Date: October 6, 2026
Author: Catie Owen, Regional Content & Insights Lead (Rewritten and Expanded for Capacity Global)
Estimated Read Time: 8–10 minutes
Executive Overview
As the artificial intelligence (AI) boom continues to reshape the global technology landscape, a profound physical reality is colliding with digital ambition. While AI models and high-performance computing (HPC) exist largely in the cloud, their physical underpinnings require vast amounts of electrical power, hardware, and—increasingly—thermal management resources.
With standard server rack densities soaring past 100 kilowatts (kW) and frequently pushing toward 130 kW, traditional air-cooling infrastructures are rapidly hitting their thermodynamic limits. The central engineering puzzle for modern data centre operators is no longer simply about how to source more power, but rather how to safely remove staggering quantities of heat without draining local municipal water supplies or destabilizing local ecosystems.
In this high-stakes environment, water and energy infrastructure have become definitive factors in a facility’s "license to operate." To unpack the complexities of liquid cooling, water conservation, and thermodynamic efficiency, Capacity spoke with Matthew McLaughlin, Corporate Account Manager for Ecolab’s data centre business. According to McLaughlin, the industry must pivot away from siloed, linear resource consumption toward an integrated, circular model where water, energy, and compute performance are managed as a single, unified ecosystem.
Detailed Chronology: The Evolution of Data Centre Thermal Constraints
The Air-Cooling Era and the AI Density Shock
For decades, traditional data centres relied on straightforward air-cooling systems. Fans, Computer Room Air Conditioner (CRAC) units, and chillers moved colossal volumes of air across server racks to dissipate heat. While this model was adequate for low-density workloads running at 5 kW to 10 kW per rack, it was never designed for the extreme thermal loads generated by modern generative AI models and GPU clusters.
As rack densities multiplied tenfold, air cooling became unsustainable. Air is a poor thermal conductor; it moves slowly and lacks the specific heat capacity required to pull heat rapidly away from densely packed microprocessors. Relying on air alone at today’s scale demands excessive fan speeds and chiller loads, consuming a massive share of a facility’s total power budget (expressed as Power Usage Effectiveness, or PUE).
The Pivot to Liquid and Circular Models
Recognizing that air could no longer keep pace with chip evolution, the industry began introducing liquid cooling variants—ranging from rear-door heat exchangers to direct-to-chip liquid cooling systems. However, early implementations often treated water and liquid coolants as disposable utilities, pulling them from local grids and discharging them with little regard for circularity.
By 2026, regulatory scrutiny, community pushback in water-stressed regions, and grid capacity constraints forced a structural evolution. Leading operators began adopting closed-loop, circular water systems and advanced liquid-to-air or liquid-to-liquid heat rejection chains. Ecolab emerged as a key facilitator in this transition, helping international operators optimize water chemistry, implement real-time fluid monitoring, and balance complex thermal chains without compromising operational uptime.
Supporting Context & Metrics: The Physics and Math of Cooling
To understand why the industry is undergoing such a radical architectural shift, one must first look at the fundamental physics of heat transfer.
The Superior Physics of Water
As Matthew McLaughlin explained during his discussion with Capacity, liquid—specifically water—is vastly superior to air when it comes to thermal dynamics.
- Heat Capacity: Water can carry roughly 3,500 times more heat than an equivalent volume of air.
- Heat Transfer Speed: Water transfers thermal energy about 23 times faster than air.
Using a familiar mechanical analogy, McLaughlin noted: "Think of a hot engine: a fan can cool its surface, but circulating coolant through the engine removes heat directly from where it is generated. The same principle applies to high-density compute. At today’s densities, that difference is decisive."
By bringing liquid loops—such as water or specialized dielectric coolants—closer to the chip, operators capture heat at higher temperatures directly at the source. This requires significantly less energy to move the heat out of the facility.
The Hidden Water Footprint of Electricity
A critical blind spot in many data centre sustainability metrics is the invisible water tied to electrical generation. Generating a kilowatt-hour of electricity at a fossil-fuel or nuclear power plant often requires substantial amounts of water for cooling and steam generation.
Consequently, energy efficiency and water conservation are inextricably linked. Every kilowatt-hour saved by an efficient cooling architecture simultaneously reduces the "embedded water" consumed off-site by the electrical grid.
- On-site: Water is utilized purposefully, treated, and kept in continuous circulation.
- Off-site: Lower total power demand directly translates to reduced water consumption at upstream power generation facilities.
Official Insights & Expert Perspectives
Breaking Down the Cooling Chain
A common misconception in facility design is that cooling is a binary choice between competing technologies. In practice, modern high-density data centres operate along a sophisticated cooling chain:
- The Chip Level: Heat moves from the processor to a cold plate via direct-to-chip liquid cooling.
- The Rack Level: Heat is transferred from the rack-level manifolds to facility loops.
- The Facility Level: Heat is ultimately rejected into the atmosphere using evaporative cooling towers, dry coolers, adiabatic systems, or hybrid configurations.
According to McLaughlin, the optimal mix depends entirely on three local variables: climate, local power grid characteristics, and regional water availability. Operators who attempt to copy-paste a generic cooling template from one region to another frequently run into performance bottlenecks.
The Danger of Isolated Metrics
McLaughlin issued a stark warning against optimizing single metrics in isolation. Facilities frequently track Power Usage Effectiveness (PUE) and Water Usage Effectiveness (WUE) on entirely separate scorecards.
However, these metrics are intrinsically linked. An aggressive water-saving measure on-site might require higher electrical loads to power pumps and chillers, which in turn drives up energy consumption and inadvertently increases water use upstream at the power plant.
"The better approach treats heat, water, and energy as one connected balance, measured inside and outside a facility based on climate, power grid, water availability, and performance requirements," McLaughlin noted.
Fluid as Critical Infrastructure
Whether an operator utilizes closed-loop systems, direct-to-chip cooling, or advanced water reuse schemes, efficiency relies entirely on the health of the fluid itself. Water and coolants are not "set-and-forget" utilities.
Over time, factors such as:
- Corrosion
- Microbiological growth
- Particulate accumulation
- Mineral variability and scaling
- Coolant drift
…can severely degrade heat-transfer efficiency. To combat this, Ecolab advocates for real-time monitoring of coolant chemistry, temperature, and flow rates.
McLaughlin highlighted real-world examples where continuous monitoring prevented catastrophic failures:
- Hyperscale Site: Real-time chemical tracking detected microscopic corrosive particles that routine visual maintenance missed, allowing engineers to resolve the issue before hardware damage occurred.
- Colocation Provider: Automated sensors flagged diluted coolant during the critical facility commissioning phase, enabling an immediate correction that kept the buildout on schedule and eliminated future outage risks.
Future Outlook: Building for a Sustainable AI Era
As the artificial intelligence boom matures through the latter half of the decade, global demand for computational power will only escalate. However, exponential technological growth does not have to result in exponential resource waste.
While electrical grid capacity can theoretically be expanded through new generation projects and transmission lines, water remains a finite, highly localized resource. Over-extracting water or mismanaging local watersheds can lead to severe ecological consequences and community friction.
To navigate this tightening regulatory and environmental landscape, McLaughlin offers a four-point roadmap for data centre operators planning high-density builds or retrofits:
- Design for the Site: Let local climate, power sources, water availability, and workload profiles shape your cooling design—never the other way around.
- Unify Metrics: Model WUE and PUE side by side, explicitly factoring in the embedded water footprint of your upstream power generation sources.
- Match the Method to the Heat: Deploy high-efficiency liquid cooling where density demands it, and pair it with facility-level heat rejection mechanisms tailored to local environmental conditions.
- Treat Fluids as Infrastructure: Implement real-time chemical and flow monitoring from day one of commissioning to ensure that thermal efficiency does not degrade over the lifecycle of the system.
Data centre operators who successfully merge water and energy management into a single, cohesive system will be best positioned to meet the insatiable demands of the AI revolution—delivering massive compute capacity while protecting the vital resources communities depend upon.
For further discussions on digital infrastructure and sustainability, industry leaders will gather at Capacity Europe 2026 on October 13, 2026, marking the 25th anniversary of the event with over 3,500 global decision-makers.
