With advances in technology, including artificial intelligence, cloud computing, and advanced semiconductors, digital economies around the world are rapidly transforming. This swift transformation is built on infrastructure that is often overlooked, which is water.

Behind each AI model trained, semiconductor chip produced, and hyperscale data centre operated is an enormous water footprint that goes unnoticed. Cooling high-density computing clusters, generating electricity required to support AI workloads, and manufacturing advanced chips require vast amounts of water, usually in areas experiencing significant water shortages.

The reliance on water will only continue to grow as the demand for AI infrastructure grows. Global AI infrastructure is projected to consume approximately 312 billion to 764 billion litres of water each year (primarily due to data centre cooling and electricity generation). Additionally, estimates indicate that AI-only data centre water consumption could exceed 1 trillion litres by 2028, an increase of nearly 1,000 per cent from 2024 levels. The manufacturing of semiconductors will also contribute significantly to this increasing demand, as one advanced semiconductor manufacturing facility utilizes approximately 10 million gallons of ultra-pure water each day.

As the world races to increase its AI capacity, water is becoming one of the most important resources to consider in building digital infrastructure. Managing this hidden reliance will be critical to ensuring that digital growth is feasible and sustainable.

Cooling challenge of AI infrastructure Modern AI systems run on high-performance GPUs and specialised chips that generate enormous heat. Cooling these systems is not a marginal operational issue; it is central to data-centre design.

Traditional evaporative cooling systems rely heavily on water. Studies estimate that around 80% of water used in many data-centre cooling processes is lost through evaporation, making water efficiency a structural challenge rather than a temporary operational one.

As AI workloads scale, these cooling requirements are intensifying. AI-optimised data centres operate at significantly higher power densities than traditional cloud infrastructure, driving both electricity and water demand upward. Importantly, much of this water footprint is indirect: power plants generating electricity for data centres also consume large volumes of water.

This dual dependency, cooling and energy, means the true water footprint of AI is embedded across the entire infrastructure ecosystem.

The water cost associated with intelligence Another factor in the digital water equation is the semiconductor industry. Chip fabrication processes require ultra-pure water during wafer cleaning, etching, and chemical processing. Even a small contaminant (particle) can ruin the yield of a specific chip, so the supply and purity of water are two items that cannot be negotiated.

Modern semiconductor fabrication plants utilize millions of gallons of ultra-pure water on a daily basis for cleaning, etching, and processing. Much of this ultra-pure water must undergo complex purification systems before reuse or discharge.

As more chips are connected to each other via AI Accelerators, large amounts of high-bandwidth memory, and advanced packaging materials are required to produce these next-generation chips, so there will be an incredible increase in global capacity for semiconductor manufacturing. This directly translates into increased demand for industrial water, especially in areas that are preparing for or are already facing a water crisis.

In effect, the physical infrastructure of intelligence, from chips to servers, is also an infrastructure of water.

Geography of digital infrastructure and water risk In an era when digital infrastructure is expanding, the locations where it is built have become increasingly important.

Hyperscale data centres and semiconductor fabrication are often chosen based on the availability of energy, cost of land, and level of connectivity. However, the decision to locate these centres or fabs can unintentionally exacerbate water stress, especially in places where demand for cooling is high due to increased ambient temperatures.

This situation creates a strategic conflict between the need for resilient infrastructure in the digital economy and the fact that resilient infrastructure depends upon having access to sufficient amounts of water. Consequently, it is becoming an important factor for policymakers, regulatory bodies, and the companies involved in building digital infrastructures; the availability of water will be critical to future digital infrastructures.

As countries around the world increase their investments in AI and semiconductor ecosystems, water governance will become an essential part of their overall technology-related strategic policies.

Towards a water-aware digital economy The next phase of digital transformation must recognise that computing power is not purely a function of silicon and electricity; it is also a function of natural resources.

To tackle the unseen water footprint of artificial intelligence, we need to increase the transparency of water use reporting, the usage of advanced cooling systems, and the use of environmental metrics in planning digital infrastructure; however, there is another key issue here. Water must also be recognized as a critical component of the digital value chain, similar to energy & data; most importantly, pertinent to countries that wish to establish themselves as global digital hubs, they need to create the computing power necessary for the AI era with sustainable resource foundations.

The outcome of the digital economy won’t be determined by just the power of their computers alone. It will be increasingly determined by how effectively we manage the resources necessary for computing.

By: DocMemory
Copyright © 2023 CST, Inc. All Rights Reserved