Data Centre Cooling Systems Explained: Types, Rack Density and Costs in Australia

Data Centre HVAC · Precision Cooling · Australia

Cooling is the single largest non-IT energy cost in a data centre, commonly accounting for roughly 35 to 40 percent of total facility electricity. It is also the system most likely to take a site down. A UPS failure is usually visible and dramatic; a cooling failure is quiet, and in a well-packed hall it can push rack inlet temperatures past safe limits in a matter of minutes. Understanding which cooling system a room actually needs is therefore both an availability decision and an operating cost decision.

This guide sets out how data centres are cooled in Australia, how the options compare, the rack density thresholds where each one stops working, what redundancy really means, and what a system costs to install. It is written from the perspective of the people who size, install and service the plant rather than the people who sell a single product.

What is a data centre cooling system?

A data centre cooling system is the complete chain that moves heat from a processor to the outside air. It starts at the rack, where servers convert almost all of the electricity they draw into heat, and ends at a condenser, dry cooler or cooling tower outside the building. Every component in between exists to carry that heat one step further along the chain without letting the equipment inlet temperature drift outside a safe band.

The sizing rule is simple and worth memorising: IT load in kilowatts is very close to cooling load in kilowatts. A 40 kW rack row needs roughly 40 kW of cooling, plus an allowance of about 10 to 20 percent for lighting, building envelope gain, people and the UPS itself. Floor area is almost irrelevant. This is why a 12 square metre comms room with four dense racks can need more cooling than a 60 square metre office. Our server room heat load calculator works this out for a specific room.

The target condition is set by ASHRAE Technical Committee 9.9, whose recommended envelope for class A1 to A4 equipment is 18 to 27 degrees Celsius at the equipment inlet, with relative humidity generally held below 60 percent to control both condensation and static discharge risk. Note that the measurement point is the front of the rack, not the wall thermostat. Rooms fail audits routinely because the wall reads 22 degrees while the top of the third rack reads 34.

The main types of data centre cooling systems

There are five practical families in Australian sites, running from a small comms cupboard to an AI training hall. Most real facilities use two or three of them together.

CRAC vs CRAH: the difference that matters

A CRAC unit is a Computer Room Air Conditioner. It contains its own refrigeration circuit with a compressor, and it rejects heat through an outdoor condenser or a condenser water loop. A CRAH unit is a Computer Room Air Handler. It has no compressor at all. It is a fan and a chilled water coil, and the actual refrigeration happens at a central chiller plant somewhere else in the building.

The practical consequence is scale. CRAC units are self contained, so a single room can be cooled without any central plant, which is why almost every Australian comms room and switchroom runs on DX CRAC. CRAH units are far more efficient once a site is big enough to justify a chiller plant, because one large chiller is more efficient than twenty small compressors, and because chilled water systems can take advantage of free cooling in cooler weather. The crossover in practice tends to sit somewhere around a few hundred kilowatts of IT load. Our CRAC topology selector walks through the DX, condenser water and chilled water decision in detail.

Airflow management: the cheapest cooling you can buy

Before adding capacity, most rooms should fix airflow. Cooling capacity that never reaches a server inlet is wasted capacity, and mixing of hot exhaust with cold supply air is the most common reason a room with adequate kW on paper still has hot spots.

Hot aisle and cold aisle layout arranges racks front to front and back to back so that all cold air is delivered into one aisle and all exhaust is collected in the other. Containment then puts a physical barrier around one of those aisles, typically doors at each end and a roof or curtain over the top, so the two air streams cannot mix at all. Blanking panels in every unused rack unit stop air short circuiting through the rack itself, and brush grommets seal cable cut-outs in a raised floor.

These measures are cheap, they need no refrigeration work, and in a poorly managed room they routinely free up 20 to 30 percent of existing cooling capacity. That is often the difference between needing a new unit and not needing one. Any competent cooling assessment should start here.

Heat rejection and the Australian ambient problem

Whatever happens inside the room, the heat still has to leave the building. That is the job of the outdoor half of the system: air cooled condensers, dry coolers, adiabatic or evaporative coolers, and cooling towers on larger chilled water plants. This is where Australian conditions bite, and where imported design assumptions cause the most trouble.

Condenser capacity falls as ambient temperature rises. A unit selected against a 35 degree design ambient will struggle on a 42 degree day in Adelaide or western Sydney, which is precisely when the room is under most stress and when the grid is least reliable. Summer design dry bulb temperatures used for condenser selection commonly sit in the mid 30s in Brisbane and Sydney, the high 30s to around 40 degrees in Melbourne, Adelaide and Perth, and in the mid 30s in Darwin but at very high humidity that limits evaporative options. Any design should be confirmed against current AIRAH design weather data for the specific site, and the condenser should be selected with headroom above the design day rather than exactly at it.

Free cooling, where outside air is cool enough to do part or all of the work without running compressors, is genuinely valuable in Melbourne, Canberra, Hobart and Adelaide overnight and through winter. It delivers far fewer usable hours in Brisbane, Townsville, Cairns and Darwin. Water based evaporative systems save energy but consume water and add legionella management obligations under AS/NZS 3666, which is a real operating commitment rather than a footnote.

Redundancy: what N+1 actually means

Cooling redundancy is described in the same language as power redundancy, and it is just as often misused. N is the number of units required to carry the full design load. N+1 means one additional unit beyond that, so any single unit can fail or be taken offline for service without losing the room. 2N means a fully duplicated system.

The common mistake in smaller rooms is to install two units, each sized for 100 percent of the load, and then run both continuously at 50 percent. That is genuine N+1 and it is a good arrangement, but only if the control system rotates lead and lag duty so both units accumulate similar run hours and both are proven to start. Two units where the second has not run in three years is not redundancy. The other frequent gap is that cooling redundancy is meaningless if all units share one power supply, one condenser water loop, or one drain. Concurrent maintainability means every part of the chain can be isolated for service without dropping the load, and it needs to be designed in rather than assumed.

What data centre cooling systems cost in Australia

Installed cost depends on capacity, condenser location, electrical work and how much builder's work the site needs. As indicative bands for design, supply, installation and commissioning, Indigi projects typically fall as follows.

Running cost is the larger number over a system's life. A room with poor containment, a badly set economiser or units fighting each other on humidity can easily consume double the energy of a well commissioned equivalent. Maintenance is the cheap insurance on both: filters, coil cleanliness, refrigerant charge, condensate drainage and control set points all drift, and they drift silently.

Data centre cooling design, installation and service across Australia

Indigi Power and Cooling covers precision cooling nationally from Brisbane, Sydney and Melbourne hubs, including regional and remote sites.

Frequently asked questions

What type of cooling system do most data centres use?

Most data centres still use air cooling. Smaller Australian sites and comms rooms almost always run direct expansion CRAC units, while larger halls use CRAH units fed from a central chilled water plant. Liquid cooling is growing quickly but is currently confined to high density AI, GPU and HPC deployments.

How do data centres cool their systems?

Cool air is delivered to the front of each rack, servers draw it through and exhaust it hot at the rear, and cooling units capture that hot air and transfer its heat into a refrigerant or chilled water circuit. The heat is then rejected outside through a condenser, dry cooler or cooling tower. Containment keeps the cold supply and hot exhaust air streams separated so they do not mix.

Who builds cooling systems for data centres in Australia?

Precision cooling is designed and installed by specialist contractors holding refrigeration and electrical licences, rather than by general air conditioning installers. Indigi Power and Cooling carries out heat load assessment, unit selection, installation, commissioning and ongoing maintenance nationally, and is Supply Nation and ICN Gateway registered for government and corporate procurement.

How much do data centre cooling systems cost?

In Australia, a small comms room precision cooling project typically runs $8,000 to $15,000 installed, a medium server room with N+1 redundancy $25,000 to $45,000, and a larger multi unit hall $45,000 to $60,000 and above. Cost is driven by capacity, condenser location and electrical work rather than by room size.

What is the difference between CRAC and CRAH?

A CRAC unit has its own compressor and refrigeration circuit and rejects heat to an outdoor condenser. A CRAH unit has no compressor and instead passes room air over a chilled water coil supplied by a central chiller plant. CRAC suits self contained rooms, while CRAH is more efficient once a site is large enough to justify central chillers.

What temperature should a data centre be kept at?

ASHRAE TC 9.9 recommends 18 to 27 degrees Celsius measured at the equipment air inlet, with relative humidity generally held below 60 percent. Measuring at the rack face rather than a wall thermostat matters, because inlet temperature at the top of a rack is often several degrees higher than room average.

Related services

• Server room cooling design and installation
• CRAC unit maintenance services
• CRAC installation and commissioning
• CHW CRAC maintenance (chilled water)
• CRAC topology selector (DX, CDW, CHW)
• Server room heat load calculator
• Server room UPS and cooling

Not sure whether your room needs more cooling or better airflow? Send us your rack count, IT load in kilowatts and current unit models and we will tell you what the room actually needs. Indigi Power and Cooling is an Indigenous and Veteran-owned critical power and cooling specialist, registered with Supply Nation and ICN Gateway. Contact Indigi Power and Cooling