Structural megatrends are pressuring water resources, increasing scarcity and challenging long‑term sustainability and economic growth. This creates opportunities for investors to invest in companies providing solutions, ranging from upstream activities such as procurement & equipment, mid‑stream production & distribution, to downstream operation & end use.
In this paper, the Environmental Strategies Group focuses on water use, with particular attention to the relationship between artificial intelligence (AI) and water, to demonstrate why water is becoming a driver of both investment risk and opportunity.
The rapid expansion of AI workloads and the associated boom in hyperscale datacentre construction have drawn attention to the soaring demand for electricity. Yet an equally critical resource is slipping under the radar: water. Across the tech value chain, water is consumed for cooling massive server farms and producing the semiconductor chips that power AI models.
The convergence of escalating demand and water scarcity creates a risk that could substantially reshape investment, regulation and operational strategy for the tech sector.
Rising water demand in tech
Direct use – datacentre cooling
Modern data centres generate substantial heat, making cooling an essential operational consideration. At the server level, cooling has moved from simple fan‑driven air circulation to liquid‑based approaches such as direct‑to‑chip refrigerant loops, cold plates, and immersion cooling. These technologies can mitigate the heat from AI workloads.
At the facility level, the heat removed from servers is expelled from the building in two ways: using chillers, cooling towers and HVAC equipment, or by water‑intensive evaporative and free‑cooling technologies that draw on external water sources.
To keep pace with compute density, direct‑to‑chip refrigerant loops, immersion cooling and hybrid designs now dominate new builds, but these methods rely on substantial water flows. In the US alone, datacentres withdrew an estimated 66 billion litres of water in 2023 – roughly the annual domestic consumption of a mid‑sized city.
Indirect use – semiconductor manufacturing
The chips that drive AI models are fabricated in ultra‑clean environments where ultrapure water (UPW) is indispensable. At a global level, total water consumption by the semiconductor industry is estimated to be comparable to that of a city with a population of approximately 7.5 million people. These amounts are expected to more than double from 2024 to 2028.
Power generation linkage
Water is also included in the AI supply chain: global water use for fossil fuel power generation has risen sharply. Although the transition to renewables should reduce future water use, the current generation mix means that AI‑driven computing carries a hidden water cost.
Constrained supply and heightening tension
Many new datacentre clusters are emerging in regions already experiencing moderate to extreme water stress such as arid zones favourable for solar power, but lacking abundant freshwater. Similarly, a large share of semiconductor fabs sit in water‑stressed basins, exposing them to supply‑side volatility.
A further water-related risk is that of conflict and social opposition as water that could serve growing urban populations or irrigate crops is diverted to power‑intensive tech infrastructure.
At the same time, governments worldwide are moving from voluntary sustainability guidelines to mandatory water‑use reporting, efficiency thresholds, and permitting conditions. The European Union’s Energy Efficiency Directive and US state‑level water‑use permits for datacentre projects represent instances of greater regulation, exposing companies to the risk of delayed approvals, higher compliance costs, and reputational damage.
Emerging solutions to mitigate water risk
Our paper discusses a range of solutions that can help tackle water risk. Here, we present a sample:
Hybrid systems that blend air, liquid, evaporative and free‑cooling techniques enable operators to match cooling intensity with local climate conditions. Closed‑loop liquid cooling can turn water into a reusable heat‑transfer medium. Using ambient air or naturally cool water bodies can dramatically cut water withdrawals, while optimised evaporation can reduce consumption.
Using sensors and other data sources, machine‑learning models can detect leaks, forecast demand spikes, and optimise pumping schedules. Predictive analytics can pinpoint inefficiencies in real time. Water‑use data can be used to support more informed decisions.
Advanced membrane filtration, reverse osmosis and ion exchange technologies enable the reclamation of ultrapure water, reducing fresh‑water intake and lowering discharge fees. Onsite treatment modules can continuously purify and recirculate water, creating a near‑closed circuit.
Conclusion
Water is becoming a central factor of operational resilience for the tech sector, with firms that embed water‑risk assessments into capital decisions gaining easier access to permits, lowering their operating costs, and strengthening their social licences — the acceptance and trust from key stakeholders (communities, regulators etc.) needed to ensure they can continue to operate.
Investment is already flowing to companies that provide ultrapure‑water systems, advanced recycling technologies, and closed‑loop cooling, while operators that continue to rely on high‑withdrawal evaporative cooling or site new capacity without regard to regional water availability risk stranded assets, tighter regulation, and rising utility expenses.
As AI workloads drive compute demand, and countries look to secure supply chain independence and resilience, addressing water scarcity through engineering innovation, data‑driven management and proactive policy will be essential to sustain growth across the entire technology value chain.
