In a modern tech-driven world, AI is the central interest of people and institutions from different layers and so is it for studies accentuating its rate of carbon emissions, especially with the latest increasing worries about carbon footprint which is proliferating day after another. AI’s effects have been often highly controversial, but the topic has been marked with a lack of sufficient data and studies.
Nature Magazine, the UN, the electronics giant Nvidia, and other concerned institutions have conducted studies whose focus was on how AI could mitigate the risks of carbon footprint looming in the landscape. Their studies emphasised that AI-driven projects can prevent potential upcoming climate risks and hence use convenient optimisations to run the involved forecasts.
Logistic companies will be often able to save fuel consumption through AI’s ability to draw shorter routes for the delivery addresses; meanwhile, farmers will be able to use fewer fertilisers and water with more efficiency, thanks to AI detecting the exact areas to be fertilised and watered, it will also treat all kinds of leaks from city pipelines networks, the studies illustrate. Yet, how much energy in turn is produced by AI when run on the former sectors?
Much Ado about Nothing?
AI, particularly generative is based on training and learning. Those require more data centres that house servers that are subject to expansion, meeting the increasing upcoming volume of updated information and the ones provided by users. Accordingly, more electricity is being consumed by servers as well as the air conditioners responsible for keeping them at a cool temperature and preventing them from overheating.
Large Language Models training (LLM) alone as a part of chat GPT-3 consumes an amount of energy equal to 500mt of carbon which is like blazing coal for continuous 10 hours, a late study by Cornell University indicates.
Big Tech’s Carbon Emissions Increase as Their AI Features Do
During its environmental report in 2024, Google announced that its emissions surged by almost 50% compared to 2019 which hinders its 2030 zero emissions plan. In 2023, data centres’ electricity consumption increased to 17%, marking a surge of 13% in terms of emissions which represent more than 14.3mmt.
According to Microsoft’s annual Environmental Sustainability Report released at the beginning of May, Carbon emissions resurrected from Microsoft after developing AI features hit 30% in 2020, due to the many data centres being built. The company’s direct emissions decreased by more than 6%, yet it was set back by AI necessary upgrades.
In contrast, Meta’s emissions have reached zero in 2020, achieving its zero-value chain emissions plan for 2030. However, the growing need for the furtherance of AI and thus the establishment of more data centres have raised Meta’s footprint, reaching 99% in 2022, as set in their 2023 sustainability report.
It is indeed necessary to balance AI’s benefits with its environmental impact, which determines the continuous development of both AI technology and sustainable endeavours. This has been treated in these companies’ reports, providing new solutions that involve collaboration with giant green-reliant corporations able to mitigate these ever-growing emissions.