Microsoft unveils a data center design that uses zero water for cooling, down from the 125M+ liters per year a typical center uses, rolling out from August 2024
New design will use no water to cool the facilities' servers and chips. — Microsoft Corp., trying to mitigate the climate impact …
Context & Ripple Effects
Microsoft had already set a 95% data-center water-reduction goal and was testing nonconducting-liquid server cooling with startups. This design marks a progression from efficiency targets and experimental cooling approaches to a facility-level specification.
The shift makes cooling architecture a more central part of Microsoft's data-center buildout decisions, rather than solely an operational-efficiency measure.
First-order effects
- Microsoft's facilities using the new design no longer require water for server and chip cooling, avoiding the annual cooling-water use cited for a typical center.
- Data-center engineering and procurement move toward the equipment and operating processes needed to support the new cooling design as rollout begins.
Second-order effects
- A zero-water option could improve Microsoft's ability to develop capacity where water use is a binding local concern, while increasing pressure on rival operators to show comparable cooling efficiency.
- Cooling-system vendors and data-center builders face demand for designs that reduce or remove water dependence, not just incremental improvements in water use.
Third-order effects
- If deployed broadly, water intensity may become a more consequential differentiator in data-center siting and design, alongside power availability and build economics.
- The pattern points toward AI infrastructure being evaluated as utility-scale infrastructure: operators will need to manage local resource constraints as part of capacity expansion.
The trend: Hyperscalers are redesigning data-center infrastructure around local resource constraints, with cooling becoming as strategic as powered capacity.