E03.01 — Understanding Data Centre Cooling Systems

How Modern Data Centres Remove Heat While Balancing Reliability, Energy Efficiency, Water Use and Environmental Performance

Key Takeaways

• Cooling is one of the largest engineering systems within any modern data centre because every watt consumed by computing equipment ultimately becomes heat.
• Modern data centres employ different cooling architectures depending on their computing requirements, climate, energy efficiency objectives and operational priorities.
• Cooling system design directly influences electricity consumption, water use, reliability, operating costs and environmental performance.
• Malaysia’s tropical climate presents unique engineering challenges that affect cooling efficiency, water demand and infrastructure design.
• Understanding cooling systems provides the engineering foundation for interpreting later MDCO discussions on sustainability, power infrastructure and low-frequency noise.

Introduction

Every modern digital service—from cloud computing and online banking to artificial intelligence (AI), video streaming and government digital platforms—depends upon one fundamental engineering challenge: removing heat.

Although data centres are often described in terms of servers, processors and networking equipment, computing itself is only part of the story. Every watt of electricity consumed by electronic equipment is ultimately converted into heat. Unless that heat is removed continuously, temperatures rise rapidly, reducing equipment performance, shortening component life and potentially leading to system failures.

Cooling is therefore not an optional support function. It is one of the core engineering systems that enables modern digital infrastructure to operate safely, reliably and continuously.

As discussed in →E03 Why Data Centres Are Built This Way, data centres are designed around reliability rather than convenience. The extensive cooling infrastructure found in hyperscale facilities—including chillers, cooling towers, pumps, fans, heat exchangers and sophisticated control systems—exists because computing equipment operates around the clock and must remain within carefully controlled environmental conditions.

This article forms part of the Malaysia Data Centre Observatory (MDCO) Explain Series and serves as the engineering foundation for several related MDCO topics. Readers interested in the environmental implications of cooling systems may also refer to →A09 Understanding Low-Frequency Noise from Data Centres, where cooling infrastructure is examined as the principal source of operational low-frequency noise.

Rather than focusing on a single piece of equipment, this article examines cooling as an integrated engineering system. It explains why cooling is necessary, how different cooling architectures operate, the function of major components, and how engineering decisions influence electricity consumption, water use, operational reliability and environmental performance.

MDCO Insight: Modern data centres are fundamentally heat-management systems. Cooling infrastructure exists not to produce cold air, but to remove the heat generated by continuous computing.

Why Data Centres Need Cooling

Every Watt of Electricity Eventually Becomes Heat

One of the most fundamental principles in engineering is the conservation of energy. Electricity consumed by computing equipment does not disappear. Instead, it is converted into other forms of energy, with almost all of it ultimately becoming heat.

Within a data centre, servers consume electrical power to perform computations, store information, communicate across networks and operate internal components such as processors, memory, storage devices and cooling fans. Although these activities perform useful digital work, virtually all of the electrical energy supplied to the equipment is eventually released as thermal energy.

A server consuming 500 watts of electrical power therefore produces approximately 500 watts of heat. A hyperscale facility supporting 50 megawatts (MW) of IT equipment similarly generates approximately 50 MW of heat that must be removed continuously to maintain safe operating conditions.

This relationship has an important implication: larger computing capacity inevitably requires larger cooling capacity.

As demand for cloud computing and AI infrastructure continues to grow, the engineering challenge is no longer simply supplying sufficient electrical power. It is also removing the corresponding quantities of heat efficiently, reliably and economically.

Heat Limits Computing Performance

Electronic components operate most reliably within specified temperature ranges.

Processors (CPUs), graphics processing units (GPUs), memory modules, storage devices and power electronics all generate heat during normal operation. If temperatures rise beyond acceptable limits, electronic performance begins to deteriorate.

Modern processors incorporate protective mechanisms that automatically reduce operating speed—a process known as thermal throttling—to prevent overheating. Although this protects the hardware, it also reduces computing performance.

If temperatures continue to rise, equipment reliability declines further. Excessive heat accelerates component ageing, increases the likelihood of electronic failure and may eventually trigger automatic shutdowns designed to prevent permanent damage.

For hyperscale operators supporting financial services, healthcare systems, government platforms or AI computing clusters, even brief interruptions can have significant operational and economic consequences.

Maintaining stable operating temperatures is therefore essential not only for equipment longevity but also for service continuity.

Cooling Never Stops

Unlike office buildings, shopping centres or residential developments, data centres rarely experience significant reductions in operation.

Cloud services remain available throughout the day and night. Online transactions occur continuously across different time zones. Artificial intelligence training may run uninterrupted for weeks. Financial systems, telecommunications infrastructure and government digital services likewise require continuous availability.

Because the computing equipment operates continuously, the cooling systems supporting it must also operate continuously.

Large ventilation fans circulate air throughout the facility.

Pumps transport chilled water or cooling fluids.

Cooling towers reject heat to the atmosphere.

Chillers compress refrigerants and transfer thermal energy.

Control systems constantly adjust operating conditions to maintain stable temperatures while minimising energy consumption.

Even when IT loads fluctuate throughout the day, the cooling infrastructure remains in continuous operation.

This continuous duty cycle distinguishes data centre cooling from many other building air-conditioning systems and explains why cooling represents one of the largest consumers of electricity within modern facilities.

Understanding the Heat Journey

Although cooling systems appear complex, their underlying purpose can be understood as a simple sequence of energy transfer.

The process begins inside the server, where electrical power is converted into computational work and heat.

That heat is transferred into the surrounding air or directly into cooling liquid.

The heat is then collected by cooling equipment within the data hall before being transported to larger mechanical systems located elsewhere within the facility.

Finally, the heat is rejected into the outdoor environment through cooling towers, dry coolers or air-cooled condensers.

At every stage, the objective remains the same:

Move heat away from sensitive computing equipment as efficiently as possible.

Understanding this sequence provides the foundation for examining the different cooling architectures used throughout the industry.

MDCO Insight: The engineering challenge is not creating cooling, but continuously transferring enormous quantities of heat away from computing equipment.

Major Data Centre Cooling Architectures

There Is No Single Cooling System

When people refer to “the air-conditioning system” of a data centre, they often imagine a single technology similar to that found in office buildings.

In reality, modern data centres employ a wide variety of cooling architectures depending on their size, computing density, climate, reliability requirements and sustainability objectives.

Smaller enterprise facilities may rely on relatively simple direct expansion (DX) systems similar to commercial buildings. Larger hyperscale facilities often employ central chilled-water systems with extensive mechanical infrastructure. Emerging AI facilities increasingly supplement or replace conventional air cooling with direct liquid cooling technologies capable of removing substantially greater heat loads.

Rather than representing competing technologies, these systems often coexist within the same facility.

Understanding their differences helps explain why data centres vary significantly in electricity consumption, water use, operating costs and environmental performance.

Direct Expansion (DX) Cooling Systems

Direct Expansion (DX) systems are among the simplest cooling architectures used within smaller data centres.

In these systems, refrigerant circulates directly between indoor Computer Room Air Conditioners (CRAC units) and outdoor condensers.

Warm air from the server room passes through cooling coils inside the CRAC unit, where heat is absorbed by the refrigerant. The refrigerant is then compressed and transported to outdoor condensers, where the heat is released into the surrounding atmosphere before the cycle repeats.

Because cooling occurs directly through the refrigeration circuit, these systems require relatively little supporting infrastructure.

Advantages include:

• relatively simple installation;
• lower initial capital costs;
• compact equipment layouts; and
• suitability for smaller facilities.

However, DX systems generally become less efficient as facilities increase in size. Their ability to support extremely high computing densities is also limited compared with larger chilled-water systems.

Chilled Water Cooling Systems

Most modern hyperscale data centres employ chilled-water cooling.

Rather than circulating refrigerant throughout the building, chilled-water systems produce cold water within central chiller plants and distribute it through insulated pipe networks.

Within the data halls, Computer Room Air Handlers (CRAH units) use chilled-water coils to cool the circulating air before returning it to the servers.

Outside the building, chillers remove heat from the water while cooling towers or air-cooled condensers reject the heat to the atmosphere.

Although considerably more complex than DX systems, chilled-water architectures provide several important advantages:

• improved energy efficiency;
• better scalability for very large facilities;
• greater operational flexibility;
• easier maintenance; and
• compatibility with high-density computing environments.

For this reason, chilled-water systems have become the dominant architecture for many hyperscale cloud providers.

Air-Cooled and Water-Cooled Chillers

Even within chilled-water systems, different methods exist for rejecting heat.

Air-cooled chillers transfer heat directly to outdoor air using large condenser fans.

They require little operational water but generally consume more electricity, particularly in warm climates.

Water-cooled chillers reject heat through cooling towers.

Because water-cooled chillers reject heat through evaporative cooling, they take advantage of the latent heat of vaporisation, allowing the condenser to operate closer to the ambient wet-bulb temperature rather than the higher dry-bulb temperature. This generally delivers higher energy efficiency than air-cooled systems. The trade-off is significant water consumption for evaporation, together with the need for continuous water treatment to control scaling, corrosion and biological growth such as Legionella.

Selecting between these alternatives involves balancing electricity consumption, water availability, environmental conditions and operating costs.

Indirect Evaporative Cooling

Some facilities reduce energy consumption by taking advantage of favourable outdoor weather conditions.

Indirect evaporative cooling uses outdoor air to cool a heat exchanger without allowing outside air to enter the server room directly.

This approach can significantly reduce compressor operation in cooler, drier climates.

However, the consistently high temperature and humidity found in tropical countries such as Malaysia limit the opportunities for this type of “free cooling” compared with temperate regions.

Direct Fresh-Air Cooling

Certain facilities located in cool climates introduce filtered outdoor air directly into the data halls.

This approach eliminates much of the mechanical refrigeration required during favourable weather conditions.

Although highly energy efficient in suitable climates, direct fresh-air cooling presents challenges where outdoor temperatures, humidity levels or airborne contaminants remain consistently high.

For Malaysia, this approach is generally less suitable because maintaining the environmental conditions required for sensitive electronic equipment becomes considerably more difficult.

Liquid Cooling

The rapid growth of artificial intelligence has accelerated interest in liquid cooling technologies.

Instead of relying primarily on air, liquid cooling transfers heat directly from processors into specially designed cooling liquids.

Several approaches are now commercially available:

• direct-to-chip cooling;
• rear-door heat exchangers;
• immersion cooling; and
• hybrid air-liquid systems.

Because liquids can transport heat far more effectively than air, these technologies support substantially higher computing densities while reducing the energy required to move cooling air throughout the facility.

Although still emerging within many commercial deployments, liquid cooling is expected to become increasingly important as AI computing continues to increase processor power densities.

Hybrid Cooling Systems

Many modern hyperscale facilities combine multiple cooling technologies within a single integrated system.

For example, conventional servers may continue using air cooling while high-density AI racks employ direct liquid cooling.

Similarly, chilled-water systems may operate alongside evaporative cooling or free-cooling modes whenever weather conditions permit.

Rather than relying on one universal solution, operators increasingly optimise cooling architectures according to computing density, local climate, operational objectives and sustainability priorities.

This systems approach allows facilities to balance reliability, energy efficiency, water consumption and long-term operational flexibility.

MDCO Insight: Modern data centres rarely rely on a single cooling technology; instead, they integrate multiple systems to optimise reliability, efficiency and environmental performance.

Understanding the Major Cooling Components

Cooling Systems Are Networks, Not Individual Machines

Modern data centre cooling systems are often described using terms such as CRAH, chiller or cooling tower. While these components are important, none of them operates independently.

Instead, cooling is achieved through a coordinated network of mechanical, electrical and control systems that continuously transfer heat from the servers to the outside environment. Each component performs a specific engineering function within this larger heat-transfer chain.

Understanding the role of each component helps explain why cooling systems consume significant amounts of electricity, why some systems use large quantities of water, and why different components contribute differently to maintenance requirements, operational costs and environmental impacts.

The following sections describe the principal components commonly found in modern hyperscale data centres.

Computer Room Air Conditioner (CRAC)

The Computer Room Air Conditioner (CRAC) is one of the oldest cooling technologies still widely used in data centres.

Unlike ordinary office air-conditioning units, CRAC systems are designed specifically for mission-critical environments that require precise temperature and humidity control throughout continuous operation.

A CRAC unit operates much like a conventional refrigeration system.

Warm air returning from the server room passes through an evaporator coil containing refrigerant. As the refrigerant absorbs heat, compressors raise its pressure before the heat is rejected outdoors through condensers.

The cooled air is then recirculated into the data hall.

Because each CRAC contains its own refrigeration circuit and compressors, these units consume relatively large amounts of electricity.

Their main components typically include:

• evaporator coil
• compressor
• expansion valve
• condenser
• circulation fans
• electronic controls

CRAC units remain common in enterprise facilities and smaller colocation centres but are becoming less common in very large hyperscale campuses, where chilled-water systems generally provide greater efficiency.

Computer Room Air Handler (CRAH)

Modern hyperscale facilities increasingly employ Computer Room Air Handlers (CRAHs) instead of CRACs.

At first glance, CRAHs appear very similar to CRAC units because both circulate cooled air throughout the data hall. The crucial difference lies in how cooling is produced.

A CRAH does not contain compressors or a refrigeration circuit.

Instead, chilled water produced by central chillers flows through cooling coils inside the CRAH. Large fans draw warm server exhaust air across these coils, transferring heat into the circulating water before returning cooled air to the servers.

This arrangement provides several advantages:

• lower energy consumption within the data hall;
• reduced maintenance inside white-space areas;
• improved scalability;
• greater operational flexibility.

Because refrigeration equipment is centralised elsewhere, CRAHs primarily consume electricity through their fan motors and electronic controls.

Chillers

The chiller is often regarded as the heart of a chilled-water cooling system.

Its purpose is to remove heat from circulating water before returning that cooled water to the CRAHs.

Although chillers vary considerably in design, most operate using the vapour-compression refrigeration cycle.

Within the chiller:

1. refrigerant absorbs heat from chilled water through the evaporator;
2. compressors increase refrigerant pressure;
3. refrigerant releases heat through the condenser;
4. expansion valves reduce pressure before the cycle repeats.

Large hyperscale facilities frequently employ multiple chillers operating in parallel.

This configuration provides:

• redundancy;
• maintenance flexibility;
• improved efficiency under varying loads;
• enhanced operational resilience.

Chillers represent one of the largest consumers of electrical power within the entire cooling system.

Cooling Towers

Cooling towers are commonly misunderstood.

Their purpose is not to cool servers directly.

Instead, they remove heat from condenser water leaving water-cooled chillers.

Warm condenser water is pumped to the cooling tower, where it is distributed across fill media while large fans draw ambient air upward.

A small proportion of the water evaporates.

Because evaporation absorbs large quantities of thermal energy, the remaining water becomes cooler before returning to the chiller.

Cooling towers therefore function primarily as heat rejection devices.

Their main components include:

• axial fans;
• distribution nozzles;
• fill media;
• drift eliminators;
• water basins;
• circulation pumps.

Cooling towers generally improve energy efficiency compared with air-cooled condensers.

However, this efficiency comes at the cost of increased water consumption.

Dry Coolers

Where water conservation is a priority, operators may install dry coolers instead of cooling towers.

Dry coolers resemble very large outdoor radiators.

Rather than relying on water evaporation, they reject heat directly into the atmosphere using large finned heat exchangers and high-capacity fans.

Advantages include:

• minimal water consumption;
• reduced water treatment requirements;
• simpler maintenance;
• elimination of evaporative losses.

The principal disadvantage is lower cooling efficiency during hot weather.

Consequently, dry coolers generally consume more electrical energy than cooling towers, particularly in warm tropical climates.

Pumps

Although often overlooked, pumps are essential components within almost every large cooling system.

Pumps continuously circulate:

• chilled water;
• condenser water;
• glycol mixtures;
• cooling liquids.

Large hyperscale campuses may contain dozens or even hundreds of pumps operating simultaneously.

Pump selection significantly influences overall energy consumption because pumping power increases rapidly as flow rates and pressure requirements increase.

Engineers therefore devote considerable attention to:

• hydraulic design;
• pipe sizing;
• pressure balancing;
• variable-speed drives;
• pump efficiency.

Even modest improvements in pump efficiency can produce substantial long-term electricity savings.

Piping Networks

Connecting all of these components is an extensive network of insulated piping.

Depending on the cooling architecture, separate pipe circuits may carry:

• chilled water;
• condenser water;
• refrigerants;
• glycol solutions;
• liquid cooling fluids.

Pipe design influences:

• pumping energy;
• heat losses;
• maintenance accessibility;
• operational reliability.

Large hyperscale campuses may contain several kilometres of piping connecting multiple buildings, cooling plants and data halls.

Heat Exchangers

Heat exchangers transfer thermal energy between fluids without allowing them to mix.

They appear throughout modern cooling systems.

Examples include:

• chiller evaporators;
• chiller condensers;
• plate heat exchangers;
• rear-door heat exchangers;
• liquid cooling distribution units.

Because heat exchangers contain no moving parts, they are generally highly reliable.

However, fouling, corrosion and scaling can gradually reduce their thermal performance, increasing electricity consumption.

Regular maintenance therefore remains essential.

Fans

Fans move enormous volumes of air throughout the cooling system.

Depending on the application, fans may be located within:

• CRAHs;
• CRACs;
• cooling towers;
• dry coolers;
• condensers;
• ventilation systems.

Modern hyperscale facilities increasingly employ electronically commutated (EC) fans equipped with variable-speed drives.

These allow airflow to adjust continuously according to IT load, improving efficiency while reducing unnecessary electricity consumption.

Because fans operate continuously, they also represent one of the principal sources of operational environmental noise.

Large axial fans associated with cooling towers, dry coolers and condensers are particularly relevant to MDCO’s separate series on →A09 Understanding Low-Frequency Noise from Data Centres.

Control Systems

Modern cooling systems are increasingly software-controlled.

Thousands of sensors continuously monitor:

• temperature;
• humidity;
• airflow;
• pressure;
• water flow;
• energy consumption;
• equipment health.

Building Management Systems (BMS) and Data Centre Infrastructure Management (DCIM) platforms analyse these data and optimise equipment operation in real time.

Variable-speed drives automatically adjust:

• fan speeds;
• pump speeds;
• compressor loading;
• valve positions.

Artificial intelligence is increasingly being applied to optimise cooling performance by predicting thermal loads before they occur.

Rather than operating individual machines independently, modern cooling systems function as integrated cyber-physical systems.

MDCO Insight: Data centre cooling depends on the coordinated operation of multiple specialised components, each performing a distinct role in transferring heat safely and efficiently away from computing equipment.

Cooling Efficiency: Understanding COP, PUE and System Performance

Cooling Consumes Energy

Cooling systems remove heat, but they also consume significant amounts of electricity.

Compressors require electrical power.

Fans consume electrical power.

Pumps consume electrical power.

Control systems consume electrical power.

Consequently, an efficient cooling system is not simply one that produces cold air.

It is one that removes the required amount of heat while using as little additional energy as possible.

Engineers therefore evaluate cooling systems using several complementary performance indicators.

What Is COP?

The most widely used engineering metric is the Coefficient of Performance (COP).

COP measures how effectively a cooling system converts electrical energy into useful cooling.

It is defined as:

COP = Cooling Output ÷ Electrical Input

For example:

If a chiller produces:

• 1,000 kW of cooling

while consuming:

• 250 kW of electricity,

its COP equals:

COP = 1,000 ÷ 250 = 4.0

This means every unit of electricity supplied to the chiller produces four units of cooling.

The higher the COP, the more efficient the cooling system.

Typical COP Values

Actual COP varies considerably depending on equipment design, operating conditions and climate.

Approximate values are:

Cooling Technology	Typical COP
Small DX systems	2.5–3.5
Air-cooled chillers	3–5
Water-cooled chillers	5–7
High-efficiency centrifugal chillers	7–9
Liquid cooling systems (system equivalent)	Often significantly higher effective efficiency

Actual performance changes continuously throughout the year.

A chiller achieving a COP of 7 under favourable conditions may perform substantially less efficiently during extremely hot weather.

COP Is Not the Whole Story

Although COP is extremely useful, it measures only individual equipment.

It does not include:

• pumps
• cooling tower fans;
• CRAH fans;
• control systems;
• distribution losses.

Consequently, two chillers with identical COP values may produce different overall facility performance depending on how the complete cooling system is designed.

Engineers therefore assess performance at both equipment level and system level.

Understanding PUE

Readers familiar with data centres often encounter another important metric:

Power Usage Effectiveness (PUE).

Unlike COP, which evaluates cooling equipment, PUE evaluates the efficiency of the entire facility.

It is defined as:

PUE = Total Facility Power ÷ IT Equipment Power

Suppose:

• servers consume 10 MW;
• cooling consumes 2 MW;
• electrical infrastructure consumes 1 MW.

Total facility power becomes:

• 13 MW.

The resulting PUE is:

13 ÷ 10 = 1.3

Lower PUE values indicate greater overall efficiency.

Because cooling often represents the largest non-IT energy consumer, improvements in cooling efficiency directly improve PUE.

COP and PUE Measure Different Things

These two metrics are often confused.

In reality, they answer different engineering questions.

Metric	Measures	Used For
COP	Efficiency of cooling equipment	Mechanical engineering
PUE	Efficiency of the whole data centre	Facility performance

A highly efficient chiller contributes to lower PUE, but excellent PUE also depends upon:

• efficient electrical distribution;
• high server utilisation;
• effective airflow management;
• efficient power supplies;
• optimised control systems.

Partial Load Matters

One common misconception is that cooling systems always operate at full capacity.

In practice, computing demand fluctuates continuously.

Modern chillers, pumps and fans therefore spend much of their operating lives at partial load.

Because energy efficiency changes under different loading conditions, manufacturers increasingly optimise equipment for high efficiency across a wide operating range rather than only at maximum output.

Variable-speed compressors and variable-frequency drives have become especially important in improving partial-load performance.

Optimising the Entire Cooling System

Modern hyperscale operators increasingly optimise cooling as an integrated system rather than focusing on individual components.

Examples include:

• sequencing multiple chillers according to load;
• adjusting chilled-water temperatures dynamically;
• varying fan speeds automatically;
• predicting future thermal demand using AI;
• selecting the most efficient operating mode based on weather conditions.

This systems approach can reduce energy consumption significantly without changing the underlying cooling equipment.

It also illustrates why cooling engineering increasingly combines mechanical engineering, electrical engineering, control systems and software optimisation.

MDCO Insight: Cooling efficiency depends not only on efficient equipment, but on how the entire cooling system is designed, controlled and operated as an integrated whole.

Electricity Consumption: Where the Energy Goes

Cooling Is One of the Largest Consumers of Electricity

After the IT equipment itself, cooling systems typically represent the largest consumer of electricity within a modern data centre.

Every fan, pump, compressor and control system requires electrical power to remove the heat generated by the servers. As computing capacity increases, the amount of heat that must be removed also increases proportionally.

Historically, cooling could account for 30–50% of a facility’s total electricity consumption. Advances in equipment efficiency, airflow management and control systems have significantly reduced this proportion in many modern hyperscale facilities. Nevertheless, cooling remains one of the largest non-IT electrical loads and continues to play a major role in determining a facility’s overall energy efficiency.

The exact proportion varies according to:

• cooling architecture;
• IT equipment density;
• climate;
• operating conditions;
• equipment efficiency; and
• facility utilisation.

For this reason, cooling is often one of the first engineering systems targeted when operators seek to improve sustainability and reduce operating costs.

Which Components Consume the Most Power?

Not every cooling component contributes equally to electricity consumption.

The largest consumers typically include:

Component	Primary Function	Relative Electricity Consumption
Chillers	Produce chilled water	Very High
Compressors (DX systems)	Refrigeration	Very High
CRAH / CRAC fans	Air circulation	High
Cooling tower fans	Heat rejection	Moderate
Pumps	Water circulation	Moderate
Dry cooler fans	Air-side heat rejection	Moderate
Building management systems	Monitoring and optimisation	Very Low

Although pumps and fans individually consume less electricity than chillers, their continuous operation means they collectively contribute a significant share of the cooling energy budget.

Engineers therefore optimise the entire system rather than focusing solely on the largest individual machines.

Fan Energy Increases Rapidly with Speed

Fans are particularly important because of the fan affinity laws.

A modest increase in fan speed can produce a disproportionately large increase in electricity consumption.

For example:

• doubling airflow does not merely double energy use;
• power demand increases approximately with the cube of fan speed.

Consequently, reducing fan speed by even a small amount can produce substantial electricity savings.

This explains why modern data centres increasingly employ:

• electronically commutated (EC) fans;
• variable frequency drives (VFDs);
• intelligent airflow control;
• hot-aisle and cold-aisle containment.

Rather than operating every fan at maximum speed continuously, operators adjust airflow dynamically according to actual server demand.

Pump Optimisation

Pumps also benefit from variable-speed operation.

Traditional pumping systems often maintained constant flow regardless of thermal demand.

Modern facilities instead adjust pump speeds continuously according to:

• chilled water temperature;
• differential pressure;
• server load;
• cooling demand.

Variable-speed pumping reduces unnecessary electricity consumption while maintaining adequate cooling throughout the facility.

The Cost of Inefficient Cooling

Even small improvements in cooling efficiency can produce substantial long-term savings.

Consider a hypothetical 100 MW hyperscale data centre.

If cooling efficiency improvements reduce electricity consumption by only 1 MW on average:

• annual electricity savings exceed 8,700 megawatt-hours (MWh);
• depending on electricity prices, this may represent millions of ringgit in annual operating cost savings;
• associated greenhouse gas emissions also decline proportionally where electricity is generated from fossil fuels.

For operators managing dozens of hyperscale campuses worldwide, relatively modest efficiency improvements can therefore produce significant financial and environmental benefits.

AI Is Changing Cooling Energy Requirements

Artificial intelligence is fundamentally changing cooling engineering.

Traditional enterprise servers typically consumed a few hundred watts per server.

Modern AI servers equipped with multiple high-performance GPUs may consume several kilowatts per server, while individual AI racks may exceed 100 kW, with some next-generation designs targeting 200–600 kW per rack.

Higher power densities generate proportionally greater heat.

Removing this heat efficiently increasingly requires:

• liquid cooling;
• advanced heat exchangers;
• higher-capacity pumps;
• improved thermal controls.

Although AI increases cooling demand, newer cooling technologies may simultaneously improve overall cooling efficiency by transferring heat more directly from the processors.

Electricity Efficiency Is Becoming a Strategic Issue

Cooling efficiency is no longer simply an engineering objective.

It increasingly influences:

• operating costs;
• sustainability reporting;
• carbon emissions;
• grid capacity;
• investment decisions;
• regulatory approvals.

As electricity demand from AI infrastructure continues to grow globally, improvements in cooling efficiency will play an increasingly important role in enabling future digital infrastructure expansion.

MDCO Insight: Cooling efficiency affects far more than electricity bills. It influences operating costs, carbon emissions, grid demand and the long-term sustainability of digital infrastructure.

Water Consumption and the Engineering Trade-Offs

Why Do Some Data Centres Use Water?

One of the most frequently discussed aspects of data centre sustainability concerns water consumption.

Not all data centres consume significant amounts of water.

The primary source of operational water use is evaporative cooling, particularly cooling towers used in water-cooled chilled-water systems.

Water is an exceptionally effective medium for transferring heat.

When water evaporates, it absorbs a large amount of thermal energy—known as the latent heat of vaporisation—allowing cooling towers to reject heat far more efficiently than many air-cooled alternatives.

This thermodynamic property explains why many hyperscale operators continue to employ evaporative cooling despite increasing public attention on water use.

Where Does the Water Go?

Cooling tower water is not “consumed” in the conventional sense.

Instead, water leaves the system through three principal mechanisms:

Evaporation

Most water loss occurs through evaporation.

This is the intended cooling mechanism.

As a small proportion of water evaporates, it removes heat from the remaining circulating water.

The evaporated water enters the atmosphere as water vapour.

Drift

Small water droplets may be carried away by airflow leaving the cooling tower.

Modern cooling towers incorporate drift eliminators that substantially reduce this loss.

Properly maintained towers typically limit drift to a very small fraction of circulating water.

Blowdown

As water evaporates, dissolved minerals remain behind.

Without intervention, mineral concentrations gradually increase, leading to scaling, corrosion and reduced equipment performance.

Operators therefore periodically discharge a portion of the circulating water—a process known as blowdown—and replace it with fresh make-up water.

Effective water treatment reduces blowdown frequency while protecting cooling equipment.

Water Usage Depends on Cooling Architecture

Different cooling systems exhibit very different water consumption profiles.

Cooling System	Electricity Use	Water Use
Air-cooled DX	Higher	Very Low
Air-cooled chillers	Higher	Very Low
Water-cooled chillers	Lower	Higher
Dry coolers	Moderate–High	Minimal
Direct liquid cooling	Depends on heat rejection system	Varies

No single system performs best across every criterion.

Instead, operators select cooling architectures according to:

• climate;
• electricity prices;
• water availability;
• environmental objectives;
• operating costs;
• reliability requirements.

The Water-Energy Trade-Off

One of the central engineering challenges is balancing electricity consumption against water consumption.

Water-cooled systems generally achieve higher energy efficiency because evaporation is an extremely effective cooling mechanism.

However, this advantage comes with increased water demand.

Conversely, air-cooled systems minimise water consumption but usually require greater electricity input, particularly in warm climates where ambient air temperatures remain high.

This creates an important trade-off:

• lower electricity use often requires more water;
• lower water use often requires more electricity.

Neither approach is inherently superior.

The optimal solution depends upon local environmental conditions, resource availability and sustainability priorities.

Water Treatment

Cooling water cannot simply circulate indefinitely.

Without treatment, biological growth, mineral scaling and corrosion gradually reduce cooling performance while increasing maintenance requirements.

Water treatment programmes commonly include:

• filtration;
• chemical dosing;
• biological control;
• corrosion inhibitors;
• conductivity monitoring.

These systems improve equipment reliability while reducing unnecessary water losses.

Water Use Effectiveness (WUE)

Just as PUE measures electrical efficiency, some operators report Water Usage Effectiveness (WUE).

WUE measures the amount of annual water consumed relative to the IT energy delivered.

Although definitions may vary slightly across reporting frameworks, WUE provides a useful indicator for comparing water performance between facilities.

Like PUE, WUE should be interpreted carefully.

A facility located in a cool northern climate may naturally achieve lower WUE than one operating in a hot tropical environment.

Consequently, climate and local operating conditions remain important considerations when comparing facilities internationally.

Water Stewardship Is Becoming Increasingly Important

Public discussion surrounding data centres increasingly extends beyond electricity consumption.

Communities, regulators and investors now also examine:

• water sourcing;
• local water availability;
• drought resilience;
• competing community water demands;
• wastewater management;
• long-term sustainability.

Many hyperscale operators have therefore introduced broader water stewardship initiatives, including:

• recycled water use;
• rainwater harvesting;
• non-potable water supplies;
• improved water monitoring;
• water-positive commitments.

Although these initiatives vary considerably between operators and locations, they reflect growing recognition that water has become an important sustainability consideration alongside energy efficiency.

Cooling Design Is Always About Trade-Offs

There is no universally “best” cooling system.

Every engineering decision involves balancing multiple objectives simultaneously.

A system optimised for maximum energy efficiency may consume more water.

A system designed to minimise water consumption may require additional electricity.

Equipment selected for lower operational noise may increase capital costs or occupy more land.

Similarly, cooling systems optimised for AI workloads may differ substantially from those designed for conventional enterprise computing.

Rather than seeking a perfect solution, engineers evaluate competing priorities and optimise the overall system according to the specific operational, environmental and economic context of each facility.

MDCO Insight: Every cooling technology involves trade-offs. Engineering decisions balance energy, water, reliability, cost and environmental performance rather than optimising a single objective.

Cooling Systems in Malaysia’s Tropical Climate

Climate Shapes Cooling Design

Cooling systems cannot be designed independently of their surrounding environment.

A data centre located in northern Sweden, Arizona or Johor faces fundamentally different operating conditions, even if the computing equipment inside is identical. Ambient temperature, humidity, rainfall, air quality and seasonal variability all influence how efficiently heat can be removed.

For Malaysia, climate is one of the defining factors shaping data centre cooling design.

Unlike temperate regions that experience cool winters and mild spring or autumn temperatures, Malaysia maintains consistently high temperatures and humidity throughout the year. While this provides a relatively stable operating environment with few seasonal extremes, it also limits opportunities to exploit naturally cool outdoor conditions.

Consequently, cooling systems in Malaysia generally operate under higher and more consistent thermal loads than equivalent facilities in cooler climates.

MDCO Insight: Data centre cooling systems are designed not only for computing demand but also for the climate in which they operate.

High Ambient Temperature

Malaysia’s average daytime temperatures typically range between 30°C and 35°C, with nighttime temperatures often remaining above 24°C.

These temperatures are significantly higher than those experienced in many established hyperscale markets in northern Europe or parts of North America.

Higher outdoor temperatures affect cooling in several ways.

First, compressors must work harder to reject heat because the temperature difference between the cooling system and the surrounding air is smaller.

Second, cooling towers become less effective as the ambient wet-bulb temperature increases.

Third, chillers generally operate at lower Coefficients of Performance (COP) during hotter periods because the refrigeration cycle requires greater compressor work.

Although modern chillers remain highly efficient, sustained tropical temperatures inevitably increase cooling energy requirements compared with cooler climates.

High Humidity

Malaysia’s relative humidity commonly exceeds 70–90% throughout much of the year.

Humidity presents engineering challenges that differ from temperature alone.

Excessive moisture can contribute to:

• condensation;
• corrosion;
• reduced equipment reliability;
• indoor environmental control challenges.

Cooling systems must therefore regulate not only temperature but also humidity within acceptable ranges for sensitive electronic equipment.

Historically, data centres maintained relatively narrow humidity limits.

Modern IT equipment has become considerably more tolerant, allowing operators to widen acceptable humidity ranges and reduce unnecessary energy consumption.

Nevertheless, humidity control remains an important aspect of cooling system design in tropical environments.

Limited Opportunities for Free Cooling

Many hyperscale facilities located in Europe or northern North America achieve substantial energy savings through free cooling.

During cool weather, outdoor air or naturally cooled water can remove heat without requiring compressors to operate continuously.

Malaysia offers far fewer opportunities for this approach.

Because outdoor temperatures remain consistently high, most facilities rely on mechanical refrigeration throughout the year.

Indirect evaporative cooling and economiser modes may still contribute under certain conditions, but their annual operating hours are generally much lower than those achieved in cooler climates.

Consequently, Malaysian facilities depend more heavily on efficient chillers, advanced control systems and optimised airflow management.

Rainfall and Water Availability

Malaysia receives abundant annual rainfall.

However, abundant rainfall does not necessarily eliminate concerns regarding operational water use.

Cooling towers require reliable make-up water throughout the year, regardless of rainfall patterns.

Operators must therefore consider:

• municipal water supply capacity;
• water treatment requirements;
• drought resilience;
• competing regional water demands;
• long-term sustainability planning.

Many newer facilities increasingly investigate alternative water sources such as:

• reclaimed water;
• treated industrial water;
• harvested rainwater;
• recycled cooling water.

These approaches seek to reduce dependence on potable water supplies while improving long-term environmental performance.

Tropical Air Quality

Malaysia periodically experiences regional haze episodes resulting from transboundary biomass burning.

Outdoor air also contains:

• dust;
• pollen;
• salt particles in coastal regions;
• industrial pollutants in urban environments.

These airborne contaminants influence cooling design.

Facilities employing outside-air cooling require sophisticated filtration systems to protect sensitive electronic equipment.

Heat exchangers and cooling coils also require regular cleaning because accumulated contaminants gradually reduce thermal efficiency.

Maintaining air quality is therefore an integral part of cooling system maintenance.

Cooling Design for Continuous Operation

Malaysia’s climate presents relatively few sudden seasonal changes.

This stability simplifies certain aspects of system design because operators do not need to accommodate freezing temperatures or extreme winter conditions.

However, the absence of cooler seasons also means cooling equipment receives little natural respite.

Chillers, pumps, cooling towers and fans operate continuously under relatively demanding environmental conditions throughout the year.

Equipment selection therefore places particular emphasis on:

• reliability;
• corrosion resistance;
• maintainability;
• energy efficiency under tropical conditions.

These factors help explain why preventive maintenance plays an especially important role in Malaysian data centre operations.

Malaysia’s Growing AI Infrastructure

Malaysia is rapidly becoming one of Southeast Asia’s largest destinations for hyperscale and AI infrastructure investment.

The emergence of high-density AI computing introduces new thermal challenges.

AI servers generate substantially higher heat loads than conventional enterprise servers.

Consequently, future Malaysian facilities are expected to incorporate increasing levels of:

• direct liquid cooling;
• rear-door heat exchangers;
• higher-capacity chilled-water systems;
• advanced thermal monitoring;
• AI-assisted cooling optimisation.

The country’s tropical climate will therefore continue to shape both present and future cooling technologies.

MDCO Insight: Malaysia’s consistently warm and humid climate makes cooling efficiency a year-round engineering priority rather than a seasonal operational consideration.

Cooling Systems and Low-Frequency Noise

Cooling Systems Also Generate Sound

Cooling systems are designed to remove heat—not to generate noise.

Nevertheless, whenever large volumes of air are moved, water is circulated or mechanical equipment operates continuously, sound is inevitably produced.

In most situations, this sound remains an accepted aspect of industrial infrastructure.

However, as hyperscale data centres become larger and are increasingly located closer to residential communities, operational noise has become an important planning and community consideration.

Among the various forms of environmental noise associated with data centres, low-frequency noise (LFN) has attracted growing international attention (→A09.01 Why Low-Frequency Noise Becomes a Community Conflict).

Readers seeking a detailed multidisciplinary discussion should refer to → A09 — Understanding Low-Frequency Noise from Data Centres, together with the accompanying MDCO Low-Frequency Noise series.

Which Cooling Components Produce Noise?

Different components generate different acoustic characteristics.

Cooling Towers

Cooling towers are often among the most significant contributors to environmental noise.

Primary sources include:

• large axial fans;
• water falling through fill media;
• airflow turbulence;
• mechanical vibration.

Because these fans often operate continuously, they can become noticeable during quieter nighttime conditions.

Dry Coolers and Condensers

Dry coolers rely on multiple large-diameter fans operating continuously to reject heat.

Fan blade rotation and airflow turbulence generate broadband sound, while certain operating conditions may also produce tonal characteristics.

Variable-speed control can reduce unnecessary noise during periods of lower cooling demand.

Chillers

Chillers contain:

• compressors;
• pumps;
• refrigerant flow;
• control valves.

Modern centrifugal chillers are generally quieter than many older designs, but compressor operation can still contribute to the overall acoustic profile.

Proper equipment selection and maintenance significantly influence operational noise.

CRAHs and CRACs

Inside the data hall, CRAHs and CRACs primarily generate noise through:

• circulation fans;
• airflow;
• internal motors.

Although much of this sound remains contained within the building, it may contribute to overall facility noise if transmitted through ventilation systems or building structures.

Pumps

Pumps typically generate relatively little airborne noise compared with fans.

However, poorly isolated pumps may transmit mechanical vibration through pipework and structural elements.

Proper vibration isolation therefore forms an important aspect of acoustic engineering.

Why Low-Frequency Noise Receives Particular Attention

Not all sound behaves in the same manner.

Higher-frequency sounds tend to attenuate more rapidly over distance and are more readily absorbed by buildings and vegetation.

Low-frequency sound behaves differently.

Its longer wavelength allows it to:

• travel greater distances under certain conditions;
• diffract around obstacles;
• penetrate buildings more readily;
• remain noticeable indoors, particularly at night.

Consequently, relatively modest low-frequency sound may be perceived differently from higher-frequency environmental noise.

Understanding these physical characteristics is essential for interpreting many community discussions surrounding data centre operations.

Engineering Measures for Noise Reduction

Modern data centre design increasingly incorporates acoustic engineering from the earliest planning stages.

Common mitigation measures include:

• selecting lower-noise equipment;
• variable-speed fan operation;
• acoustic louvres;
• silencers;
• equipment enclosures;
• vibration isolation;
• resilient pipe supports;
• optimised equipment layout;
• increased setbacks from residential areas.

Rather than relying on a single solution, operators generally combine multiple engineering measures to reduce overall acoustic emissions.

Balancing Noise and Performance

Noise reduction often involves engineering trade-offs.

For example:

Reducing fan speed generally decreases noise.

However, lower airflow may reduce cooling capacity.

Installing additional acoustic barriers improves sound attenuation but may increase airflow resistance.

Changing equipment layouts may require larger sites or higher construction costs.

Engineers therefore seek solutions that maintain:

• reliable cooling;
• acceptable energy efficiency;
• operational flexibility;
• reasonable acoustic performance.

These objectives must be balanced simultaneously rather than optimised individually.

Engineering Alone Is Not Always Sufficient

International experience demonstrates that technically compliant facilities may nevertheless become the subject of continuing community concern.

Acoustic measurements, engineering standards and regulatory compliance represent important components of responsible infrastructure management.

However, community experience may also be influenced by:

• operating patterns;
• environmental conditions;
• cumulative infrastructure;
• communication;
• public expectations;
• trust.

For this reason, MDCO examines low-frequency noise not only as an engineering issue but also through the perspectives of governance, regulation, planning, public health research and community engagement.

MDCO Insight: Low-frequency noise begins as an engineering phenomenon but often becomes a multidisciplinary issue involving acoustics, governance, planning and community experience.

Emerging Cooling Technologies and the Future

Cooling Is Evolving Alongside Computing

For much of the history of data centres, cooling technologies evolved gradually. Improvements focused on incremental gains in chiller efficiency, airflow management and control systems while the underlying architecture remained largely unchanged.

Artificial intelligence is changing this trajectory.

Modern AI processors consume substantially more electrical power than traditional enterprise servers, producing correspondingly higher heat loads within increasingly compact spaces. As rack power densities continue to rise, conventional air-cooling approaches are approaching practical engineering limits for certain applications.

This shift is driving a new generation of cooling technologies that seek not only to improve efficiency but also to enable computing capacities that would otherwise be difficult to achieve.

MDCO Insight: The future of data centre cooling will be driven as much by advances in computing as by advances in mechanical engineering.

Direct Liquid Cooling

One of the most significant developments is Direct Liquid Cooling (DLC).

Rather than cooling the surrounding air, specially designed cold plates are attached directly to high-power components such as CPUs and GPUs. A circulating liquid removes heat immediately from the processor before transferring it to a secondary cooling loop.

Compared with conventional air cooling, DLC offers several advantages:

• significantly higher heat removal capability;
• reduced fan energy consumption;
• higher rack power density;
• improved thermal stability;
• lower overall airflow requirements.

Although air cooling continues to serve many conventional workloads effectively, direct liquid cooling is increasingly becoming the preferred solution for large AI clusters.

Immersion Cooling

Another emerging technology is immersion cooling.

Instead of blowing air across servers, complete servers are immersed in specially formulated dielectric fluids that do not conduct electricity.

Heat generated by electronic components transfers directly into the surrounding liquid before being removed through heat exchangers.

Potential benefits include:

• extremely high cooling capacity;
• minimal server fan requirements;
• reduced airborne contamination;
• quieter operation;
• improved energy efficiency for certain high-density applications.

However, immersion cooling also introduces new engineering considerations involving equipment compatibility, maintenance procedures and operational practices.

For these reasons, widespread deployment remains concentrated in specialised high-density environments.

Artificial Intelligence Optimising Cooling

Artificial intelligence is influencing cooling in two different ways.

First, AI increases computing density and therefore cooling demand.

Second, AI is increasingly being used to optimise cooling systems themselves.

Modern control platforms analyse thousands of sensor readings continuously, including:

• server temperatures;
• airflow patterns;
• chilled-water temperatures;
• weather forecasts;
• equipment loading;
• historical operating data.

Machine learning algorithms can then predict thermal demand and adjust chillers, pumps and fans proactively rather than reactively.

This allows cooling systems to maintain stable operating conditions while reducing unnecessary energy consumption.

Heat Reuse

Traditionally, most data centres simply rejected waste heat into the surrounding environment.

Increasingly, operators are investigating opportunities to recover and reuse this thermal energy.

Examples include:

• district heating;
• industrial process heating;
• greenhouse agriculture;
• domestic hot water systems;
• absorption cooling.

Such opportunities depend heavily on local geography, nearby heat demand and economic feasibility.

While large-scale heat reuse remains relatively uncommon in tropical climates such as Malaysia, future industrial developments may create new opportunities.

Designing for Sustainability

Future cooling systems will increasingly be evaluated against multiple objectives simultaneously.

Operators are no longer concerned only with maintaining acceptable temperatures.

Cooling systems are now expected to support:

• higher computing densities;
• lower energy consumption;
• reduced water use;
• lower carbon emissions;
• improved resilience;
• quieter operation;
• greater operational flexibility.

These objectives occasionally complement one another but may also require engineering trade-offs.

Consequently, future cooling design is expected to become increasingly integrated across mechanical engineering, electrical engineering, digital controls, environmental management and sustainability planning.

MDCO Insight: Future cooling systems will optimise multiple objectives simultaneously, balancing computing performance, efficiency, sustainability and operational resilience.

Why Understanding Cooling Matters

To many people, cooling systems appear to be little more than large industrial air-conditioners.

In reality, they represent one of the most sophisticated engineering systems within a modern data centre.

Cooling determines:

• whether servers operate reliably;
• how much electricity a facility consumes;
• how much water may be required;
• how efficiently computing resources can be delivered;
• the facility’s environmental footprint;
• many of the operational costs incurred throughout its lifetime.

It also influences wider public-interest discussions concerning sustainability, energy infrastructure, water resources and environmental noise.

Understanding cooling therefore provides an essential foundation for understanding many other aspects of modern data centre development.

Why MDCO Is Examining Cooling Systems

Cooling systems illustrate one of MDCO’s recurring themes.

Complex infrastructure cannot be understood by examining individual components in isolation.

A cooling tower is not simply a mechanical device.

It influences:

• electricity demand;
• water consumption;
• greenhouse gas emissions;
• operational expenditure;
• environmental noise;
• community experience.

Similarly, a chiller is not merely a refrigeration machine.

Its performance affects energy efficiency, sustainability reporting, carbon intensity, grid demand and long-term infrastructure planning.

Understanding these relationships requires a systems perspective rather than a single engineering discipline.

This article therefore forms part of MDCO’s broader effort to explain how technical systems interact with governance, regulation, sustainability and society.

MDCO Insight: Cooling systems demonstrate how engineering decisions influence energy, water, sustainability, economics and community outcomes simultaneously.

The Observatory Perspective

Throughout the Explain Series, MDCO has examined the engineering principles that underpin modern digital infrastructure.

→E01 What Is a Data Centre? introduced the fundamental role of data centres within the digital economy.

→E02 How Data Centres Interact with the World explained how facilities exchange electricity, water, telecommunications and heat with the surrounding environment.

→E03 Why Data Centres Are Built This Way examined the engineering philosophy behind redundancy, resilience and continuous operation.

This article extends that discussion by examining one of the most important engineering systems within every modern facility: cooling.

Understanding cooling also provides the foundation for many subsequent MDCO topics.

Electricity consumption, water use, low-frequency noise, sustainability reporting, AI infrastructure and environmental performance all depend, to varying degrees, upon how heat is managed.

Future articles within both the Explain and Analyse Series will build upon these engineering principles to explore the wider governance, environmental and societal implications of modern data centre development.

As computing technologies continue to evolve, cooling will remain one of the defining engineering challenges shaping the future of digital infrastructure.

MDCO Insight: Cooling lies at the intersection of engineering, sustainability and digital infrastructure, making it one of the defining systems shaping the future of modern data centres.

Selected References

• ASHRAE. Thermal Guidelines for Data Processing Environments (5th Edition). Atlanta: American Society of Heating, Refrigerating and Air-Conditioning Engineers. Technical guidance on environmental conditions, thermal management and cooling design for data centres. https://www.ashrae.org/
• ASHRAE Technical Committee 9.9. Mission Critical Facilities, Data Centers, Technology Spaces and Electronic Equipment. Guidance on data centre cooling, airflow management and thermal design. https://www.ashrae.org/
• The Green Grid. Power Usage Effectiveness (PUE): A Comprehensive Examination of the Metric. Industry guidance on measuring and interpreting energy efficiency in data centres. https://www.thegreengrid.org/
• Uptime Institute. Global Data Center Survey. Annual analysis of data centre operations, cooling technologies, resilience and sustainability trends. https://uptimeinstitute.com/
• International Energy Agency (IEA). Energy and AI. Analysis of global electricity demand, AI infrastructure and energy implications for digital infrastructure. https://www.iea.org/
• Open Compute Project (OCP). Advanced Cooling Solutions. Technical resources covering liquid cooling, immersion cooling and future thermal management technologies. https://www.opencompute.org/

Citation

Malaysia Data Centre Observatory (MDCO). Understanding Data Centre Cooling Systems. MDCO Explain Series No. E03.01 (Version 1.0, July 2026).

MDCO Note

This article forms part of the Malaysia Data Centre Observatory (MDCO) Explain Series, which aims to improve public understanding of data centre development through evidence-based, accessible and balanced analysis. It is intended for educational and informational purposes only and does not constitute legal, engineering, planning, environmental or professional advice.

Malaysia’s rapidly evolving data centre ecosystem includes facilities developed, owned or operated by organisations such as AirTrunk, Amazon Web Services (AWS), Bridge Data Centres, DayOne, EdgeConneX, Google, K2 Data Centres, Microsoft, NTT Global Data Centers, Princeton Digital Group (PDG), ST Telemedia Global Data Centres (STT GDC), STACK Infrastructure, Vantage Data Centers, YTL Data Centre Park and many others. MDCO is independent of these organisations, as well as governments, regulators, utilities and advocacy groups. Its role is to facilitate transparency, structured understanding and equal access to information by presenting publicly verifiable evidence, relevant context and multiple stakeholder perspectives. MDCO does not endorse, oppose or advocate for any particular organisation, project or policy position.