AI’s carbon footprint and climate change effects
Jun. 26, 2025 • Tania Kukreja, LL.M. Student, Amity University, Mohali, Punjab
Student's Pen
Introduction:
In today's world, AI (Artificial Intelligence) is increasingly dominating various sectors with its groundbreaking technology. Its sophisticated software is designed to understand, think, and learn from its surroundings, producing highly relevant responses. These days, many articles, headlines, and other written content are often generated by AI. However, our focus isn't on whether AI will replace human jobs or if it contributes to a decline in human intelligence due to over-reliance on these tools. Instead, we aim to explore how the expanding carbon footprint of AI is impacting the environment.
What is a generative AI?
Before we understand generative AI, we shall know what is meant my Artificial Intelligence (AI). It is a broad range of computer system designed to simulate human intelligence by performing tasks that typically require human cognitive functions. It includes learning, reasoning, problem-solving, perception, language understanding and more.
Generative AI is a subset of AI, specifically focusing on creating new content, such as images, texts, and videos, based on its training data and user prompts. Generative AI works by utilizing generative models also known as “Large Language Models (LLMs)”, which learn the patterns and structures of their training data to produce new outputs that mirror this learned content. These models often rely on neural networks, such as transformers, which process vast amounts of data to understand relationships within it.
During training, they predict the next output in a sequence (like the next word in a sentence) and learn from the feedback—this process is repetitive, allowing the model to improve its predictions over time.
Some examples of foundation models are LLMs, GANs, VAEs, GPT-3 and Stable Diffusion, which allows users to leverage the power of language. Widely used applications such as ChatGPT, which utilizes GPT-3, enable users to create essays from brief text prompts. In contrast, Stable Diffusion lets users produce photorealistic images based on textual input.
Some popular AI generators are ChatGPT, DALL-E3, Microsoft Copilot, Google Gemini and many more.
Major factors contributing to environmental impacts:
While AI has emerged as a powerful tool due to the rapid advancement of new technologies, its environmental impact is concerning. According to the International Energy Agency (IEA), engaging with AI could use up to ten times more electricity than a typical Google search.
To train the AI models, there are data centers which are a temperature-controlled buildings that houses servers, data storage drives and network equipment. AI just computes but a generative AI but according to Noman Bashir, a generative AI training cluster could use seven or eight times more energy than a standard computing workload.
According to the Organization for Economic Co-operation and Development, global electricity consumption by data centers reached 460 terawatts in 2022, positioning them as the 11th largest electricity consumer in the world, just between Saudi Arabia (371 terawatts) and France (463 terawatts).
While training the Generative AI maximum amount of electricity is consumed which then leads to the generation of tons of carbon dioxide. In 2019, researchers from the University of Massachusetts Amherst discovered that training a large AI model can generate over 626,000 pounds of CO2 emissions, which is comparable to the lifetime emissions of five cars.
A more recent study indicated that training GPT-3, which has 175 billion parameters, utilized 1287 MWh of electricity and produced 502 metric tons of carbon emissions, equivalent to the annual emissions from driving 112 gasoline-powered cars.
The training also causes pressures on the power grid operators causing fluctuations, for this issue they usually employ diesel-based generators.
Each time an individual uses ChatGPT to summarize an article, the computing hardware that performs those functions consumes the energy. But the actual consumption of electricity by the AI models cannot be estimated as the companies are not providing the precise data on it.
With the electricity demands being the major problem, there is also a need for a huge amount of chilled water supply so as to cool down the hardware that are being used in training, deploying and fine-tuning generative AI models, in order to absorb heat from the computing equipment. This causes an impact on the biodiversity and can cause a strain on the municipal water supplies.
Another cause for environmental degradation is the production of high-performing chips, such as GPUs (Graphics Processing Units), which uses significant energy and water for its production, resulting in substantial carbon emissions and increased water consumption.
The extraction and processing of raw materials, such as silicon, rare earth elements, and other metals, also pose environmental challenges, including habitat destruction and pollution. With the AI industry growing at a fast pace, the chip industry is moving quickly leading to increase in e-waste.
The United Nations reports that global e-waste generation is increasing at a rate five times greater than the recorded levels of e-cycling. In 2022, a staggering 62 million tons of e-waste were generated, representing an 82% increase since 2010.
It is anticipated that this figure will rise by an additional 32% by 2030, reaching 82 million tons. Less than 25% of this e-waste was recycled appropriately, and only 1% of the demand for rare earth elements was fulfilled through e-waste recycling.
Sustainable Approach to Mitigate AIs Carbon Footprint:
Adopting sustainable approaches is essential to mitigate this growing carbon footprint and ensure that technological advancements do not come at the expense of our planet. Here, we explore various strategies that can be implemented to create a more environmentally friendly AI landscape.
1- Utilizing Renewable Energy Sources: Transition AI data centers and operations to renewable energy sources such as solar, wind, or hydroelectric power. This shift significantly reduces the carbon emissions associated with electricity usage in AI processes
2- Improving Data Center Energy Efficiency: Invest in energy-efficient data center infrastructure and cooling technologies to minimize energy consumption. Implementing best practices for data center design can drastically lower overall energy usage.
3- Leveraging Cloud Computing: Utilize cloud-based solutions that optimize resource use by distributing workloads across efficient, scalable data centers. Cloud providers often have advanced technologies to manage energy and cooling more effectively than individual organizations can achieve
4- Promoting Hardware Innovation: Encourage the development of next-generation AI hardware, such as neuromorphic chips and optical processors, designed specifically for energy efficiency, further decreasing the carbon footprint associated with AI computations
5- Recycling and E-Waste Management: Establish effective recycling programs for outdated AI hardware and electronic components to mitigate the environmental impact of e-waste. Recycling metals and other materials can help reduce the demand for new resource extraction
6- Facilitating Cross-Disciplinary Research: Foster collaboration between AI researchers, environmental scientists, and policymakers to identify innovative solutions and establish best practices for sustainable AI development and deployment.
There are numerous additional strategies that can be employed to reduce the carbon footprint of AI. The strategies mentioned above are just a few key examples.
Conclusion:
As generative AI continues to gain traction in our daily lives from writing articles and generating images to powering virtual assistants its environmental impact is becoming increasingly difficult to overlook. Behind the convenience and creativity lies an energy-intensive process involving massive data centers, high-performing chips, and immense water usage to cool overheated servers.
OpenAI CEO Sam Altman recently posted a light-hearted message on X, asking users to "chill out a bit," referencing the Ghibli-style avatars created by AI. While playful in tone, his message underscores a deeper truth: the pressure to meet rising AI demands is taking a toll—not just on developers and teams but on the planet itself. AI’s carbon footprint is rapidly expanding due to the power required to train and operate large language models. Additionally, the water used to cool systems and the electronic waste from hardware production contribute to a growing environmental crisis.
While AI offers immense potential to improve lives and reshape industries, its growth must be responsible and ethical. Striking a balance between innovation and sustainability is not just ideal—it’s necessary.
References:
4- Tech Target