Why HVAC Strategy Is the Critical Decision
For data centre fit-outs, HVAC represents the largest single capital cost and the primary determinant of operational efficiency. Get it wrong and you'll overspend on capital, underperform on PUE, and potentially compromise the N+1 resilience your uptime commitments demand.
This guide is for data centre owners, project managers and facilities directors navigating their first — or second — fit-out.
Understanding PUE
Power Usage Effectiveness (PUE) is the ratio of total energy consumed by the facility to the energy consumed by the IT equipment. A PUE of 1.0 is theoretical perfection; a PUE above 2.0 indicates significant inefficiency.
For a modern fit-out, a PUE target of 1.4 or below is achievable with the right cooling strategy. Every 0.1 reduction in PUE represents meaningful operational cost savings at scale.
The Main Cooling Strategies
1. Traditional CRAC Unit Cooling
Computer Room Air Conditioning units have been the default for decades. They're well-understood, but increasingly over-specified for modern high-density IT loads. Capital cost is high; PUE performance is typically poor (1.7–2.2 range).
2. In-Row Cooling
In-row cooling units are positioned within server aisles, delivering cool air directly at source. They're more efficient than traditional CRAC for high-density loads and support lower PUE targets.
3. Adiabatic Free-Cooling
For UK climates (where ambient temperatures allow free-cooling for a large proportion of the year), adiabatic systems can dramatically reduce mechanical cooling requirements and achieve PUE targets of 1.3–1.5.
4. Liquid Cooling
For extreme density deployments, direct liquid cooling or rear-door heat exchangers may be required. Capital cost is higher; PUE performance can be exceptional (sub-1.2).
N+1 Resilience: What It Actually Means
N+1 resilience means that for every N components required to operate the system, one additional component exists as backup. For HVAC, this means a single cooling unit failure must not compromise the IT load.
This has direct implications for system design — particularly the number of units, their individual capacity relative to the total load, and the controls strategy for automatic failover.
The Most Common Mistakes
- Over-specifying for peak load without considering phased IT load growth — results in capital waste and poor part-load efficiency
- Assuming CRAC units are the only option — frequently leads to missing the PUE target
- Ignoring structural constraints on plant room layout until detailed design — costly late-stage changes
- Not modelling free-cooling hours for the specific site location
How B&W INFRA Can Help
We provide independent HVAC design reviews and value engineering for data centre fit-outs. Our recent project for a technology firm achieved a PUE of 1.38 while reducing capital costs by £480K versus the original design proposal.
Get in touch to discuss your project.