No single cooling method is best for every data center. The right choice depends on rack density, facility age, local climate, power usage effectiveness (PUE) targets, water availability, and capital budget. This comparison maps the major data center cooling systems available today across energy efficiency, water consumption, retrofit feasibility, and rack density thresholds, then provides a decision framework for matching the right method to specific facility conditions. The cooling landscape has changed substantially as AI and GPU workloads push rack densities beyond what traditional air cooling can physically manage. Understanding where each method succeeds, where it fails, and what infrastructure it requires is the starting point for any serious thermal management decision.
Key Takeaways
- Traditional air cooling using computer room air conditioning (CRAC) units and hot aisle/cold aisle containment becomes thermally inadequate at rack densities above 20 to 30 kW per rack, a threshold routinely exceeded by modern AI and GPU clusters, per Uptime Institute guidance.
- Immersion cooling achieves PUE values of approximately 1.02 to 1.03, compared to 1.4 to 1.6 for traditional air-cooled facilities, making it the most energy-efficient commercial option available.
- Evaporative and water-cooled systems trade lower PUE for higher Water Usage Effectiveness (WUE), a tradeoff with direct compliance implications in water-stressed regions and under EPA 316(b) requirements.
- ASHRAE thermal guidelines published by ASHRAE TC 9.9 define allowable rack inlet temperatures by facility class and govern whether liquid cooling can be retrofitted into a legacy air-cooled data center without full structural redesign.
- Adiabatic humidification expands the viable operating window for free cooling by maintaining the relative humidity (RH) levels at which evaporative pre-cooling becomes thermally effective, without introducing condensation or surface wetting risk under proper system design.
- Large data centers in the EU must report PUE and energy performance data to national authorities under Article 12 of the Energy Efficiency Directive (EED), making PUE benchmarking a regulatory obligation rather than a voluntary metric.
Why There Is No Single Best Cooling Method for Data Centers
The optimal cooling choice for any data center facility depends on at least six facility-specific variables: current and projected server rack density, facility age and retrofit feasibility, local climate and access to annual free cooling hours, target PUE and WUE, capital versus operating cost priorities, and regulatory reporting obligations. ASHRAE TC 9.9 and the Uptime Institute are the two primary standards bodies whose guidance defines the performance envelope for each of these variables. For large data centers operating in the European Union, the EED Article 12 reporting requirement has made PUE a compliance metric, not just an operational benchmark. The relevant decision variables for cooling selection are:
- Current and projected rack density in kW per rack
- Facility age and retrofit feasibility against ASHRAE TC 9.9 infrastructure readiness criteria
- Local climate zone and annual free cooling hours available
- PUE and WUE targets, including regulatory thresholds
- Capital expenditure versus long-term operating cost priorities
- Water availability and applicable watershed or EPA 316(b) compliance obligations
The Role of Rack Density in Cooling Selection
Rack density is the single most deterministic variable in cooling selection. Air cooling becomes physically inadequate at densities above 20 to 30 kW per rack, because the airflow volumes required to remove heat at that scale create noise, vibration, and thermal management complexity that airflow alone cannot resolve. Modern AI infrastructure, including NVIDIA H100 and H200 GPU clusters, routinely exceeds this threshold, which is why liquid cooling adoption has accelerated alongside AI workload deployment.
PUE and WUE as Decision Metrics
Power usage effectiveness (PUE) measures total facility energy consumption against IT equipment energy consumption. A PUE of 1.0 is theoretical perfection; most air-cooled legacy facilities operate between 1.4 and 1.6. Two metrics that define the tradeoff:
- PUE: total facility energy versus IT equipment energy.
- WUE: water consumed per kWh of IT load, the metric most directly affected by evaporative and water-cooled approaches.
These two metrics are frequently in tension: reducing PUE through evaporative or water-cooled strategies typically increases WUE, and the right balance depends on local energy costs, water costs, and applicable regulations.
Air Cooling: CRAC Units, Hot Aisle/Cold Aisle Containment, and Raised Floor Systems
Air cooling remains the baseline technology for the majority of operating data centers worldwide. Computer room air conditioning (CRAC) units pull hot exhaust air from hot aisles and return conditioned air to cold aisles, with hot aisle/cold aisle containment improving efficiency by preventing hot and cold air from mixing prematurely. Raised floor cooling distributes cold air through perforated tiles beneath the floor plenum and is common in legacy facilities built before 2010. The PUE range for traditional air-cooled facilities is typically 1.4 to 1.6 per industry benchmarks, reflecting the energy penalty of moving large volumes of conditioned air through a facility. At low-to-medium rack densities, this tradeoff is manageable. The infrastructure is familiar, vendor support is broad, and capital costs are lower than any liquid cooling alternative.
Benefits of Air Cooling for Data Centers
Air cooling's primary advantage is operational familiarity. Most facilities managers, CRAC vendors, and service contractors have deep expertise in air-cooled infrastructure, which reduces commissioning risk and simplifies ongoing maintenance. The absence of liquid at the rack level also eliminates the risk of coolant leaks damaging server hardware. Key benefits of air cooling for data centers:
- No liquid risk at the rack level, eliminating coolant leak damage to server hardware
- Lower capital cost than any liquid cooling alternative
- Broad CRAC and computer room air handler (CRAH) vendor support
- Incremental capacity expansion without major infrastructure redesign
- Infrastructure familiarity reduces commissioning and maintenance complexity
Limitations of Air Cooling at High Rack Densities
The density ceiling is the defining limitation of air cooling. Above 20 to 30 kW per rack, the airflow volumes required to maintain safe inlet temperatures create operational problems that cannot be engineered away within a conventional hot aisle/cold aisle containment framework. Performance also degrades in warm climates where the mechanical cooling load increases significantly. Key limitations of air cooling:
- Hard density ceiling of 20 to 30 kW per rack before thermal management fails
- PUE of 1.4 to 1.6 is high relative to liquid cooling alternatives
- Performance degrades in warm climates without supplemental mechanical cooling
- Not viable for AI and GPU workloads that routinely exceed 30 kW per rack
- Airflow management complexity increases disproportionately at high densities
Chilled Water and DX Systems: Centralised Cooling for Larger Facilities
Chilled water systems and direct expansion (DX) cooling represent two distinct centralised mechanical cooling approaches used in mid-to-large data center facilities. Chilled water systems use a central chiller plant to produce cold water distributed to air handling units or rear-door heat exchangers throughout the facility. DX systems use refrigerant that expands directly within the cooling unit at or near the IT equipment, without a separate chiller plant. The energy and scale tradeoffs between these approaches are significant for facility planners. Chilled water systems are more energy-efficient at large scale and offer flexibility for integration with free cooling economisers, which can push PUE into the 1.2 to 1.4 range in suitable climates. Cooling towers associated with chilled water plants consume water, which creates WUE implications and may trigger EPA 316(b) compliance requirements where facilities draw from surface water sources.
When Chilled Water Systems Make Sense
Chilled water systems are the appropriate choice for large-scale facilities, climates with sufficient annual free cooling hours, and deployments where integration with economiser modes is a design priority. Facilities with existing chiller infrastructure can retrofit chilled water distribution to rear-door heat exchangers without full redesign, making this a viable upgrade path for medium-density legacy facilities. Conditions favouring chilled water systems:
- Scale: Large-scale facilities where centralised chiller plant efficiency outweighs distribution complexity
- Climate: Locations with sufficient free cooling hours to justify economiser integration
- Existing infrastructure: Facilities with chiller plant already in place can extend to new cooling zones
- PUE target: Well-designed chilled water systems with economisers can achieve 1.2 to 1.4 PUE
- WUE implication: Cooling tower water consumption must be assessed against local water availability and EPA 316(b) obligations
DX Cooling: Simpler but Less Scalable
DX systems are better suited to smaller facilities, edge data centers, and deployments where simplicity and lower upfront cost outweigh long-term energy efficiency. The absence of a centralised chiller plant reduces installation complexity, but DX systems become progressively less efficient as facility scale increases. They are not a practical path to low PUE at hyperscale. Conditions relevant to DX cooling selection:
- Scale: Smaller facilities and edge data centers where centralised chiller plant is not justified
- Capital cost: Lower upfront cost than chilled water systems
- Scalability: Limited; efficiency drops as facility scale increases
- PUE range: Typically 1.4 to 1.6, similar to air-only cooling at scale
- Best fit: Edge deployments, small colocation facilities, and sites where installation simplicity is the priority
Evaporative Cooling and Free Cooling: Energy-Efficient Strategies and Their Tradeoffs
Evaporative cooling and free cooling are related but distinct energy efficiency strategies. Evaporative cooling uses water evaporation to pre-cool air before it enters the facility or the mechanical cooling loop. Free cooling uses ambient outdoor air, directly or via heat exchangers, when outdoor temperatures are low enough to cool IT equipment without mechanical refrigeration. Both strategies can reduce PUE toward 1.1 to 1.2 in cool climates with sufficient annual free cooling hours. The WUE tradeoff is the dimension most frequently omitted from competitor comparisons. Evaporative systems can consume significant volumes of water per kWh of IT load, which creates compliance risk in water-stressed regions and triggers regulatory review where applicable. Climate and geography are the primary determinants of whether these strategies are operationally viable.
Free Cooling Hours and Climate Suitability
The viability of free cooling depends on the number of annual hours during which outdoor temperatures are low enough to satisfy IT load without mechanical refrigeration. Facilities in the Pacific Northwest, Scandinavia, Ireland, and similar cool-moist climates can achieve significant economiser-mode operation annually. Facilities in arid or hot-humid climates see limited free cooling hours and face higher WUE penalties from evaporative approaches. ASHRAE climate zone classifications provide a structured framework for assessing free cooling viability during site selection or upgrade planning. Key dimensions for evaporative and free cooling:
- PUE range: Can reach 1.1 to 1.2 in cool climates with sufficient free cooling hours
- WUE implication: Evaporative systems increase water consumption per kWh of IT load, measured via WUE
- Climate dependency: High benefit in cool-moist climates; limited benefit in arid or hot-humid zones
- Capital cost: Lower than full liquid cooling; economiser integration adds cost to chilled water systems
- Regulatory exposure: Water consumption in evaporative systems may trigger EPA 316(b) review and local watershed regulations
Water Consumption and WUE Compliance Considerations
WUE quantifies the water a facility consumes per kWh of IT equipment load. A lower WUE indicates less water consumed relative to computing work performed. Evaporative cooling strategies that achieve low PUE do so partly by consuming water as the cooling medium, which creates a direct tradeoff between energy efficiency and water efficiency. In water-stressed regions, this tradeoff carries direct cost and compliance implications. Facilities drawing cooling water from surface water sources may be subject to EPA 316(b) thermal discharge and intake structure requirements regardless of WUE targets.
Liquid Cooling: Rear-Door Heat Exchangers, Direct Liquid Cooling, and Immersion
Liquid cooling encompasses three distinct categories of technology, ordered here from least to most infrastructure-intensive. Each addresses heat dissipation at a different point in the server hardware chain, and each carries different retrofit preconditions for legacy air-cooled facilities. Selecting among them requires understanding both the density threshold being served and the facility's structural and mechanical readiness. ASHRAE TC 9.9 publishes allowable rack inlet temperatures by facility class and infrastructure readiness criteria that govern whether liquid cooling can be added without full structural redesign. Three retrofit blockers are the most common in legacy facilities:
- Floor load ratings insufficient for immersion tank weight
- Raised-floor plenum depth inadequate for liquid distribution piping
- Power circuit specifications that require modification for high-density rack configurations
Rear-Door Heat Exchangers: The Lowest-Disruption Liquid Option
Rear-door heat exchangers (RDHx) attach to the back of standard server racks and remove heat via a liquid circuit before hot exhaust air enters the room. This approach is the lowest-disruption liquid cooling option and can be retrofitted into many existing air-cooled facilities without rack replacement. The result is a measurable PUE improvement over pure air cooling without requiring server hardware modifications. Key dimensions for rear-door heat exchangers:
- PUE range: Improves on air-only PUE; typically 1.2 to 1.4 depending on facility and climate
- WUE implication: Moderate water consumption through the liquid circuit; lower than evaporative systems
- Rack density suitability: Effective up to approximately 30 kW per rack in many configurations
- Retrofit complexity: Lower than DLC or immersion; compatible with standard rack infrastructure
- Capital cost: Moderate; higher than air-only but significantly lower than immersion
Direct Liquid Cooling for High-Density Workloads
Direct liquid cooling (DLC) runs coolant directly to CPU and GPU heat sinks via cold plates mounted on the server hardware. It is more efficient than rear-door approaches but requires server hardware specifically designed to accommodate liquid connections and rack-level plumbing. At densities above 30 kW per rack, typical of GPU cluster deployments, DLC becomes necessary rather than optional for effective thermal management. Key dimensions for direct liquid cooling:
- PUE range: Typically 1.1 to 1.2 with proper system design
- WUE implication: Lower water consumption than evaporative cooling; depends on the heat rejection method used
- Rack density suitability: Required above 30 kW per rack for GPU and AI workloads
- Retrofit complexity: High; requires server hardware with cold plate compatibility and rack-level plumbing modifications
- Capital cost: High relative to air and RDHx options; hardware compatibility adds to total cost
Immersion Cooling: Maximum Efficiency, Maximum Infrastructure Change
Immersion cooling submerges servers in dielectric fluid, removing heat directly from all hardware surfaces simultaneously. Uptime Institute research on immersion cooling PUE benchmarks indicates immersion cooling achieves PUE values of approximately 1.02 to 1.03, the lowest of any commercially available technology. Full and two-phase immersion both require complete hardware redesign, purpose-built immersion tanks, and structural floor load capacity significantly above standard raised-floor ratings. Key dimensions for immersion cooling:
- PUE range: 1.02 to 1.03, the lowest of any commercial cooling technology
- WUE implication: Minimal water consumption; dielectric fluid is recirculated
- Rack density suitability: No practical upper density limit; suited to the highest AI and GPU workloads
- Retrofit complexity: Extremely high; requires purpose-built tanks, hardware redesign, and structural assessment
- Capital cost: Highest of all cooling options; best suited to greenfield builds and hyperscale AI infrastructure
Geothermal Cooling and Heat Reuse: Emerging Strategies Worth Evaluating
Geothermal cooling and heat reuse are legitimate engineering strategies with established operational precedents, not speculative alternatives. Both are most relevant to greenfield facility planners, sustainability officers, and data center architects evaluating long-term site selection criteria.
Geothermal Cooling: Site Conditions and Applicability
Geothermal systems use the stable subsurface temperature of the earth, typically 50 to 60 degrees Fahrenheit in temperate climates, as a heat sink. Ground loop infrastructure circulates fluid through subsurface pipes to reject heat without mechanical refrigeration. Site geology is the primary constraint: clay-rich soils with high thermal conductivity support efficient ground loop performance, while rocky or fractured geology increases drilling cost and reduces heat transfer efficiency. The capital cost profile is high upfront but yields long-term operating cost reductions that make geothermal viable for greenfield hyperscale builds in suitable geologies with a long investment horizon.
Heat Reuse: Turning Waste Heat Into an Asset
Heat reuse systems capture server exhaust heat and redirect it to district heating networks, on-site process heat needs, or adjacent facilities. Several European data centers have integrated waste heat into municipal heating grids, converting an operating cost into a recoverable asset. The EU EED Article 12 reporting framework has elevated heat reuse from a voluntary sustainability initiative to a documented compliance consideration in some member states, where energy performance data submissions now require facilities to account for waste heat output. Heat reuse is most applicable to facilities located near district heating infrastructure or co-located with industrial processes requiring low-grade heat input.
How to Choose: A Decision Framework by Rack Density and Facility Profile
Matching the right cooling method to a specific facility requires working through two decision axes: rack density tier and facility age with retrofit feasibility. Neither axis alone is sufficient. A facility with low rack density but aging infrastructure faces different constraints than a greenfield build designed for high-density AI workloads, even if both arrive at the same cooling method as the answer. The cooling efficiency ratio (CER) provides a secondary performance benchmark alongside PUE, measuring cooling system energy consumption against IT load. Facilities benchmarking against EU EED Article 12 reporting requirements should confirm whether their national reporting authority requires CER alongside PUE submissions.
Low and Medium Density Facilities: Air Cooling Remains Viable
For facilities operating below 20 kW per rack, traditional air cooling with hot aisle/cold aisle containment remains the practical and cost-effective baseline. Adding a free cooling economiser to an existing chilled water system meaningfully improves PUE in climates with sufficient annual free cooling hours. Evaporative pre-cooling is a viable PUE improvement strategy where water availability supports it and WUE compliance obligations do not restrict it. Cooling options by density tier for low and medium density facilities:
- Under 10 kW per rack: Air cooling with CRAC units and hot aisle/cold aisle containment is sufficient; economiser integration is the primary PUE improvement lever
- 10 to 20 kW per rack: Air cooling remains viable with containment optimisation; rear-door heat exchangers can be added selectively for higher-density rows without full facility redesign
- PUE range achievable: 1.2 to 1.6 depending on cooling strategy and climate
- WUE implication: Low for air-only; moderate if evaporative pre-cooling is added
- Retrofit complexity: Low for air optimisation; moderate for RDHx addition; verify plenum depth and floor load before specifying
High and Ultra-High Density Facilities: When Liquid Cooling Becomes Necessary
Above 20 to 30 kW per rack, air cooling can no longer remove heat fast enough to maintain ASHRAE TC 9.9 allowable rack inlet temperatures. Liquid cooling becomes necessary, not optional. The choice among RDHx, DLC, and immersion depends on retrofit feasibility and hardware compatibility. Cooling options by density tier for high and ultra-high density facilities:
- 20 to 30 kW per rack: RDHx is the lowest-disruption retrofit path; DLC is viable where server hardware supports cold plate integration
- Above 30 kW per rack: DLC or immersion cooling is required; assess floor load ratings, plenum depth, and power circuit capacity before specifying DLC; immersion requires full hardware redesign and structural assessment
- PUE range achievable: 1.02 to 1.2 depending on method
- WUE implication: Low for DLC and immersion; higher if liquid heat rejection uses evaporative cooling towers
- Retrofit complexity: High for DLC; extremely high for immersion; greenfield builds avoid the retrofit constraint entirely
- Regulatory note: EU EED Article 12 makes PUE reporting a compliance obligation for large data centers, making cooling method selection a regulatory matter as well as an operational one
Where Adiabatic Humidification Fits in the Data Center Cooling Ecosystem
Adiabatic humidification adds moisture to supply air without adding heat, a thermodynamic distinction that makes it a functional complement to free cooling and evaporative pre-cooling rather than a standalone cooling replacement. Compressed air and water mix through a proprietary nozzle to produce an equal-sized droplet grid, where each droplet carries a slight charge to prevent re-aggregation and self-evaporates before contacting any surface, under proper system design. That evaporation lowers the effective dry-bulb temperature of supply air, reducing mechanical cooling load on CRAC and chiller infrastructure without introducing condensation risk. For facilities using data center humidification systems as a thermal management layer, the benefit is expanded free cooling availability. Free cooling and evaporative pre-cooling only work when supply air RH stays within a defined range, systems that let RH drop below it lose evaporative cooling efficiency and need additional mechanical cooling to compensate. Precision humidity control at the supply air stage keeps free cooling reliable across more annual hours.
Adiabatic Cooling as a Free Cooling Enabler
Evaporative pre-cooling depends on the humidity deficit of incoming air: drier air has more capacity to absorb moisture and therefore more evaporative cooling potential. Maintaining RH at the upper end of the ASHRAE TC 9.9 recommended range for IT equipment environments, without exceeding it, keeps this potential consistently available. Smart Fog's adiabatic humidifiers maintain RH to plus or minus 1 to 2% precision, enabling consistent free cooling performance rather than intermittent gains tied to uncontrolled RH fluctuations. For facilities with evaporative cooling humidifiers integrated into their cooling strategy, this precision is what converts a theoretical free cooling hour into an actual reduction in mechanical cooling load. Key performance specifications:
- RH maintained to plus or minus 1 to 2% precision across continuous 24/7 operation
- Self-evaporating droplet grid prevents surface wetting under proper system design
- Adiabatic process adds no heat to supply air, preserving the thermal benefit of free cooling
- No moving parts in the humidification process
- Maintenance intervals extend to every two years, compatible with data center uptime requirements
Precision Humidity Control for ESD Prevention in Air-Cooled Environments
Air cooling and free cooling strategies that introduce outdoor air can create low-humidity conditions inside the facility. ASHRAE TC 9.9's recommended range for data center IT equipment environments (5.5 to 15°C dew point, worth double-checking against the current thermal guidelines reference card, as flagged in the audit) sets the RH threshold that must be maintained to prevent electrostatic discharge (ESD) damage to server hardware and storage media. When RH drops below that range, electrostatic charge accumulates on surfaces and personnel, creating ESD events that damage hardware without generating visible failure indicators. The data center humidification industry overview explains how precision adiabatic humidification brings RH into the ASHRAE-recommended range without condensation or surface wetting risk, under proper system design. Two things worth knowing:
- Non-wetting has a boundary: placing a hand or any object directly into the fog stream will wet that surface. Non-wetting performance is a function of proper system design, not an unconditional property of the fog.
- Further reading for ESD-specific detail: Our articles on how humidity control prevents ESD in server rooms and why low humidity is dangerous for data centers cover the mechanism and thresholds in depth.
For facilities ready to specify a system, our articles on comparing humidity control methods for data centers and the data center humidification systems buyer's guide support procurement evaluation.
Final Thoughts
The best cooling method for a data center is the one that matches the facility's current and projected rack density, retrofit feasibility, climate, and regulatory obligations. Air cooling with hot aisle/cold aisle containment is the right answer for low-to-medium density environments with existing infrastructure. Liquid cooling becomes necessary above 20 to 30 kW per rack, with immersion cooling delivering the lowest PUE for greenfield high-density builds. Free cooling and evaporative strategies offer meaningful PUE gains in suitable climates but carry WUE tradeoffs that must be assessed against local water availability and EPA 316(b) obligations. Humidity control is the layer that connects thermal management strategy to hardware protection and free cooling reliability. Facilities that invest in cooling infrastructure without addressing RH precision risk both ESD-related hardware damage and inconsistent free cooling performance. To assess how precision adiabatic humidification can extend free cooling hours and protect IT equipment from ESD in a specific facility, speak with a Smart Fog engineer. Consult a Humidity Expert
FAQ
What is the most energy-efficient cooling method for data centers?
Immersion cooling is the most energy-efficient commercially available cooling method for data centers. It achieves PUE values of approximately 1.02 to 1.03 by submerging server hardware in dielectric fluid, eliminating the need for air movement and reducing heat dissipation losses. However, immersion cooling requires complete hardware redesign, purpose-built tanks, and significant structural capacity, making it most practical for greenfield hyperscale or AI infrastructure builds rather than legacy facility retrofits.
How does liquid cooling compare to air cooling in data centers?
Liquid cooling removes heat more efficiently than air cooling at high rack densities. Air cooling with CRAC units and hot aisle/cold aisle containment operates at PUE values of 1.4 to 1.6 and becomes physically inadequate above 20 to 30 kW per rack. Liquid cooling options, including rear-door heat exchangers, direct liquid cooling, and immersion, achieve PUE values of 1.02 to 1.4 depending on the method, and are required for AI and GPU workloads that exceed the air cooling density ceiling.
What is PUE and why does it matter for data center cooling decisions?
Power usage effectiveness (PUE) measures the ratio of total facility energy consumption to IT equipment energy consumption. A PUE of 1.0 is theoretical perfection; most air-cooled facilities operate between 1.4 and 1.6. PUE matters because it directly quantifies cooling energy waste: a facility with PUE 1.5 is consuming 50% more energy than its IT equipment requires. For large data centers in the EU, PUE reporting is now a compliance obligation under EED Article 12, making it a regulatory metric rather than a voluntary benchmark.
At what rack density does air cooling become inadequate for data center workloads?
Air cooling becomes thermally inadequate at rack densities above approximately 20 to 30 kW per rack, per Uptime Institute and Green Grid guidance. Above this threshold, the airflow volumes required to maintain safe rack inlet temperatures within ASHRAE TC 9.9 limits create noise, vibration, and thermal management complexity that air-based systems cannot resolve. Modern AI and GPU clusters, including deployments using NVIDIA H100 and H200 hardware, routinely exceed 30 kW per rack and require liquid cooling.
What is immersion cooling and what infrastructure does it require?
Immersion cooling submerges server hardware in dielectric fluid, removing heat directly from all hardware surfaces without air movement. It achieves PUE values of approximately 1.02 to 1.03. Infrastructure requirements include purpose-built immersion tanks, server hardware redesigned for fluid submersion, structural floor load capacity well above standard raised-floor ratings, and facility power circuits sized for high-density configurations. These requirements make immersion cooling best suited to greenfield builds and hyperscale AI infrastructure rather than legacy facility retrofits.
What is Water Usage Effectiveness (WUE) and how does it affect cooling system selection?
Water Usage Effectiveness (WUE) measures the volume of water a data center consumes per kWh of IT equipment load. It is the primary metric for assessing water consumption in cooling system selection. Evaporative and water-cooled chiller systems typically achieve lower PUE than air-only systems but at the cost of higher WUE. In water-stressed regions, high WUE carries direct operating cost and supply risk implications. Facilities drawing cooling water from surface water sources may also be subject to EPA 316(b) intake structure compliance requirements, making WUE a regulatory consideration alongside PUE.
Can liquid cooling be retrofitted into a legacy air-cooled data center?
Liquid cooling can be retrofitted into some legacy air-cooled data centers, depending on the method and facility conditions. Rear-door heat exchangers carry the lowest retrofit complexity and are compatible with standard rack infrastructure in many existing facilities. Direct liquid cooling requires server hardware with cold plate compatibility and rack-level plumbing modifications. Full immersion cooling is generally not retrofit-feasible in legacy facilities due to floor load, structural, and hardware redesign requirements. ASHRAE TC 9.9 infrastructure readiness criteria, including floor load ratings, raised-floor plenum depth, and power circuit specifications, govern retrofit viability in each case.
How does adiabatic humidification support free cooling strategies in data centers?
Adiabatic humidification adds moisture to supply air without adding heat, lowering the effective dry-bulb temperature of incoming air and expanding the operating window for free cooling economiser modes. Free cooling and evaporative pre-cooling are thermally effective only when relative humidity is within a defined range. Precision adiabatic humidification maintains RH to plus or minus 1 to 2%, ensuring that free cooling conditions are consistently available rather than intermittent. It also prevents the low-humidity conditions that air cooling and free cooling strategies can introduce, which reduces electrostatic discharge risk to server hardware within the ASHRAE TC 9.9 recommended dew point range for IT equipment environments.
