Greenfield AI factories built from scratch can incorporate liquid cooling as a foundational design assumption. Existing data centers, built for an earlier generation of lower-density air-cooled workloads, face a different and often more constrained challenge: how to retrofit liquid cooling capability into a building never designed to support it, without disrupting tenants already operating within that facility.
Rear-Door Heat Exchangers as a Transition Path
For facilities not ready for a full liquid cooling overhaul, rear-door heat exchangers offer a meaningful intermediate step — mounting a liquid-to-air coil on the back of existing racks, capturing exhaust heat before it enters the hot aisle, and converting a purely air-cooled environment into a hybrid system capable of supporting moderately higher densities than air cooling alone, generally in the range of 40 to 70 kW per rack depending on configuration. This approach requires considerably less structural and piping infrastructure change than a full direct-to-chip retrofit, making it an attractive option for facilities seeking a meaningful density upgrade without a full liquid cooling build-out.
The Real Constraints on Full Retrofit
- Raised floor or slab structural capacity, designed for an earlier generation of lighter equipment, may not support the weight of modern liquid-cooled racks and the piping infrastructure required to serve them
- Routing primary and secondary coolant loops through an occupied, operational facility is considerably more disruptive than installing the same infrastructure in an empty building under construction
- Existing facilities often lack the floor-to-ceiling height, yard space for dry coolers, or electrical capacity headroom that a purpose-built liquid-cooled facility would incorporate from the outset
A retrofit is not simply a new-build project compressed into an existing building — the existing building's own constraints become a design input as significant as the cooling technology itself.
Phased Retrofit Strategies Reduce Disruption
Many operators pursuing liquid cooling retrofits adopt a phased approach — converting specific zones or halls of a larger facility to liquid cooling capability while leaving other areas air-cooled, rather than attempting a disruptive, facility-wide conversion in a single phase. This allows liquid-cooled capacity to come online for AI tenants while air-cooled zones continue serving existing lower-density tenants without disruption, and it can spread the capital and operational disruption of the retrofit over a more manageable timeline.
When Retrofit Genuinely Makes Sense Versus When It Does Not
Retrofit is most attractive when a facility's underlying structural, electrical, and spatial characteristics can genuinely accommodate the upgrade with reasonable capital investment, and when the facility's location and existing tenant relationships carry value that would be lost by decommissioning rather than upgrading. In cases where structural or spatial constraints are severe enough that a genuine retrofit would approach the cost of new construction, building new purpose-designed capacity, even at a different site, can be the more capital-efficient path — a conclusion that requires honest technical assessment rather than an assumption that retrofit is always the lower-cost option simply because a building already exists.
Assessing Retrofit Potential Rigorously
DATAPERT supports clients in rigorously assessing whether a liquid cooling retrofit genuinely makes sense for a specific facility, as part of our technical advisory and broader data center development capabilities. Start a project to evaluate retrofit potential for an existing facility.
