RIYADH: Artificial intelligence has opened up remarkable new possibilities — but beneath its sleek, digital surface lies an overlooked environmental cost: Water.
While debates over AI’s energy demands have captured headlines, its vast and growing water footprint remains largely out of sight. That blind spot, experts warn, could carry serious consequences.
Salah Al-Kafrawi, senior consultant at EY for data and AI, told Arab News: “We have no universal approach to assess how much water is consumed while using or training AI,” leaving the true scale of the problem poorly understood.
Although a few tech firms publish rough water usage estimates, most businesses — from e-commerce to aviation — remain unaware of their consumption. “Many aren’t even aware of their water footprint,” Al-Kafrawi said
Even available data likely underestimates reality by a factor of 10 or more. In a world facing increasing water scarcity, AI’s mounting thirst is a silent crisis — one that demands innovation, transparency and smarter systems.
AI’s water usage is complex. It includes both direct consumption, such as cooling servers, and indirect usage tied to the electricity that powers them.
“AI requires significant data for training and evaluation, along with electricity to operate and cooling systems to prevent overheating,” Al-Kafrawi said.
That electricity often comes from water-intensive power sources. Coal-fired and nuclear plants, for instance, consume “millions of gallons of water daily” for steam production and cooling — a hidden environmental cost rarely accounted for in sustainability reports.
Meanwhile, data centers use vast amounts of water to stop their servers from overheating.
An artist's rendition of the data center of KAUST. (KAUST photo)
“Data centers often use water cooling towers to dissipate heat generated by their servers,” said Al-Kafrawi, noting this creates “another significant source of water usage.”
This leads to a vicious cycle: AI requires power, power generates heat, and cooling that heat requires more water.
“The combination of water needed for electricity generation and cooling systems means that AI’s water footprint extends far beyond what might be immediately apparent,” Al-Kafrawi said.
In Saudi Arabia, where water is scarce and AI infrastructure is booming, managing this balance is becoming increasingly urgent.
Abdulelah Al-Shehri, assistant professor of chemical engineering at King Saud University, said that most global data centers rely on potable water — despite its scarcity.
“Water purity directly affects a system’s lifespan,” he said. Indeed, reclaimed water can risk corrosion and microbial contamination, though efforts to safely reuse non-potable sources are gaining traction.
Most global data centers rely on potable water, putting pressure on scarce supply. (AFP photo)
Currently, most data centers in the Kingdom use hybrid cooling systems.
“Saudi data centers rely on high-efficiency mechanical cooling systems combining air-cooled and water-cooled chillers,” said Al-Shehri, referencing Microsoft’s climate-adaptive guidelines and the Saudi Telecom Company’s 2023 Sustainability Report.
Even these efficient systems come with a significant toll. Al-Shehri said that cooling infrastructure for the country’s 300 MW of operational data centers consumes between 2.3 and 2.8 liters of water per kilowatt-hour.
“If we look at current capacity operating at full load year-round, annual water use would approach 6.7 million cubic meters,” he warned — roughly the same amount used by 160,000 Saudi households each year.
And this is just the beginning. Al-Shehri projected that capacity could more than quadruple to 1,300 MW within five years — putting total water use on par with the needs of 700,000 households.
“These figures only capture direct water use for cooling,” he added. The indirect cost, tied to fossil fuel-based energy production, remains even greater.
Despite the growing demand, Al-Shehri said that there are promising solutions — but they require investment and vision.
“Multiple solutions are emerging to reduce water usage in data center cooling, though investors often hesitate because of the intensive capital investment and costly retrofits.”
Among the most promising ideas is rethinking traditional cooling standards. “Microsoft and Google have piloted these so-called ‘high-temperature data centers,’ which would effectively drive the direct water footprint to zero,” he said.
Raising operating temperatures from 21 degrees Celsius to as high as 35 Celsius allows systems to rely more on air than water.
Other technologies focus on recycling heat. “Absorption chillers can reclaim up to 40 percent of waste heat here and repurpose it for cooling,” said Al-Shehri, pointing to global examples such as Swiss company Infomaniak, which now uses server heat to warm 6,000 homes.
The Swiss company Infomaniak uses absorption chillers to reclaim waste heat, which it then sells to warm 6,000 homes. (X: @infomaniak_de)
There is also the energy mix to consider. “Diversifying energy sources for AI power isn’t a simple swap,” he said. “It hinges on resource availability, proximity to data centers, and the water footprint.”
Research shows that solar and wind are the most water-efficient sources. By contrast, biomass — a widely promoted renewable option — can consume “up to 100 times more water than natural gas,” Al-Shehri said.
Another overlooked cost comes from the treatment and transportation of water itself — a point highlighted by Abdullah Al-Otaibi, Saudi Arabia managing director at global water solutions firm Ecolab.
“Water must be moved, heated, cooled and treated to be fit for business use, which requires energy,” Al-Otaibi told Arab News. He framed water and energy as interconnected levers — what scientists call the “water-energy nexus.”
“Water serves a dual purpose in data centers. It cools infrastructure directly and gets consumed indirectly when generating the electricity that powers high-performance computing,” he said.
Failing to manage this interdependency is risky. “Overlooking water creates risk — particularly in regions like Saudi Arabia, where water is scarce, and digital infrastructure is expanding rapidly.”
DID YOU KNOW?
• AI systems consume large volumes of water through energy generation and data center cooling.
• High-temperature data centers and heat-recycling systems offer promising ways to reduce direct water use.
• Experts urge early-stage planning and water-smart designs to reduce risk and future-proof AI growth.
Al-Otaibi said that better data and AI tools can help companies to understand and reduce their water and energy footprints. “With the right data and technologies, businesses can make water use more visible — and more manageable,” he said.
Ecolab’s audits show significant gains are possible. Their tools have helped clients to cut water usage by 44 percent, energy by 22 percent and emissions by 12 percent, all while improving reliability.
“Water efficiency can become a business enabler — supporting uptime and sustainability targets at the same time,” Al-Otaibi said.
With AI infrastructure expanding rapidly across the Gulf, Al-Otaibi urged stakeholders to act now — especially in design stages. “The key is starting early — smart water strategies begin before construction,” he said.
He cited Ecolab’s partnership with global data center operator Digital Realty, where their AI-driven system is expected to reduce water use by up to 15 percent and prevent the withdrawal of 126 million gallons of potable water annually.
For Saudi Arabia, where water is precious and technology ambitions are vast, that kind of efficiency is not just smart — it is essential.
