By: Canninah Dladla - Cluster President for Sub-Saharan Africa at Schneider Electric

Nearly 60% of operators are expected to adopt liquid cooling within five years as AI-driven economics reshape infrastructure investment priorities. However, for the modern data centre operator, the shift to liquid cooling represents a considerable change in direction that will define the competitive landscape of infrastructure for the next decade. 

As AI data centre cooling demands accelerate, operators are evaluating liquid cooling data centre strategies as the primary solution for high-density environments. 

However, for CTOs and infrastructure leaders, success in the AI era depends on understanding the full financial picture of liquid cooling, starting with upfront capital investment and planning strategically for long-term operational and performance benefits.

The chiller paradox: balancing upfront capital with long-term performance

The first critical cooling element to consider is a facility’s heat rejection system. Liquid cooling for AI workloads enables higher rack densities by removing heat more efficiently than traditional air-based systems. 

Water holds 3,000 times more heat by volume than air which means liquid cooling can capture heat at significantly higher temperatures. 

However, this also creates a critical design crossroads: do you leverage an existing "low-temperature" chiller plant for a hybrid environment, or invest in a dedicated "high-temperature" system specifically for liquid-cooled loads?

While a shared chiller avoids immediate Capex, it anchors the entire facility to the inefficient, frigid temperatures required by legacy air-cooled gear. In contrast, investing in dedicated high-temperature chillers unlocks massive "economiser hours," where the system runs without energy-intensive compressors, turning a capital expense into a massive engine for operational savings.

The real benefit lies in the Technology Cooling System (TCS) fluid temperature. Boosting the TCS by 20°C (e.g., from 25°C to 45°C) can slash energy consumption by 30% to 40%. 

Again, operators must be aware e of a potential efficiency trap; running at the absolute limit maximises savings but evaporates your thermal safety buffer. And in the volatile world of AI workloads, overshooting these temperatures means expensive GPUs will throttle or shut down, creating a catastrophic downtime risk that far outweighs energy savings.

Compaction: the secret weapon of Capex savings

While many fear the "liquid cooling premium," a full lifecycle analysis tells a different story altogether.

At standard densities, the cost of liquid cooling components is often fully offset by the elimination of expensive air-cooled chillers and Computer Room Air Handlers (CRAHs). But when you push into high-density territory, liquid cooling shifts from cost-neutral to a massive financial advantage.

Compaction, however, is the magical dealbreaker – concentrating more compute capacity into fewer racks and less floor space, operators can lower costs tied to building construction, lighting, fire suppression, and power distribution while also improving overall infrastructure efficiency. 

For instance, compressing your compute power into a 20 kW/rack liquid-cooled footprint yields a 10% total facility Capex savings compared to 10 kW/rack air cooling. At 40 kW/rack, that savings jumps to 14%. By drastically shrinking the physical footprint, the costs related to building shells, lighting, fire suppression, and power distribution decreases.

The benefits extend beyond cooling to the power chain itself. Replacing the power-hungry server fans with efficient micro-pumps, the overall IT load drops, allowing for an estimated $0.14 per watt savings on UPS and switchgear sizing. In the liquid-cooled data centre, smaller is actually better, and less costly, too.

The sustainability dividend and heat re-use

Liquid cooling is essential in a resource-constrained world. These systems can slash total cooling energy by up to 60% and offer a lifeline in water-stressed regions. 

Traditional air cooling consumes a staggering 43 litres of water per minute for every 1 MW of capacity, while liquid cooling, utilising warmer water and dry coolers, can achieve elite PUE ratings without draining local water supplies.

Because liquid cooling allows the use of warmer water temperatures, it can effectively utilise dry coolers with economisers, generating massive energy savings without the need for water-intensive adiabatic evaporation. 

Furthermore, rather than venting excess heat into the atmosphere, facilities can repurpose or sell it for district heating and industrial applications, creating new revenue opportunities while advancing carbon-reduction goals. As adoption increases, market analysts project the global data centre heat reuse-to-district-heating market will grow from $7.4 billion in 2025 to $21.6 billion by 2034.

As data centre operators look to lead in the AI era, the real advantage comes when liquid cooling is treated not as an upgrade, but as a financial strategy, unlocking a more efficient, higher-density, and future-ready foundation for growth.