Artificial Intelligence (AI) has become one of the most widely used technological tools of the moment; however, it is intensifying the global environmental crisis with its growing ecological footprint. Each interaction with advanced AI models consumes large amounts of energy, requires water for cooling, and contributes to the generation of electronic waste, significantly increasing its ecological impact.
Training, inference, and data storage in AI systems demand enormous amounts of energy. For instance, a query in ChatGPT requires up to ten times more electricity than a Google search. In high-traffic platforms like virtual assistants and recommendation engines, the energy impact becomes even more critical.
Moreover, training deep learning models like GPT-3 demands intensive use of specialized hardware such as GPUs and TPUs, leading to increased electricity consumption. For example, Hugging Face consumed 433 megawatt-hours (MWh) to train a multilingual model, enough to power 40 American households for a year.
Additionally, data centers hosting these models operate continuously, with 40% of their consumption dedicated to servers and another 40% to cooling systems. Just in training GPT-3, 700,000 liters of water were used to cool the servers.
An Increasing Carbon Footprint
Major tech companies are expanding their AI infrastructure globally, leading to a surge in their carbon emissions. Over the past five years, Google’s emissions increased by 48%, while Microsoft’s grew by 30% between 2020 and 2023.
In 2022, AI and the cryptocurrency sector consumed 460 terawatt-hours (TWh), 2% of global electricity, according to the International Energy Agency. It is estimated that by 2026, this figure could double, reaching the equivalent of Japan’s total consumption.
If all Google searches incorporated AI, it would require 29.2 TWh per year, equivalent to the annual electricity consumption of Ireland. In Nvidia’s case, whose hardware is crucial for AI, its total consumption could range between 85 and 134 TWh annually, similar to Argentina’s consumption.
The use of AI leads to increased greenhouse gas emissions. Photo: Pixabay.
Seeking Sustainable Solutions
The environmental impact of AI largely depends on the energy source used. If powered by electricity generated from fossil fuels, its carbon footprint soars.
Companies like Google, Microsoft, and Amazon have pledged to reduce their emissions and operate on renewable energy by 2030 or 2040. However, these commitments were made before the massive expansion of AI, raising uncertainties about their actual viability. The challenge now is to balance the advancement of artificial intelligence with sustainable practices to minimize its impact on the planet.
AI and Its Ecological Footprint
Artificial Intelligence (AI) has a negative environmental impact, consuming significant amounts of water and electricity, and generating carbon emissions. It also produces electronic waste and requires raw materials that are unsustainably sourced.
The environmental impacts of AI include:
- Carbon emissions: AI generates carbon emissions during training and usage. Data centers, hosting AI servers, are a major source of the ecological footprint’s growth.
- Water consumption: It consumes large amounts of water, which is increasingly scarce in many regions.
- Electronic waste: The manufacturing of AI hardware generates electronic waste.
- Use of raw materials: AI hardware manufacturing requires raw materials, including critical minerals and rare elements, often unsustainably extracted.
However, AI can also help protect the environment, for example, in disaster prevention and mitigation.
Have you visited our YouTube channel yet? Subscribe now!
