Water Use: How Ecolab Tackles AI Sustainability Concerns

A surge in AI is transforming the energy and cooling demands of data centres.

The introduction of advanced chips and increased workloads is vastly increasing electricity consumption, prompting providers to re-evaluate cooling systems to ensure competitiveness and sustainability.

Estimates indicate that by 2030, infrastructure supporting large language models may need 327 GW of electricity, roughly equating to 70% of all electricity used in the US in 2024.

With each generative AI query using about 2.9 watt-hours, 10 times that of a standard search engine query, efficient cooling in data centres is more vital than ever.

Addressing this critical issue is Ecolab, which offers advanced innovations and solutions in cooling technologies.

Emilio Tentuta, Chief Sustainability Officer at Ecolab, says: “What I value and love most about my role is helping companies unlock growth by managing resources more intelligently, especially water.

“It's incredibly rewarding to see how smart water resource management drives business growth and a broader impact in the world.”

Balancing resource use

While power remains a primary constraint on data centre expansion, water is equally crucial.

Rising cooling needs exert considerable pressure on local watersheds depending on their locations.

Therefore, conserving both power and water resources is vital for operators aiming for efficiency and durability.

“When we think about data centres, one is opening every week and one semiconductor fab every month as well,” says Christophe Beck, CEO of Ecolab, speaking to CNBC.

“They all require a lot of water. We estimate that by 2030, we will need the incremental power of the whole of India, and the drinking water needs of the whole of the United States.

“The technology we’re bringing enables data centres to power even more AI in the future, and the same applies for microelectronics."

Alternative solutions like waterless cooling have been explored, but without precise analysis, they can unintentionally raise the hydro footprint.

Energy generation itself is water-intensive, with natural gas needing between 570 and 1,100 litres to produce a single megawatt-hour.

Water, however, retains inherent benefits over air cooling.

Data centres using liquid cooling typically use 10–30% less energy than air-cooled systems due to water’s superior heat-dissipating abilities.

Therefore, water-cooled systems prove efficient and strategically advantageous when strategically applied.

Christophe tells CNBC: “It’s sustainable because the new technologies use little to no water. In the past, data centres were cooled with air conditioning in large rooms. Tomorrow, it’s going to be direct-to-chip cooling. There’s no water involved. It’s just a liquid that flows directly to the chip, with much greater AI power as a result.”

Crafting resilient cooling designs

Site selection is critical for sustainable cooling.

Assessing factors such as climate, power sources, water accessibility and watershed impact can guide the most suitable solutions.

Designs must accommodate current needs and be adaptable to future changes.

This requires integrating diverse cooling strategies from site infrastructure to chip level.

Direct-to-chip cooling, for instance, can lessen the load on facilities’ chillers and towers and by redistributing cooling strains, data centres can decrease both power and water consumption.

Digital systems continually evaluate performance, pinpointing inefficiencies and enabling proactive adjustments which ensures sustained efficiency and prevents unforeseen downtimes.

A mid-sized 100-megawatt data centre with optimised use of water can consume less annually than an 18-hole golf course.

However, unlike golf courses, data centre water use fluctuates hourly with variables like cooling load, water quality, ambient conditions and equipment status.

Leveraging technology for efficiency

With rising demand, no single developer can assess every cooling decision from design to maintenance comprehensively.

Therefore, collaborating with expert suppliers becomes essential to bridge this gap.

Suppliers with comprehensive oversight use performance data for decision-making, anticipate hurdles and mitigate risks.

Digital monitoring plays a crucial role in this ecosystem, identifying irregularities early and equipping operators to optimise system performance.

These strategies culminate in more efficient water and power usage, supporting environmental goals alongside financial benefits.

“For us, it has always been a question of performance,” Christophe explains to CNBC. “It’s about helping businesses and industries make more money and invest more in the future. 

“If our solutions help data centres operate with little or no water, that’s good for the tech companies, good for consumers and good for the environment.

"In fact, most data centres in the future won’t use more water than a single car wash in a local community. It’s a good deal for everyone.”

Integrating sustainability into core strategies

For cooling systems to be sustainable, efficiency should be embedded at each development phase.

This begins before construction, aligning site choices with grid capacity, climate conditions and community needs.

During the design phase, methods like adiabatic and direct-to-chip liquid cooling remarkably reduce energy consumption.

Once operational, projects focused on water reclamation and reuse further minimise reliance on potable resources.

Incorporating digital monitoring guarantees these strategies remain dynamic, evolving with workload demands to ensure competitiveness and conserve finite resources.

Ecolab proposes that the most sustainable strategy acknowledges the interplay of power, water and community interests.

Tailoring cooling methods to local realities enables data centres to maximise efficiency and maintain resilience amid escalating demand.