Protecting groundwater so AI can keep growing
AI is growing at extraordinary speed. Behind that growth, one resource has recently been attracting more attention in the United States: water.
AI requires large data centers, where many computers operate around the clock. They generate enormous heat and need cooling. Water is one of the resources used for that purpose, and in some regions that water comes from underground.
As I wrote in my first note article, I am not opposed to AI or data centers. Having spent many years in IT, I look forward to the changes AI will bring to society.
That is precisely why I believe we need to ask: How can we make the water needed for AI’s growth sustainable?
A statement from groundwater specialists in the United States
In February this year, the National Ground Water Association (NGWA), a US organization of groundwater specialists, released a position paper titled “Data Centers and Groundwater Resources.”
NGWA: Data Centers and Groundwater Resources
The paper does not reject the construction of data centers. It recognizes the centers supporting AI and cloud services as essential infrastructure for modern society.
At the same time, it argues that data-center growth must be compatible with protecting groundwater. I found that point particularly important.
Knowing how much water is used is not enough
One aspect of the paper that caught my attention was transparency. NGWA calls on data centers to disclose information such as:
- Where their water is sourced
- Annual and peak water withdrawals
- The volume of wastewater discharged
- How much water is actually consumed
- The relationship between cooling methods, electricity and water use
These disclosures should continue beyond the planning stage, through permitting and actual operation.
In other words, saying “this data center uses a certain volume of water each year” is not enough. We need to ask where the water comes from, what will happen to local groundwater and rivers if it continues to be used, and what happens during a drought.
Water is different in every place
This is a crucial point. Using one million liters of water has a very different meaning in a water-rich area than in a water-scarce one.
Groundwater flows differ by location. So do rainfall, geology and forests. Some regions receive snow.
NGWA therefore calls for site-specific assessments of water resources and environmental capacity. It also asks that operators evaluate impacts on neighboring water users, groundwater and surface water, and carry out appropriate monitoring, rather than considering withdrawals alone.
I believe this will become extremely important in Japan too.
Technologies that use less water are advancing
Data-center technology is also progressing. The paper introduces geothermal cooling, including closed-loop systems that release heat underground. NGWA notes the potential to substantially reduce cooling-water use and, in some cases, eliminate consumptive water use.
I expect these technologies to develop further. But I do not think technology alone is the whole answer.
Think about using water and nurturing water together
Data centers use water, so conserving it matters. But could we go a step further?
How is water replenished in that region in the first place? Rain falls into forests, soaks into soil, enters the ground and slowly becomes groundwater. It emerges as springs and flows into rivers.
Could companies participate not only as users of water, but also as partners in nurturing their region’s water?
That is the framework we are considering at amenowa.
Starting small in Rusutsu, Hokkaido
At amenowa, we have begun observing the water cycle in Rusutsu, Hokkaido. We measure rain and observe snow, groundwater levels, springs and forests.
We start by asking how water circulates through this particular place.
In the future, by managing forests and creating conditions that help water infiltrate the ground, we hope to make visible the balance between the water people use and the water nature replenishes.
AI and nature do not have to be on opposite sides
I do not see this as a choice between AI and nature. AI will continue to grow, and there will probably be more data centers.
That makes it essential to build groundwater protection into the infrastructure supporting AI from the start. This discussion is already underway in the United States.
NGWA concludes with the idea that balancing data-center growth with long-term groundwater protection is both possible and necessary. I agree.
Rather than stopping technology, we should ask how to protect nature so that technology can continue for the next hundred years. We should nurture water as well as use it.
I plan to follow international news on AI and water, companies and water, and forests and water each week, and write about what we can do in Japan.