Water infrastructure
CoralDC's deployed systems use closed-loop liquid cooling, which substantially limits operational water consumption by design. This page covers what the company is researching beyond that, and why it does not describe its infrastructure as water-positive. Water consumption is a consequence of the cooling architecture rather than an independent variable that can be optimised later, which is why it is decided early.
The constraint
Evaporative cooling, the dominant approach in conventional large-scale facilities, consumes water continuously. As AI infrastructure scales into water-stressed regions and into jurisdictions where water permitting is politically contested, water consumption becomes a siting constraint and in some cases a permitting veto.
Why it matters for AI infrastructure
Water consumption is not an independent variable that can be optimised later. It is a consequence of the cooling architecture, which is a consequence of the density target, which is a consequence of the available power envelope. A system designed around closed-loop cooling and dry heat rejection has a fundamentally different water profile from one designed around evaporative rejection. Water is a design decision made early.
For scale: evaporatively cooled facilities have been characterised at approximately 2.6 million US gallons per megawatt per year, against near zero for closed-loop reference designs.
What CoralDC is studying
Site water balance methodology for pod-scale deployments across varied climate zones. Whether dry and dew-point heat rejection holds across a full annual cycle in each target market rather than at design conditions. Whether low-grade rejected heat can drive useful water recovery at single-deployment thermal scale rather than campus scale. Whether recovered water has realisable value in the markets CoralDC's customers occupy, or whether the value is entirely in avoided consumption and reduced permitting risk.
What CoralDC is evaluating
Closed-loop and hybrid heat rejection, dew-point and indirect evaporative approaches for hot climates, cooling-water recovery and blowdown reduction, treated and reclaimed water as a heat rejection input, heat-driven water recovery, and zero-liquid-discharge approaches.
What remains uncertain
CoralDC does not claim water-positive infrastructure.
The phrase requires site-specific lifecycle accounting including embodied water, source water stress and seasonal availability. A claim made without that accounting is marketing rather than engineering. Heat-driven water recovery is thermodynamically credible and its economics depend heavily on humidity, ambient temperature and local water pricing. In temperate, water-abundant, low-tariff markets, which describes much of CoralDC's home geography, the economic case is weak even where the technical case holds.
Water-positive AI infrastructure is a legitimate long-term objective and a premature claim. CoralDC intends to be in a position to demonstrate it before saying it.
Where partnership could accelerate this
Water technology companies willing to be evaluated against real thermal outputs and real climate conditions rather than laboratory parameters. Water treatment specialists, particularly for deployments in water-constrained markets where treated or brackish water as a rejection input is the difference between a viable site and an unviable one.
Who should talk to us
Water technology and treatment companies. Atmospheric water generation developers. Municipal and industrial water authorities. University water and thermal engineering groups.
Bring us the problem, not the pitch.
CoralDC evaluates technologies continuously across its research domains. Submissions go to engineering, not to a marketing queue.
Direct answers for infrastructure buyers.
What is water-positive AI infrastructure?
It would describe infrastructure returning more water to its local environment than it consumes, requiring closed-loop cooling that avoids evaporative consumption, combined with water recovery, and site-specific lifecycle accounting covering source water stress and embodied water. It is an objective rather than an established standard, and CoralDC treats it as research rather than a claim.
How much water does a Coral deployment use?
A closed-loop deployment with dry heat rejection consumes effectively no water in operation. Whether dry rejection is viable year-round depends on the climate at the site and is confirmed during the architecture review.