THE WATER FOOTPRINT
FRESHWATER RISK ASSESSMENT
Fazmina Imamudeen investigates the growing footprint of the AI revolution
AI has sparked a global race to build faster chips, larger data centres and more powerful computing infrastructure. Governments are competing to attract investments, tech companies are announcing billion dollar expansions and investors are pouring unprecedented sums into digital infrastructure.
Yet, beneath this race lies the resource of water, which is receiving far less attention.
The issue came into sharper focus in June when the UN called on AI companies to disclose the environmental cost of their technologies and warned that the rapid expansion of data centres could place growing pressure on already stressed freshwater resources.
While energy consumption has long dominated discussions surrounding artificial intelligence, water is emerging as another critical constraint.
Every data centre generates enormous amounts of heat. Without continuous cooling, servers can overheat, reduce performance and damage equipment. Therefore, many facilities rely on cooling systems that consume large quantities of freshwater while the electricity required to power such centres also carries an indirect water footprint through power generation.
Research published by the University of California estimated that global AI related water withdrawal could reach between 4.2 and 6.6 billion cubic metres annually by next year. At the facility level, the largest hyperscale data centres can consume millions of gallons of water each day depending on their design, local climate and cooling technology.
According to the UN, global freshwater demand is projected to exceed sustainable supply by up to 40 percent by 2030 if current trends continue. Agriculture remains the largest consumer of freshwater but AI infrastructure is becoming a significant new industrial user in regions where water resources are already under pressure.
Data centre developers are increasingly assessing water availability alongside electricity rates, fibre connectivity and tax incentives when selecting locations.
In several parts of the US, proposed facilities have faced public opposition because of concerns over their impact on local water supplies. Similar debates are emerging in Europe and Australia, reflecting a broader recognition that digital infrastructure has physical resource limits.
Microsoft has announced new data centre cooling systems designed to operate without consuming water for cooling during most of the year. Meanwhile, other operators are investing in closed loop cooling technologies that recycle water rather than continuously draw fresh supplies.
These innovations are driven by commercial necessities rather than sustainability goals in regions where water scarcity could constrain future expansion.
Artificial intelligence depends on advanced semiconductors, and semiconductor fabrication requires vast quantities of ultrapure water.
Taiwan’s repeated droughts demonstrate how water shortages can threaten one of the world’s most important manufacturing industries and force authorities to prioritise water supplies to chip fabrication plants.
In a global economy that’s dependent on semiconductors, local water shortages can quickly become international business risks.
Building an AI economy is no longer simply a question of providing land, tax incentives and electricity; it also requires long-term confidence that water resources can sustain decades of industrial growth.
Countries that are able to provide reliable water infrastructure may gain a competitive advantage in attracting investments while those facing chronic water stress could find themselves constrained despite strong digital ambitions.
This conversation is timely for Sri Lanka too.
As nations compete to attract tech investments, reliable water management may prove to be an overlooked economic asset. Protecting watersheds, modernising water infrastructure and improving resource efficiency are investments in the resilience needed to participate in the next phase of the digital economy.
Artificial intelligence is often described as the defining technology of this century. Yet, its success depends on resources that are far older than the technology itself – and the race to build the future may ultimately be limited not by computing power but access to freshwater.
