Photo by Sergei Starostin: https://www.pexels.com/photo/network-servers-on-an-enclosure-6466141/
Artificial intelligence has hit society just like the internet did. It is considered to be a digital revolution for society; however, there is a high demand for AI and because of this, there needs to be a strong physical infrastructure put in place. This includes rows of servers, computer chips, electricity networks, and data centres. Due to the high demand for AI, another resource in demand, and arguably one of the most valuable, is water.
As demand for AI continues, so does the requirement for data centres that are capable of training and running highly powerful models. The facilities consume a lot of electricity and generate an obscene amount of heat, which means they require highly efficient cooling systems to stop the equipment from overheating, and this is where water plays a vital role.
Despite this being a revolutionary piece of technology, there is one question that is a great concern for many, and that is: Can AI infrastructure coexist with a world facing increasing water scarcity?
The Hidden Water Footprint of AI
Data centres need to manage heat efficiently, and to do this, you need advanced cooling systems. These use water, which then absorbs and removes heat from the equipment.
The amount of water that is used depends on a number of factors such as the facility, other cooling technology, local climate, and the overall efficiency of its infrastructure. This is why researchers are unable to put a single figure to water consumption for an individual AI query or model.
One thing for certain is that the trend is very clear. AI is becoming widely used throughout the world, and the need for infrastructure is constantly increasing. The more servers used, the more heat is produced, which means more demand for water and efficient cooling systems.
Additionally, another core issue is that a large portion of data centers are located in areas where there are already water shortages.
Location Is Key
Water scarcity isn’t distributed evenly around the world. For example, a data centre that operates in a cool, water-rich area faces different constraints from those in areas that are already suffering from water shortages. This is why the location of data centres has become so important when considering where to put them.
The companies looking to construct the infrastructure of data centres need to carefully consider electricity, internet connectivity, land, and tax incentives, as well as access to reliable cooling resources.
For regions facing drought, the arrival of large data centres will raise questions about the demand for water. Farmers, industry, residents and ecosystems are already competing for water due to the limited availability in these areas. Adding other industrial facilities to the area can escalate these debates over who has access to the water and under what conditions.
Can Technology Be The Solution?
A common misconception with this debate is that technology is the cause of water scarcity; however, it can also be the solution.
There is a large investment in technology to improve cooling technologies, including systems to reduce or eliminate the constant need for freshwater consumption. For example, closed-loop cooling can recirculate water rather than constantly requiring new supplies. There are other developments in alternative cooling and air technology; however, this also needs to factor in climate and energy requirements.
Another consideration is reusing used water instead of potable supplies. If treated and regulated like ultrapure water, it will reduce pressure on drinking-water resources significantly.
Artificial intelligence could actually make it more efficient as well. With the use of AI systems, these can be used to monitor data centre temperatures, predict workloads, and even optimise cooling operations in real time. The little improvements made to efficiency, multiplied across large facilities, could potentially reduce the water and energy demand we are seeing with modern data centres.
A Question of Priorities
The debate ultimately extends past technology. The core question is how it is raised a fundamental question about how societies allocate scarce resources.
There are big promises, from scientific research and medical applications to even new forms of communication. These promises are what is creating the demand for more data centres around the world. That being said, there are physical limits to this. At the end of the day, these data centres won’t function effectively if we don’t have the electricity, raw materials for chips, and efficient cooling systems, including water to run them.
To ensure that we live in a world where both water and AI technology coexist, more research must be done for this to be possible. Solutions would include new data centres in regions where there is a good water supply, improving the efficiency of cooling systems and using recycled water, improving transparency, and ensuring that the local community is not affected in a negative way by the data centres. The future lies with how well we manage the resources that they require, and not how intelligent AI itself becomes.