When the conversation about data centres turns to environmental impact, energy and carbon take the headline and water is treated as an afterthought. That is the wrong way around for several reasons, the most consequential of which is that the water a data centre consumes through its electricity supply usually dwarfs the water it consumes through its cooling systems. The total is materially larger than the WUE metric on the site dashboard tends to suggest.

Direct and indirect water

The total water footprint of a data centre breaks down into two categories.

Direct consumption is what runs through the cooling systems on site. This is what the Water Usage Effectiveness (WUE) metric captures, and where most of the recent innovation has happened. Direct liquid cooling, evaporative cooling and advanced heat management have all materially reduced the direct number per unit of IT load.

Indirect consumption is what gets used upstream to produce the electricity the data centre draws from the grid, plus the water needed to treat the wastewater the site produces. This is the larger category, and it is almost never reported. Recent analysis of US data centres put the 2018 operational water footprint at 513 million cubic metres, with roughly three-quarters of that volume sitting in indirect consumption rather than direct.

If you are reading a sustainability report that quotes a single WUE figure as the data centre's water story, you are reading about a quarter of the actual footprint.

Why electricity production uses so much water

Most large-scale electricity generation methods are water-intensive in some way. Mekonnen's analysis of water footprints across energy sources puts the most water-intensive in the following order:

  • Hydropower: second only to firewood in water footprint per unit of energy produced (300 to 850,000 m³ per terajoule-equivalent). Reservoirs alter river ecosystems and flow patterns; the impact is structural as well as volumetric.
  • Fossil fuels (coal, gas, oil): vast cooling water requirements. Water is withdrawn from rivers, lakes or oceans and released back warmer, with downstream impacts on aquatic ecosystems.
  • Nuclear: heavy cooling water requirements for the same physical reason. Fission heat needs continuous cooling, with thermal pollution effects on receiving water bodies.
  • Concentrated solar power (CSP): often overlooked as low-water because solar is generally seen as clean. CSP relies on the same cooling-cycle physics as fossil fuels and nuclear, and lands in the same water-footprint order of magnitude.

Photovoltaic solar, wind and geothermal all use materially less water per unit of electricity than the above. Not zero, and not without other lifecycle impacts, but distinctly lower.

Location matters

Because most data centres draw from the local grid, the geography of the data centre determines the indirect water footprint of every workload running on it. A facility in a region with hydropower-heavy generation (much of Africa, Latin America and the Caribbean) carries an above-average indirect water footprint regardless of how efficient its own cooling systems are. A facility in a region with high renewables penetration and low hydropower share carries a much lower one.

Finland is a useful counterexample to the common assumption that low carbon equals low water. Finland's grid is heavy on biofuels and waste (around 52% of generation) and nuclear (around 34%), producing materially lower emissions than most European grids and a materially higher water footprint. A data centre in Helsinki and a data centre in Madrid can run identical workloads and report very different stories on these two dimensions, in opposite directions.

This matters more every year. EU monitoring put water scarcity as affecting roughly 29% of EU territory in 2019, and the trend has continued in the wrong direction. A location chosen for cheap clean electricity can sit inside a region under increasing water stress; the two questions are not the same.

None of this is a question a customer can answer by reading the data centre's WUE. It requires the grid mix of the region the data centre is in, alongside the water-stress profile of that region.

And not just water: biodiversity and land

Electricity production has impacts beyond emissions and water, on biodiversity and land use. The headline impacts sit on the fossil fuel side, where extraction and mining cause significant habitat destruction, land degradation, and ecosystem disruption, but renewables are not impact-free. Large-scale solar farms, wind installations and hydropower dams all affect wildlife and ecosystems, albeit to a lesser extent than fossil fuels.

The Ivanpah Solar Electric Generating System in California, one of the world's largest solar power plants, illustrates the point. Construction cleared large areas of desert habitat, displacing native plant and animal species. "Renewable" is the right answer on emissions; it is not always the right answer on biodiversity or land use, depending on where and how the installation goes in.

For a procurement function reading a sustainability report, the practical implication is the same one that ran through the water section above: the environmental footprint of a data centre's electricity supply is multi-dimensional, and the dimensions do not always move in the same direction. A grid that scores well on carbon may score poorly on water or on land use. Knowing which dimension matters most for the specific organisation's reporting and regulatory exposure is the precondition for choosing well.

What we measure

The Interact environmental report calculates the water footprint of the electricity consumed by a customer's estate, derived from the grid mix of the data centre's geographic location. Direct cooling water is captured separately. The total picture, direct plus indirect, is what gets reported.

For a customer evaluating data centre options across multiple geographies, this matters commercially as well as environmentally. The water-stress profile of a location is increasingly something procurement teams are asked about by their own sustainability functions, by their lenders, and by regulators in jurisdictions with mandatory water-disclosure requirements. The data centre with the better WUE may not be the data centre with the smaller total water footprint.

What this means in practice

Three implications worth carrying away.

First, ask the question. If a data centre or hosting provider answers the water question only with a WUE number, they are answering a small part of it.

Second, look at the grid mix. Where the electricity comes from drives most of the answer. Geography is the single largest variable.

Third, measure both. A complete water footprint needs direct and indirect together. Half a picture is worse than no picture; it gives a procurement decision the false confidence of a complete one.

Carbon is the metric on the headline. Water is the one that turns out to be quietly comparable in scale, and quietly easier to mismeasure.