🤖 AI may live in the cloud, but its infrastructure is decidedly physical, and increasingly resource-hungry. Behind every model is a network of data centers packed with high-performance chips that consume vast amounts of electricity and generate equally significant amounts of heat. Keeping those systems running means drawing on another increasingly scarce resource: water.

The environmental footprint is therefore becoming harder to measure through carbon emissions alone. Global data center electricity consumption could reach 945 TWh by 2030, up from an estimated 448 TWh in 2025, a June 2026 report (pdf) from the United Nations University Institute for Water, Environment, and Health (UNU-INWEH) estimates. The associated water footprint could reach 9.3 tn liters, while the land footprint could exceed 14.5k sq km, and that’s just the beginning.

Not all data centers are created equal

The facilities drawing the most scrutiny are hyperscale data centers and AI-optimized sites running dense clusters of GPUs. Unlike conventional data centers, which may handle a broad mix of workloads, AI facilities can run intensive computing operations continuously, generating high heat densities that conventional cooling systems increasingly struggle to manage. And the environmental burden does not end when an AI model has been trained.

The day-to-day process of running AI models — known as inference — can become a major source of consumption as more people and businesses use generative AI for text, images, video, and other applications. The more efficient AI becomes, the cheaper and easier it can become to use, creating a rebound effect in which lower resource consumption per task is offset by a much larger volume of tasks. The UNU-INWEH report warns that efficiency gains can therefore be absorbed by growing demand.

Then there’s the water problem: Data centers need cooling because the electricity flowing through high-performance chips is converted into heat. Traditional cooling systems can rely heavily on water evaporation, creating a direct water footprint at the facility. There’s also an indirect footprint: electricity generation itself can require substantial water, particularly when power comes from thermal generation.

That makes the environmental cost highly dependent on where a data center is built and how its electricity and cooling systems are supplied. A facility powered by low-carbon electricity may have a smaller carbon footprint while still placing considerable pressure on local water supplies. The UNU-INWEH research found that carbon, water, and land footprints can move in opposite directions, meaning a cleaner power source is not automatically a low-water or low-land option.

Where does the Middle East — and Egypt — fit into this?

The region is attracting growing interest in digital infrastructure and AI while also dealing with extreme heat and some of the world’s most severe water constraints. Higher temperatures increase cooling requirements, just as electricity demand for air conditioning rises across cities. Egypt is entering this equation with ambitions of its own. The government is working to position the country as a regional hub for data movement and digital infrastructure, leveraging its geographic location and extensive network of international submarine cables.

A national data center strategy is being developed around the expansion of digital and energy infrastructure, with the government highlighting long-term electricity supply arrangements and renewable power as part of its plans. Egypt is targeting a 45% share of renewable energy in total electricity generation by 2028, while the government is also expanding battery energy storage systems (BESS) to help stabilize the grid as solar and wind capacity grows. A project being developed with Huawei Egypt includes plans for standalone BESS plants with a combined capacity of 4k MWh.

That matters because AI data centers need 24/7 reliable power, while solar and wind generation are intermittent. Storage can help bridge that gap, allowing excess renewable electricity to be stored and dispatched when demand rises. But cleaner electricity does not, on its own, resolve the cooling problem.

Egypt doesn’t have water to spare

For Egypt, water may be the harder constraint. The country faces an annual water deficit of around 7 bcm, according to Unicef (pdf). That puts the water requirements of large-scale data center infrastructure into a very different context from countries with more abundant freshwater resources. Cooling technology will therefore be critical. Liquid cooling — including direct-to-chip systems — can transfer heat more efficiently than air and is increasingly being considered necessary for high-density AI workloads. Liquid cooling can also offer heat-transfer performance 20-30x higher than air-based systems, and Egyptian industry experts have called for data center sites to be planned alongside dedicated energy and water infrastructure.

What does responsible expansion look like?

Expansion will increasingly depend on treating water, energy, and land as interconnected constraints rather than separate environmental concerns. That means the location and design of data centers could become just as important as their access to electricity, with approaches such as closed-loop cooling and the use of recycled or non-potable water helping reduce pressure on local supplies. In water-stressed regions, availability may ultimately prove as important a consideration as grid capacity when determining where new facilities can be built.

The same logic applies to energy. Pairing renewable generation with storage and stronger grid infrastructure could help data centers manage their demand, while shifting workloads toward facilities or regions when renewable power is most abundant could ease pressure during peak periods. That broader approach is also beginning to shape how the industry measures its footprint: the United Nations University is calling for AI infrastructure to account for carbon, water, and land footprints together, while the UN has launched an initiative calling on major AI companies to disclose the environmental impacts of their systems.

The question for Egypt isn’t whether it should participate in the AI data center boom. It’s whether the country can build the infrastructure needed to capture its economic benefits without shifting the environmental cost onto resources — particularly water — that are already under pressure. Digital expansion may happen in the cloud, but its consequences will be felt firmly on the ground.