#130 WEEKEND SPECIAL - Diving Deep: Understanding AI's Water Footprint and Environmental Impact
Fresh & Hot curated AI happenings in one snack. Never miss a byte 🍔
This snack byte will take approx 4 minutes to consume.
AI BYTE # 📢: Diving Deep: Understanding AI's Water Footprint and Environmental Impact
In the era of rapid technological advancements, AI stands at the forefront of innovation, driving change across various industries. However, this progress comes at a cost, particularly concerning the environment.
One of the less discussed but significant impacts of AI is its excessive water consumption, which poses a threat to its environmental contributions.
AI’s Thirst for Water
AI’s environmental footprint extends beyond carbon emissions, with a substantial water footprint that often goes unnoticed. The production and operation of AI technologies require vast amounts of water, particularly for cooling data centers that house the servers running AI algorithms. These data centers consume water in two primary ways: directly, through cooling systems that maintain optimal temperatures for the hardware, and indirectly, through the generation of electricity that powers them.
Initial research shows that AI has a significant water footprint. It uses water both for cooling the servers that power its computations and for producing the energy it consumes. As AI becomes more integrated into our societies, its water footprint will inevitably grow.
The growth of ChatGPT and similar AI models has been hailed as “the new Google.” But while a single Google search requires half a millilitre of water in energy, ChatGPT consumes 500 millilitres of water for every five to 50 prompts.
AI uses and pollutes water through related hardware production. Producing the AI hardware involves resource-intensive mining for rare materials such as silicon, germanium, gallium, boron and phosphorous. Extracting these minerals has a significant impact on the environment and contributes to water pollution.
Semiconductors and microchips require large volumes of water in the manufacturing stage. Other hardware, such as for various sensors, also have an associated water footprint.
Data centres provide the physical infrastructure for training and running AI, and their energy consumption could double by 2026. Technology firms using water to run and cool these data centres potentially require water withdrawals of 4.2 to 6.6 billion cubic metres by 2027.
By comparison, Google’s data centres used over 21 billion litres of potable water in 2022, an increase of 20 per cent on its 2021 usage.
Training an AI at the computing level of a human brain for one year can cost 126,000 litres of water. Each year the computing power needed to train AI increases tenfold, requiring more resources.
Water use of big tech companies’ data centres is grossly underestimated — for example, the water consumption at Microsoft’s Dutch data centre was four times their initial plans. Demand for water for cooling will only increase because of rising average temperatures due to climate change.
The Environmental Impact
The water consumption associated with AI is not just about the quantity but also the quality of water being used. The process of producing AI hardware involves resource-intensive mining for rare materials, which significantly impacts the environment and contributes to water pollution. Moreover, the demand for fresh water to cool AI infrastructure strains limited water resources, exacerbating global water scarcity.
Despite these challenges, AI has the potential to make significant environmental contributions. It can improve energy efficiency, optimize renewable energy deployment, and monitor deforestation from satellite images. AI can also aid in tackling climate change by predicting weather patterns, tracking icebergs, and identifying pollution sources.
The Need for Sustainable Practices
To mitigate the water footprint of AI, the industry must prioritize sustainable practices. This includes optimizing algorithms for energy efficiency, developing more sustainable data center designs, and shifting intensive AI tasks to periods of lower electricity demand to reduce reliance on non-renewable energy sources.
Addressing AI’s water consumption requires a collaborative effort between policymakers, researchers, and industry leaders. Innovations in AI should be guided by principles of sustainability, ensuring that the technology’s development does not compromise the planet’s water resources.
Taiwan, responsible for 90 per cent of the world’s advanced semiconductor chip production, has resorted to cloud seeding, water desalination, interbasin water transfers and halting irrigation for 180,000 hectares to address its water needs.
As water becomes increasingly expensive and scarce in relation to demand, companies are now strategically placing their data centres in the developing world — even in dry sub-Saharan Africa, data centre investments are increasing.
Google’s planned data centre in Uruguay, which recently suffered its worst drought in 74 years, would require 7.6 million litres per day, sparking widespread protest.
Conclusion
As we continue to harness the power of AI, it is imperative to maintain a balance between technological progress and environmental conservation. The conversation around AI’s environmental impact must include its water consumption, prompting action towards more sustainable practices.
Only then can we ensure that AI’s contributions to the environment are not drowned out by its own demands.