Liquid cooling risks and solutions for AI data centers

Schneider Electric (blog author Kevin Brown) details the technical and operational risks of deploying direct liquid cooling at scale for AI data centers and positions its Motivair end-to-end portfolio as the recommended solution.

• Main announcement/action: Schneider Electric highlights the need for an end-to-end liquid cooling approach for AI data centers, citing concrete technical risks (corrosion, warranty/SLA complexity, unrealized energy savings) and promoting Motivair by Schneider Electric as a complete global portfolio including CDUs, RDHx, HDUs, dynamic cold plates, chillers, software and services; blog dated January 12, 2026, author Kevin Brown, contributor David McGlocklin (Cooling Development/Sustaining Engineering Manager, North America).
• Background and details: The post provides technical context and quantification: GPU power rising from ~150 W (CPU era) to 1000–1400 W per chip (NVIDIA Blackwell), rack densities from 10–20 kW historically to 142 kW per rack today and notes “1 MW per rack is on the horizon”; discusses fluid types (DI water, PG25 propylene glycol), material compatibility/corrosion, the need to meet temperature/pressure/flow specs across vendors, references White Paper SPD_WP210_EN for DLC system challenges, and quantifies a chiller-temperature trade-off (~2–2.5% electrical efficiency gain per 1°C raised).