Artificial intelligence is transforming our lives, reshaping sectors such as education, healthcare, the environment, and the workplace. It enhances accessibility with assistive technologies for people with disabilities. However, AI’s rapid development raises concerns over its ethical implications and environmental footprint.
This article explores AI’s energy consumption and ongoing efforts to reduce its environmental impact.
Understanding AI’s environmental footprint
The rise in AI usage has raised the demand for more data and computing power, placing a significant strain on our natural resources. The UN Environment Programme points out that we need to evaluate software and hardware life cycles together in assessing AI’s environmental footprint, as both are linked. The software life cycle includes data collection and preparation, model building, training, validation, deployment, inference, maintenance, and retirement. The hardware life cycle includes raw material extraction, production, transportation, and data centre construction, followed by e-waste management, maintenance, and disposal. Evaluating AI’s hardware life cycle is complex because each stage has its environmental impact, from mining and extraction to transportation, energy and water consumption, and e-waste generation.
AI’s overall environmental impact falls into three categories:
- Direct: Greenhouse gas (GHG) emissions from computing, energy and water consumption, mineral extraction, pollution, and e-waste production.
- Indirect: GHG emissions from AI applications and machine learning.
- Higher-order effects can amplify existing inequalities, biases, and poor quality in training data.
How does AI affect the environment?
Large-scale AI deployments pose several environmental concerns. Most AI servers are stored in data centres, which produce electronic waste and can contain toxic chemicals, such as mercury and lead. Data centres consume vast amounts of electricity, creating greenhouse gas emissions. They also require large amounts of water for construction and to cool the electrical components. Global AI demand is expected to consume 4.2-6.6 billion cubic meters of water by 2027, surpassing Denmark’s total annual water withdrawal of 4-6 billion cubic meters.
Although the digital economy is sometimes viewed as virtual or in the “cloud”, it is nonetheless highly reliant on physical resources and raw materials. Digital devices, hardware and infrastructure are made of plastics, glass, ceramics, and various minerals and metals. Data centres rely on minerals and rare elements, which are often mined unsustainably. Making a 2-kilogram computer requires approximately 800 kg of raw materials.
How much electricity does ChatGPT use to answer your question?
AI-powered virtual assistants such as ChatGPT use more energy than traditional search engines. According to the International Energy Agency (IEA), a single ChatGPT request requires ten times more electricity than a Google search. The average ChatGPT query costs approximately 0.36 cents (USD). Machine learning and AI accounted for less than 0.2 per cent of global electricity demand and less than 0.1 per cent of global GHG emissions in 2021. However, the demand for AI computing is increasing rapidly. In recent years, Meta has seen an annual increase in computing demand for machine learning training and inference of more than 100 per cent. As AI use grows, energy demand will increase, making the use of low-carbon energy sources essential to reducing GHG emissions.
Growing need for data centres
Data centres are the backbone for storing, processing, and distributing data for different applications, including websites, cloud, or AI services. Data centres that host AI technology consume vast amounts of energy to power their complex electronics, the majority of which still comes from fossil fuels, contributing to greenhouse gas emissions. The rapid growth of AI has increased new data centre investments to accommodate growing power demands. In 2022, data centres accounted for about 1 per cent of global electricity demand, which is only expected to rise. In Ireland, where the data centre market is developing rapidly, electricity demand from data centres represented 17% of the country’s total electricity consumption for 2022. If this trend continues, Ireland’s data centres will double their electricity consumption by 2026.
The number of data centres has increased from 500,000 in 2012 to 8 million today, and experts predict that AI’s escalating energy needs will sustain this rapid growth.
Can AI be the solution?
Despite its environmental impact, AI also has the potential to reduce its footprint. AI algorithms can identify patterns in data, detect anomalies, and anticipate and forecast future results. AI might help governments, organizations, and individuals monitor environmental changes and make more responsible decisions. AI may also accelerate innovations in energy technologies.
According to the UN Environment Programme’s Climate Technology Progress Report 2024, AI is becoming increasingly important in mapping renewable energy potential, optimizing efficiency, and facilitating interconnectivity with other sectors, such as water and agriculture. However, AI cannot fully replace the physical infrastructure and governance systems essential for the energy transition. Strong governance frameworks are needed to ensure the responsible use of AI in renewable energy projects. National policies that include circular economy strategies can help to reduce the growing demand for ICT hardware and infrastructure. However, financial barriers persist, particularly in developing countries, limiting the mainstream use of AI-driven sustainability solutions.
What is being done to address AI’s environmental impact?
Governments and international organizations are taking steps to mitigate AI’s environmental footprint. Over 190 countries in the UN system have adopted the UNESCO Recommendations on the Ethics of Artificial Intelligence, which address AI’s ethical application, including its environmental impact. The European Union has passed the AI Act, a legislative framework regulating AI’s environmental impact.
To curb the environmental fallout from AI, the UN Environment Programme recommends that:
- Countries develop standardized methods to measure AI’s environmental footprint.
- Governments develop regulations requiring companies to disclose the environmental impact of AI-based products and services.
- Tech companies make AI algorithms more energy-efficient, reducing their energy demand while recycling water and reusing components where feasible.
- Countries encourage organizations to use renewable energy and carbon offset to green their data centres. AI-related policies should also be integrated into broader environmental regulations.
While AI and digital transformation offer opportunities for social and economic progress, their environmental effects are complex, impacting planetary health, environmental sustainability, and human well-being. The rising demand for critical minerals, rare earth elements, and water resources to support expanding data centres requires careful assessment. To reduce AI’s environmental impact, it is essential to prioritize e-waste recycling, energy-efficient data centres, renewable energy adoption, and responsible resource management.
Further reading: