to pause the AI hype train to ask ourselves, not ChatGPT, some questions. 

If you’re familiar with the torment of writer’s block, then the prospect of turning to AI chat tools can be tempting. Yet many writers still avoid ChatGPT and its ilk, mostly to avoid their telltale AI tone but also out of fear of plagiarism, as well as because it simply feels like cheating. Writer’s block is part of the creative process, as frustrating as it may be in the face of a looming deadline.  

Some of us will admit to using AI tools for more menial tasks. This includes asking it to summarise sprawling notes, boil down dense reports, conduct spellchecks, delete repeated phrases, and delete repeated phrases. These functions buy back a bit of time in the workday, providing we resist the urge to procrastinate with questions like “how many apples would I need to create an apple tower tall enough to reach the moon?” But they have little to no wider societal value. So, should AI be relied on in this way? It’s 5.1 billion apples, by the way.  

In early June, a session about AI sustainability took place at the European Sustainability Week (EUSEW) in Brussels. There, a representative of the International Energy Agency showed a slide detailing how asking an AI chatbot to produce a large chunk of text uses as much energy as it would take to power a 600 lumen LED light bulb for an hour. Generating an 8-second AI cat video consumes nearly twice as much as is needed to charge a laptop, he added. If we consider that ChatGPT alone has 800 million weekly users, then we’re talking a lot of energy, and a lot of cats. 

This prompted reflection: If AI chat tools included a live emissions tracker, would we rely on them for as many tasks? What is our AI carbon footprint? How many rooms could we illuminate with these our not-so-illuminating questions? 

Beneath the surface of our daily interactions with AI lies a sprawling material infrastructure embodied by data centres. As of 2025, data centres accounted for roughly 2% of the European Union’s energy consumption, a share that is expected to triple by the end of the decade, driven by AI. By 2030, the global electricity demand from data centres could eclipse that of Japan as AI technology drums up a forecasted $15.7 trillion for the global economy.  

The demand is insatiable, but our resources are finite. In Europe, renewable energy could provide around 80% of the electricity needed to sustain data centres by 2035, according to the IEA, with nuclear, natural gas, and coal filling the gap. In the US and China, the two big players in AI, projected renewables for data centres drops to around a third with fossil fuels propping up much of the rest.  

Data centres are thirsty and need a flow of water to keep their systems cool. AI’s global water consumption is predicted to exceed that of Denmark’s by 2027.   This is a particularly contentious issue in water-scarce countries like Spain, which is being touted as a digital hub as Amazon and Meta scramble to open up hyperscale data centres, prompting environmental concerns.  

Of course, our use of AI extends far beyond uncanny social media videos and humdrum tasks. AI tools have already offered breakthroughs in medical diagnostics, agricultural innovation, early warning climate systems, transport, and energy efficiency, as well as less noble pursuits like warfare and as a possible replacement to human workers, which would be nice only if we get to keep our salaries and retire early. 

In the corridors of Brussels, AI is touted as a tool that can power the “clean energy transition,” although the details of how are scant. In an ideal world, renewable energy growth would keep up with data centre demand, big corporations would be bound by strict sustainability rules and reporting requirements, and AI would only be used for the betterment of society.  Unfortunately, we don’t live in an ideal world.  

As of 2025, AI development, including the physical infrastructure underpinning it, is in the hands of profit-driven multinationals whose innovation is outpacing policymaking. And in the fog of a perpetual omnicrisis it can be hard to take stock of our aims and ask: Who is this for? What is this for? And at what cost? 

Speaking at the EUSEW, Anastasia Tsougka, a program manager at the Environmental Coalition on Standards, put it well: “When we’re in the process of continuous growth, it’s hard to stop and see we should be aiming for something less.” 

Sensible and sustainable AI use will only be brought about with bottom-up pressure on Big Tech and politicians, which in turn requires citizens and AI users to be aware of the environmental costs of the technology.  At a time when the EU and the US are rolling back climate commitments whilst also trying to remain competitive with China, this will be a challenge.  

Calling for frugal and planet-centred AI models, which prioritise efficiency, quality, and sustainability, would be a step in the right direction. But bringing them about requires a wholesale rethink of the AI industry’s present trajectory, which is heady and sometimes feels beyond our control – when it very much is not.  

Politicians will have to be bold, informed, and transparent when it comes to drawing the regulatory lines around Big Tech’s rapid AI expansion, which should be counterbalanced with strict environmental reporting that carry sanctions if not adhered to. The alternative is to hand unchecked power to data-hungry multinationals, whose motivations are rarely altruistic.  

Hey ChatGPT, think of a catchy conclusion.