Schneider says hotter coolant can make AI datacenters less thirsty

Schneider Electric's modeled 100MW datacenter analysis finds that shifting from conventional air cooling to 45°C liquid cooling can cut onsite water consumption by at least 50% in water-constrained locations such as Dallas. The report estimates cooling towers can use five to 20 times more water than dry coolers, while higher-temperature liquid cooling can also reduce equipment requirements and capital expenditure. The findings underscore rising water constraints for AI datacenter expansion, though Schneider has a commercial interest through its 75% stake in liquid-cooling specialist Motivair.

Analysis

This is less a near-term Schneider earnings catalyst than evidence that AI datacenter site selection is shifting from a power-availability problem to a combined power, water-permitting, and thermal-design problem. In water-stressed U.S. markets, the ability to specify higher-temperature liquid loops early can shorten entitlement timelines and reduce community opposition—potentially more valuable to developers than incremental PUE gains. That favors integrated electrical-and-thermal vendors with design-in influence, notably Schneider Electric (SU.PA/SBGSF), Vertiv (VRT), and private cooling specialists, while raising execution risk for developers whose land banks assumed conventional evaporative cooling.

The second-order effect is a mix shift rather than simply higher cooling spend. Liquid cooling raises content per MW for CDU, pumps, heat exchangers, controls, and deployment services, but reduced external equipment requirements can partially offset hardware volumes; suppliers with proprietary controls and service contracts should capture the best margin pool. VRT is the most liquid public proxy for AI thermal capex, but its valuation already embeds aggressive AI infrastructure growth. Schneider's claims should be treated as vendor-sponsored modeling until independently corroborated by actual customer water permits, project-level capex, and operating data.

Over 1-3 months, permitting disputes or water restrictions in Texas, Arizona, Virginia, and the UK would be a catalyst for a cooling-design premium and incremental VRT/SU.PA orders. Over 6-18 months, the key variable is whether GPU roadmaps sustain high rack densities and permit warm-water operation without compromising server reliability; if customers remain conservative on temperature set points, the projected water and capex advantages narrow. The supplied ticker SU is Suncor (SU), not Schneider Electric, so this article offers no direct fundamental read-through to SU.

Key Decisions for Investors

• Do not trade Suncor (SU) on this item; the ticker mapping is not economically relevant. Use Schneider Electric (SU.PA or SBGSF) and Vertiv (VRT) as the appropriate public-market proxies.
• Maintain a watch-list long bias in VRT versus a broad datacenter-infrastructure basket over the next 1-3 months, but wait for evidence of liquid-cooling backlog conversion or water-driven project redesigns. Entry trigger: disclosed cooling/order growth exceeding management's existing AI assumptions; invalidation: cooling backlog growth decelerates for two consecutive quarters or gross margin falls on project mix.
• For a lower-beta expression, accumulate SU.PA/SBGSF on market weakness rather than chase a single white paper. The upside is greater attachment of thermal controls and services to electrical distribution projects; risk/reward deteriorates if AI datacenter capex is delayed or customers standardize on lower-cost, non-integrated cooling architectures.
• Set regulatory alerts for water-use restrictions, permit denials, or mandatory water-disclosure rules in major hyperscale markets. Such events would strengthen the relative case for liquid/dry-cooling suppliers but could be negative for datacenter developers with unpermitted capacity, including Digital Realty (DLR) and Equinix (EQIX).