A simple question to an AI generates a monumental impact. The ecological footprint of tools like ChatGPT and DeepSeek often remains unknown. Evaluating their energy consumption is equivalent to measuring the brightness of a light left on for several hours.
The environmental issues related to the use of artificial intelligences require rigorous analysis. Understanding the energy consumed by each request is imperative. The recent summit in Paris on AI highlights this dilemma.
The complexity of the calculations required to process a simple input raises crucial questions. The resources used, such as water and electricity, are astronomical. This interview reveals unprecedented perspectives on our relationship with these technologies.
Environmental impact of artificial intelligence queries
As the AI summit approaches in Paris on February 10 and 11, the question of energy and resource consumption generated by artificial intelligence tools, such as ChatGPT and DeepSeek, becomes crucial. Researchers are closely examining the quantification of this impact, particularly to determine how much water and electricity are actually used during a simple query.
An analysis of the energy consumed
Generative AI tools, capable of quickly writing texts or creating images, hide behind their operation a non-negligible energy complexity. Experts are asking: Do ChatGPT or DeepSeek turn out to be energy hogs? Each question posed can be equivalent to leaving a bulb on for several hours. A single request can consume a surprising amount of energy. The International Energy Agency (IEA) has determined that each inquiry on ChatGPT could require at least ten times more electricity than a standard Google search.
Responses from AI creators
The teams responsible for ChatGPT and DeepSeek have been questioned about their environmental impact. They were vague in their responses. ChatGPT stated: “It is difficult to precisely quantify my carbon footprint”. For their part, the developers of DeepSeek claimed that “a simple query consumes very little energy”, which seems contradictory to the data analyzed by various independent studies.
Revealing figures
A study conducted by MIT Technology Review disclosed that training the GPT-3 model produced 552 tons of CO₂, a quantity equivalent to the CO₂ emissions of 205 round trips between Paris and New York. The impact is not limited to the initial training. Each question posed activates data centers that operate continuously, consuming water and electricity.
Assessment of daily use
Amélie Cordier, a doctor in AI at the University of Lyon, highlighted that tools like ComparAI allow for measuring the environmental impact of an AI query. This becomes tangible when considering that a query can be equivalent to the lighting of an LED bulb for several hours.
Towards an eco-score for AI?
The need for an eco-score for artificial intelligence tools is becoming pressing. This would raise public awareness of the consequences of their requests. Indeed, the user often has no idea of the billions of calculations performed by processors in vast data centers.
Responsible and concise use
DeepSeek recommends to “favor short and targeted interactions” to minimize environmental impact. Each additional request increases the complexity of the necessary calculations, equivalent to a balloon that grows larger with each interaction.
Impact on the ecological transition
The applications of artificial intelligence are not limited to recreational processes. AI optimizes ecological solutions, such as climate forecasting and the strategic placement of solar panels. However, improper use of AI tools for trivial queries can counterbalance these beneficial contributions.
Experts’ conclusion
The development of AIs and their ecological impact require thorough reflection on their use. The benefits of ethical and targeted usage can far outweigh the disadvantages related to their consumption. The question of whether these tools truly help counter climate change deserves to be asked seriously, taking into account all dimensions of their operation.
Frequently asked questions
What is the carbon footprint of a single query to ChatGPT or DeepSeek?
A query to these artificial intelligence models can emit a quantity of CO₂ equivalent to several hours of use of an LED lamp, thus having a more significant environmental impact than one might imagine.
How much energy does a question posed to an AI like ChatGPT or DeepSeek consume?
Asking a question can consume the equivalent of the energy needed to run an LED bulb for an hour, and sometimes even more, depending on the complexity of the request.
Can we evaluate the precise environmental impact of interacting with these AIs?
There are tools like ComparAI that allow measuring the energy consumption of a query submitted to an AI, but results can vary depending on technical specifics and resources used.
How do data centers impact our environment when using AI?
Data centers, required for AI operation, consume enormous amounts of electricity and water for cooling, contributing to the increase of the carbon footprint associated with their use.
What does a “rebound effect” mean in the context of AI use?
A “rebound effect” occurs when the reduction in the costs of using an AI leads to an increase in its use, which could negate the initial environmental benefits of that reduction.
Can conversational AIs be used more responsibly to reduce their impact?
Yes, it is recommended to formulate precise and concise requests to minimize the necessary calculations and, consequently, the energy consumed during interactions with these systems.
What are the alternatives to using AI for information retrieval?
Before resorting to AI, it is wise to consider other search methods, such as books, specialized articles, or in-person discussions, to reduce our dependency on energy-intensive systems.
What role does AI play in the ecological transition?
AI can be beneficial for the ecological transition by optimizing processes such as resource management, positioning of solar panels, and climate projections, but it depends on responsible use.
Are users aware of the environmental impact of their daily use of AIs?
Often, users are not fully aware of the environmental impact that their daily use of AIs can generate, highlighting the need for increased awareness on this issue.
How can we raise users’ awareness of the impact of AIs on the environment?
Awareness campaigns, eco-scores for AI models, and studies on their energy consumption can help raise public awareness and encourage more responsible practices in using these tools.
