Retrofitting legacy data centers for higher density

Many data centers in Canada were designed for racks of a few kilowatts, with perimeter CRAC or CRAH units and a raised floor. Today the same owners are being asked to host equipment that draws several times as much power per rack, including accelerators for AI workloads. Building a new facility is not always possible or necessary. In many cases, a legacy room can be upgraded to support higher density, at least in part.

Retrofitting a live data center is harder than building new. Work must happen without disrupting operations, existing infrastructure constrains choices, and limits in power, cooling, structure and space all interact. This guide outlines a practical approach.

Step 1: assess what you have

Start with an honest assessment of the existing facility.

Cooling

• Installed capacity and actual load on each cooling unit
• Redundancy, and whether it still holds at current load
• Chilled water plant capacity, temperatures and spare connections
• Airflow management condition: blanking, cutouts, containment
• Rack inlet temperatures and hot spots
• Age and condition of equipment

Power

• Utility service, UPS and generator capacity
• Distribution to rows and racks, including PDU and branch circuit capacity
• Space for additional electrical equipment

Structure and space

• Raised floor load rating and condition
• Structural slab capacity for heavy racks or immersion tanks
• Ceiling height and overhead space for piping, busway and containment
• Plant room and roof space for added heat rejection

Operations

• Change windows and tolerance for disruption
• Redundancy needed during the work
• Staff skills for new technologies such as liquid cooling

A CFD model, validated with field measurements, is a valuable part of this assessment. It shows where capacity exists and where problems will appear as density rises.

Step 2: capture quick wins

Before adding equipment, recover the capacity already installed. Many legacy rooms have far more cooling than they use, wasted through bypass and recirculation.

1. Install blanking panels and seal cable cutouts.
2. Remove abandoned cables from the underfloor plenum.
3. Relocate perforated tiles to match load.
4. Add hot or cold aisle containment.
5. Convert constant-speed fans to variable speed with EC fan retrofits or drives.
6. Centralize humidity control and widen deadbands.
7. Raise supply temperatures within the ASHRAE TC 9.9 recommended envelope.

These measures often free significant capacity and improve efficiency. See airflow management and hot and cold aisle containment.

Step 3: choose a density strategy

There are three broad strategies for adding density.

The high-density zone is often the most practical. It concentrates investment, avoids disturbing the whole room and can be built in phases.

Step 4: select cooling technologies

Close-coupled air cooling

In-row coolers and rear-door heat exchangers add capacity at the rack without changing the whole room. Rear doors in particular can be installed rack by rack with minimal disruption, provided chilled water can be routed to each position.

Direct-to-chip liquid cooling

For racks of accelerators, direct-to-chip liquid cooling is often the only practical option. A retrofit needs:

• Facility water extended to the rows, usually through a new secondary header
• CDUs, either in-row or centralized
• Leak detection and drip protection
• A plan for the residual air-cooled heat
• Water temperatures compatible with the IT vendor's requirements

Liquid-to-air CDUs as a bridge

Where facility water cannot reach the row quickly, liquid-to-air CDUs let cold plate servers run with heat rejected to the room. This adds air-side load and is usually an interim step.

Step 5: plant and distribution

Higher density raises total heat load and may require changes to the plant:

• Additional chiller or heat rejection capacity
• Warmer water temperatures for close-coupled and liquid systems, possibly with a separate warm-water loop
• Waterside economizer or dry coolers to take advantage of warm water and Canadian winters
• Thermal storage for ride-through if densities make temperature rise during outages faster

Piping retrofits in a live room require careful planning. Hot taps, line stops and freeze plugs can connect new piping to live systems without shutdown, but each carries risk and should be planned with the operations team.

Step 6: power and structure

Every additional kilowatt of IT load needs both power and cooling. Electrical upgrades often take longer than mechanical ones because of equipment lead times and utility coordination. Verify floor load capacity before installing heavy racks; liquid-cooled racks and immersion tanks can exceed legacy floor ratings. A structural engineer should review point loads.

Step 7: phase the work

Live retrofits are usually done in phases:

1. Prepare: install piping, power and supports during maintenance windows.
2. Build the new zone: install containment, cooling and power for the first pod.
3. Commission the zone with load banks before moving IT equipment.
4. Migrate equipment into the new zone.
5. Repeat for the next zone.

Maintain redundancy throughout. Temporary cooling may be required during plant changes. Each phase should be commissioned before it carries production load.

Common pitfalls

• Planning cooling without matching power, or the reverse
• Forgetting the residual air load of liquid-cooled racks
• Underestimating floor loads
• Running piping over critical equipment without drip protection
• Skipping commissioning of new zones because of schedule pressure
• Letting redundancy erode as load grows

Frequently asked questions

Should I retrofit or build new?

It depends on the scale of density needed, the remaining life of the existing infrastructure, available power and space, and timelines. A retrofit with a high-density zone often meets near-term needs at lower cost.

Can a raised-floor room support liquid cooling?

Yes. The underfloor plenum is often a good route for facility water piping to rows, provided it is coordinated with airflow and cabling and protected with leak detection.

Key takeaways

• Start with a full assessment of cooling, power, structure and operations.
• Recover existing capacity with airflow management before adding equipment.
• High-density zones with close-coupled or liquid cooling are often the most practical path.
• Plan plant, piping, power and structural upgrades together.
• Phase the work, keep redundancy throughout and commission each phase under load.