Fast Tech, IA and Energy Waste
- 11 hours ago
- 3 min read
IA and Energy Efficiency : A Global Challenge
1. AI, an energy-intensive revolution
Artificial intelligence (AI) has become a cornerstone of innovation, but its rise is accompanied by a sharp increase in energy consumption and CO₂ emissions. According to the International Energy Agency, electricity demand related to AI could grow by 15 to 20% by 2026, while data centers could double their electricity consumption compared to 2022 (source-aivancity ). By 2025, the intensive use of AI has already generated as much CO₂ emissions as the city of New York, or about 80 million tons. (source-cnews). Producing a 100-word email with GPT-4 would consume approximately half a liter of water, depending on where the model is hosted; generating a single image using AI would consume as much energy as fully charging an average smartphone; and submitting a query to generative AI could consume up to 30 times more energy than using a conventional search engine. (Source Commission de l'éthique en science et technologie du Québec).
Training large language models, such as GPT-3, consumed up to 1,287 MWh per run, emitting the equivalent of 552 tonnes of CO₂, or the annual emissions of 123 gasoline cars (source-jeconsommeresponable) . Inference (the daily use of models) became even more resource-intensive than training once a model exceeded 200 to 600 million uses, a threshold quickly surpassed by the most popular generative AIs. (source-jedha)
2. The impact of the energy-intensive revolution on the climate and resources
Beyond CO₂, AI puts a strain on natural resources: data centers could soon consume six times more water than Denmark, and the extraction of rare metals needed to manufacture electronic chips exacerbates the pressure on ecosystems. (notre-environnement.gouv.fr). According to Green IT, 69% of the environmental impacts of AI are not related to carbon, but to resource depletion and pollution (bpifrance.fr).
3. Corporate Responsibility: Balancing Innovation and Sobriety
Faced with these challenges, businesses are on the front lines. In Europe, 64% of large companies now integrate the environmental footprint of AI into their investment decisions, favoring resource sharing and balancing performance with sobriety (aivancity.ai). Some, like Google and Microsoft, have seen their emissions rise by 29 to 48% since 2019 due to AI infrastructure development, calling into question their carbon neutrality (commitmentsinstitut-superieur-environnement.com).
However, solutions are emerging: algorithm optimization, eco-design of services, transition to renewable energy, and greater transparency about emissions. The European regulatory framework (AI Act) is beginning to address these issues, but remains insufficient to enforce true sobriety (code-climat.com)
4. Risks to the Planet and Pathways for Action
If no action is taken, the unchecked growth of AI could worsen pressures on water, electricity, and metals, while deepening its carbon footprint. Experts call for collective awareness: developing AI "with sobriety" is no longer optional but a condition for survival (newsentreprises.com)
Companies must:
Systematically assess the real utility of each AI application.
Prioritize more efficient models and shared infrastructure.
Integrate digital sobriety into their CSR strategies and train their teams (institut-superieur-environnement.com)
Conclusion : AI offers immense opportunities, but its development cannot come at the expense of planetary boundaries. Corporate responsibility is crucial to reconciling innovation with environmental preservation, adopting a sober, transparent, and sustainable approach.
And then, at a time when desalination plants are being destroyed in the Gulf countries because of the war in Iran, perhaps we should think more carefully before using AI to generate a beautiful image or ask a question that search engines know very well how to answer.
Interested in this topic? Contact us to schedule an appointment with our Energy Efficiency Manager..
Comments