A cartoon shows two individuals facing a computer surrounded by hundreds of servers, with ‘AI’ written on the screen. One of the characters says: ‘And how are we going to find enough electricity for all this?’ The other replies, staring at the screen: ‘All we have to do is ask.’ This humorous illustration by the cartoonist Kak accompanies an article published on 9 April by L’Opinion, whose headline sums up what is at stake: ‘Hello progress? This is Earth. Too fast, too far, too hard: technological advances consume a lot of electricity, rare metals and land… without any thought or planning. It’s not too late to put things right.’ This materiality of AI and its environmental impact is emphasised by the main parties involved themselves. Sam Altman, founder of OpenAI, said at the recent Davos Summit1 that ‘a major breakthrough in energy is needed’ because ‘it’s absolutely right to say that AI is going to need a lot more energy.’ Bill Gates concurred, mentioning a ‘dizzying amount of energy required’, with electricity being ‘a key variable in the profitability of a data centre.’ In total, AI could consume 85 to 134 TWh of electricity in 2027, equivalent to the consumption of Argentina or Sweden.2 This raises serious questions about the compatibility of the development of AI with the energy transition. In terms of opportunity, this issue is at the heart of one of the recommendations in the report ‘AI: Our ambition for France’ submitted by the Artificial Intelligence Commission to Emmanuel Macron in March Given that France is lagging far behind in the development of AI but is a leader in the environment and low-carbon energy, with world leading companies like Suez, Veolia, Saur, EDF, Engie, Sonepar, Rexel, Schneider and Legrand, the report recommends adopting a specialisation strategy. ‘To think that the only option (when it comes to AI) is to copy the American model is absurd.’ On the contrary, we need to ‘make France a pioneer in AI for the planet by strengthening transparency, research into low-impact models, and the use of AI to support energy and environmental transitions.’3 On this last point, the solution lies – as is often the case – in the question of use and putting all technology back in its place as a simple tool for a specific purpose… which must be defined collectively and owned politically. Will AI, an energy guzzler, make it possible to reduce and rationalise global energy consumption? As Kak’s cartoon suggests, if you ask it a question, AI will answer. All that remains is to ask the right question.