Cooling without water
Designs that consume almost none exist and cost more. The industry is discovering which regulators will make it pay the difference.
Ricardo Gomez Angel via UnsplashThe technical part of the data centre water problem is largely solved. Designs exist that consume almost no water directly, and analysis of the question describes them as available rather than speculative, at a modest cost premium.
That framing matters because it moves the argument. Policy analysis describing the water problem as soluble with current technology reframes the issue from a technical constraint into a question about who pays for the better option.
The premium is real and it is mostly electricity. Rejecting heat without evaporation means moving it with mechanical cooling, and that consumes power on a site that is already competing for a grid connection.
So the choice is rarely between good and bad. It is between a facility that consumes water in a place where water is contested and a facility that consumes more electricity in a place where the interconnection queue runs to years.
Operators default to the cheaper option in the absence of a rule, which is the entirely predictable behaviour of a competitive industry with thin margins on capacity and a customer that shops on price.
That is why disclosure regimes matter more than they appear to. Once per-facility water use is published, the cheaper design carries a reputational cost in the community hosting it, and that cost is felt in the next permit.
Retrofitting is the harder half and it gets discussed least. Changing the cooling architecture of a running facility means taking capacity offline, which is expensive in a market where capacity is the scarce good.
New builds are where the shift will show first, and the siting decision is where it is actually made. A design chosen for a location with abundant water is a commitment for the life of the building.
Heat reuse remains the genuinely interesting idea and remains rare, because it requires a neighbour who wants low-grade heat and a district system to move it, which is a planning question rather than an engineering one.
The realistic near-term outcome is regional divergence. Water-constrained jurisdictions will require the more expensive design, others will not, and the industry will site accordingly rather than standardise.