Liquid Cooling
Definition
Liquid cooling uses liquid-based systems, such as direct-to-chip cooling or immersion cooling, to remove heat from servers, GPUs, and accelerators. It is increasingly important for AI infrastructure because dense GPU clusters can exceed the practical limits of traditional air cooling.
Why it matters
Cooling capability can determine whether a data center is AI-ready. Liquid cooling can support higher rack density and better efficiency, but it introduces capex, maintenance complexity, retrofit risk, water or fluid management, and operational skill requirements. Investors should treat cooling design as a core underwriting variable, not a technical footnote.
Common misconceptions
- •Liquid cooling is not one technology; direct-to-chip and immersion systems have different economics and risks.
- •Liquid cooling does not eliminate power constraints.
- •A facility advertised as AI-ready may still require major cooling retrofits for next-generation GPUs.
Technical details
Common approaches
Direct-to-chip cooling moves liquid through cold plates attached to processors or accelerators. Immersion cooling submerges hardware in dielectric fluid.
Hybrid designs may combine air cooling for some components with liquid cooling for high-heat chips.
Economic impact
Liquid cooling can improve PUE and support higher density, potentially increasing revenue per square foot or per megawatt.
Offsetting costs include cooling distribution units, plumbing, facility retrofits, leak management, maintenance, and potential hardware compatibility limitations.
Diligence questions
Which cooling system is installed or planned, and what densities does it support?
Is the facility retrofitted or purpose-built for liquid cooling?
Who bears retrofit cost, downtime risk, and maintenance responsibility under tenant contracts?