Artificial Intelligence Prosperity Demands Water - Local Communities Face the Burden
In the rapidly evolving world of artificial intelligence, a growing concern is the impact of data centers on water resources. As these facilities expand, policymakers, communities, and tech companies are increasingly focusing on solutions to mitigate the rising water demands.
Recent research estimates that a series of 20 to 50 queries to a large language model like ChatGPT can require the consumption of roughly half a liter of fresh water for cooling the servers processing the request. However, the water usage can be much higher for physical data centers. For instance, a single older data center like Meta's can use 500,000 gallons of water per day.
The Netherlands has seen a notable example of this, with a Microsoft data center consuming 84 million liters of drinking water, significantly higher than the estimated 12-20 million liters. Across Newton County, Georgia, a larger water crisis is unfolding due to data centers, with newer facilities seeking permits for millions of gallons daily. One permit in Georgia requested nine million gallons a day. The Morrises, a retired couple in rural Newton County, have experienced water issues since Meta began building a data center near their home. Three of their neighbors also report similar problems.
In response to these concerns, several key approaches are being considered:
- Using alternative water sources: Companies like Google are already utilizing canal water at their data centers, while services providers such as Veolia offer water treatment and reclamation to reduce dependency on drinking water supplies.
- Improving cooling system efficiency: Upgrades to cooling tower controls and treatment protocols have doubled cycles of concentration in some data centers, cutting water use by up to 50% and saving millions of gallons annually.
- Deploying waterless or closed-loop cooling systems: Microsoft’s data center in Wisconsin uses a closed-loop system designed to consume zero water for cooling. However, these systems often require higher energy consumption, which can indirectly increase water use at electricity generation plants.
- Site selection policies: Policymakers should prioritize water sustainability alongside energy costs when selecting sites for future data centers. This would help prevent exacerbating water stress in drought-prone areas.
- Regulatory frameworks and community engagement: Public backlash and regulatory reconsiderations, like those faced by Google in Chile, highlight the need for environmental impact assessments that rigorously consider water impacts in permitting decisions.
In Phoenix, Arizona, developers have paused homebuilding due to drought exacerbated by tech facilities. In Uruguay, Google's plan to build a data center that would consume 7.6 million liters of potable water per day sparked public outrage. The AI revolution is expected to increase water demand, and communities should not be left dry in the race to power artificial intelligence.
If a data center brings jobs and investments into an area, a part of that money should go to water investments. Policymakers should treat water as a critical and finite resource, not a disposable afterthought. Transparent reporting of water usage is necessary for governments to make informed decisions. The Morrises, who have spent $5,000 trying to fix the water issues near their home, are a stark reminder of the real-world impacts of data center water consumption. Beverly Morris stated that she is scared to drink their own water.
In Colorado, water-hungry data centers are part of renegotiations over the Colorado River. As the world continues to embrace AI, it is essential that we find solutions to balance the needs of technology with the needs of our planet's most precious resource: water.
- The concerns about artificial intelligence's impact on water resources are growing, as data centers expand, requiring more water for cooling servers.
- A large language model like ChatGPT requires roughly half a liter of fresh water for cooling each series of 20 to 50 queries, while physical data centers can use significantly more.
- For instance, Meta's older data center uses 500,000 gallons of water per day, and a Microsoft data center in the Netherlands consumed 84 million liters of drinking water.
- In Newton County, Georgia, a larger water crisis is unfolding due to data centers, with newer facilities seeking permits for millions of gallons daily.
- To mitigate this issue, companies are exploring various solutions, such as using alternative water sources, improving cooling system efficiency, deploying waterless or closed-loop cooling systems, and selecting water-sustainable sites for data centers.
- Policymakers should also prioritize transparency in reporting water usage, engage communities, and establish regulatory frameworks that consider water impacts in permitting decisions.
- As the AI revolution continues, it's crucial to find ways to balance the needs of technology with the needs of the planet's water resources for sustainability.
