show index hide index
Artificial intelligence, fascinating for its capabilities and potential, is also at the heart of heated debates regarding its energy impact. As millions of users turn to cutting-edge tools like ChatGPT, the crucial question of energy consumption and CO₂ emissions associated with these technologies is becoming increasingly pressing. However, a lack of clarity persists around the true carbon footprint of AI systems, casting a shadow over their long-term sustainability. Artificial intelligence (AI) is transforming our daily lives, but it also brings significant challenges, particularly regarding energy consumption. While spectacular advances are being made, details related to AI’s carbon footprint often remain unclear. This article explores recent claims regarding the energy used by artificial intelligence models, limited transparency in the sector, and the importance of clear regulation to address these environmental challenges. Energy Consumption: Disputed Figures Artificial intelligence gurus often maintain a certain opacity regarding the energy consumption of the models they develop. For example, Sam Altman, CEO of OpenAI, recently stated that an average query onChatGPT uses 0.34 watt-hours of electricity. This is equivalent to the consumption of an oven in just over a second or that of a low-energy light bulb in a few minutes. However, without further details provided by OpenAI, this figure raises many questions. What is meant by « average query »? Is image generation included? Are the energy costs of model training or server cooling taken into account? The lack of environmental transparency One of the main criticisms leveled by experts is the lack of transparency from AI companies regarding their carbon footprint. Sasha Luccioni, climate specialist at Hugging Face, highlights this opacity. A recent effort is to estimate the environmental impact of AI. A peer-reviewed study reveals that by May 2025, 84% of large language model (LLM) usage involved models that did not provide any environmental indicators. This lack of transparency is concerning in the context of a climate emergency. The Importance of Regulation and Precise Measurement The need for strict regulation and precise measurement is imperative. As Luccioni points out, the daily use of AI tools without clearly defined efficiency metrics or emission factors is worrisome, especially when compared to the transparency required in other industries, such as the automotive industry, where fuel consumption is measured per kilometer. In the absence of reliable data, it becomes easy for speculative assumptions to become established truths. One example is the claim that a query on ChatGPT
consumes ten times more energy than a Google search. This figure comes from a comment by John Hennessy, president of Alphabet, making it an unreliable source for such a claim.
Future Challenges for AI and the Environment
As artificial intelligence becomes increasingly ubiquitous, its energy implications must be understood and proactively addressed. Research and innovation must be accompanied by clear environmental responsibilities. Advances in AI promise tremendous potential, but developers and regulators must ensure that these technologies do not compromise the ecological future of our planet.
