Two Numbers, One Ledger: Water and Power Effectiveness

The data-center industry runs on two headline efficiency numbers, and the whole water-energy story turns on understanding how they pull against each other. The famous one is power use effectiveness, the ratio of total facility energy to the energy that actually reaches the computing equipment. A perfect building would score 1.0, spending nothing on overhead; real facilities score somewhat above that, and every fraction shaved off has been treated as a badge of honor for more than a decade. The less famous, younger number is water use effectiveness, the volume of water a facility consumes per unit of computing energy delivered, usually stated in liters per kilowatt-hour. One measures wasted watts; the other measures spent water. Read together they describe a cooling strategy honestly. Read apart they invite exactly the self-deception this module warns against.

Here is the trap, stated as plainly as possible. The single cheapest way to drive power overhead down is to cool with evaporation, because letting water carry off heat spares the compressors that would otherwise dominate the overhead. So an operator optimizing power effectiveness alone is quietly pushed toward the thirstiest possible design. The building posts a gorgeous power number and a water number it may not even publish. An observer watching only the famous metric sees an efficiency champion; the aquifer sees a heavy customer. The two numbers are not independent virtues to be maximized side by side. They are ends of a seesaw, and pressing one down often lifts the other up.

That coupling is why a single number, however impressive, tells you almost nothing about whether an operator has made a good tradeoff. A facility with a strong power figure and a heavy water figure has chosen to spend water. A facility with a light water figure and a mediocre power figure has chosen to spend electricity, and probably carbon. Neither is automatically better; the right choice depends entirely on where the building sits, which is the subject of the next module. But you can only even see the choice if both numbers are on the table.

Water effectiveness carries a hidden subtlety that power effectiveness does not, and it trips up careful readers. There is the water a facility runs through its own cooling plant, and then there is the water consumed far away at the power station that generates the facility's electricity. Thermal power plants, whether coal, gas, or nuclear, often cool themselves with enormous quantities of water, so every kilowatt-hour a data center draws carries an invisible upstream water cost. This is the source-versus-site distinction. A building that goes fully dry to post a beautiful on-site water number may, by burning more electricity, have increased the water consumed at the power plant that feeds it. The water did not disappear; it moved upstream, out of the facility report and onto the grid. An honest accounting follows it there.

This upstream water is the quiet revenge of the water-energy tradeoff on anyone who thinks they can escape it. Spend water on site and it shows up in your own report. Spend electricity to avoid that water and much of it reappears as water spent at a distant, thirsty power plant, plus carbon on top. The only way to know whether a dry design genuinely saved water or merely relocated it is to account for both the site figure and the source figure, and precious few public reports do. When one does, it is a strong signal that the operator actually understands its own footprint rather than gaming a single dashboard.

The practical upshot for anyone evaluating a facility or a portfolio is a short set of questions. Are both effectiveness numbers disclosed, or only the flattering one? Is the water figure a site number, a source number, or both? Over what period and season is each measured, since a desert campus averaged over a mild spring hides its brutal summer draw? And what water source feeds the plant, potable, reclaimed, or non-potable, since a large withdrawal of reclaimed wastewater is a very different thing from the same volume of drinking water. An operator that answers these crisply is telling you it keeps an honest ledger. One that offers a single number and changes the subject is telling you something too.

The mental model to carry out of this lesson is the ledger with two columns that must always be read together. Watts on one side, water on the other, with a footnote reminding you that the columns leak into each other, water hiding inside the watts and watts hiding inside the water. Efficiency is not a single triumphant figure to be minimized. It is a balance to be struck and, crucially, to be disclosed on both sides, so that the person reading it can judge whether the tradeoff made sense for the place where the building actually stands.