In recent years, the rapid development of artificial intelligence has significantly increased computing demand in data centers. Large language model training, AI inference, and high-performance computing are all driving the deployment of more GPU servers.

As server power continues to rise, one practical issue is becoming increasingly important:

How can a data center remove large amounts of heat from high-density computing equipment in a stable and efficient way?

Traditional data centers have relied mainly on air cooling. However, as rack power density increases, air cooling becomes more difficult to manage.

Higher rack power requires greater airflow, more fan power, and more complex hot-aisle and cold-aisle design.

For this reason, liquid cooling and water-based cooling systems are becoming more common in high-density data centers.

Once a data center begins to use large chilled water systems, another issue becomes important: how cooling capacity is produced, stored, and delivered.

This is where chilled water storage tanks and thermal energy storage systems begin to play a larger role.

At the same time, large chilled water tanks do not operate at one uniform temperature. Temperature stratification develops inside the tank, creating a need for multi-point temperature monitoring.

Why Are AI Servers Increasing the Need for Water Cooling?

Compared with traditional servers, AI servers usually have much higher power density.

GPU servers can continuously generate large amounts of heat during model training and inference.

In lower-density data centers, cooling can often be handled through room air conditioning, fan systems, and controlled airflow.

As rack power increases, however, relying only on air becomes less efficient and more difficult to manage.

Water and other liquids can transport heat much more effectively than air. This allows heat to be removed more directly from CPUs, GPUs, and other high-power components.

Common cooling systems used in modern data centers include:

• direct-to-chip liquid cooling;
• coolant distribution units, or CDUs;
• facility water systems;
• chilled water loops;
• cooling towers;
• dry coolers.

Heat is first removed from the server components by cold plates and coolant. It is then transferred through CDUs, heat exchangers, and chilled water systems before being rejected to the outside environment.

For this reason, the growth of AI computing is not only increasing demand for GPUs and servers.

It is also increasing demand for:

cooling equipment, heat exchangers, chilled water systems, and thermal energy storage systems.

If a Data Center Already Has Chillers, Why Is a Chilled Water Storage Tank Needed?

This is a common question in large data center cooling projects.

A chiller can already produce cold water.

Why add a large chilled water storage tank?

The reason is simple:

Cooling energy does not always need to be produced at the exact time it is consumed.

A chilled water storage tank allows a facility to produce and store cooling capacity during lower-load periods.

That stored cooling can then be used later when cooling demand increases.

In simple terms, a chilled water storage system works like a form of energy storage.

A battery stores electrical energy.

A chilled water tank stores thermal energy in the form of cooling capacity.

1. Managing Cooling Load Changes

AI server loads are not always constant.

When large groups of GPUs begin training or inference workloads at the same time, IT power consumption can increase quickly.

More electrical power means more heat.

That means the cooling system must also respond to a higher load.

If all cooling must be provided directly by the chillers at every moment, chillers, pumps, and electrical infrastructure may need to be sized around higher peak loads.

With chilled water storage, the system can use:

chillers + stored cooling

to support part of the peak cooling demand.

This gives the cooling system more flexibility when server loads change.

2. Reducing Cooling Power During Peak Periods

Large chillers consume a significant amount of electricity.

If AI servers are already operating at high power while the cooling plant is also running at maximum output, the facility may experience:

high IT power demand + high cooling power demand

at the same time.

A thermal storage system can produce chilled water during lower-load periods.

For example:

cooling energy can be stored at night or during periods of lower demand,

then discharged during higher-load periods.

This can reduce part of the cooling power demand during peak operating hours.

For large data centers, this can be useful for energy management and peak demand control.

3. Reducing Peak Chiller Capacity Pressure

A data center may operate at a moderate cooling load for most of the day but experience higher cooling demand for shorter periods.

If all equipment is sized only for the maximum short-term load, the cooling plant may require larger chillers, pumps, and electrical capacity.

A chilled water storage system can allow chillers to handle the base cooling load while stored cooling supports part of the peak demand.

Actual system design depends on many factors, including:

• IT load profile;
• cooling load;
• electricity pricing;
• redundancy requirements;
• data center design level;
• operating strategy.

Chilled water storage is not necessary for every project, but it can provide additional operational flexibility in large centralized chilled water systems.

4. Providing Cooling Buffer Capacity

AI data centers require high levels of operating continuity.

Servers may have UPS systems and standby generators, but cooling must also remain available.

If server power remains available while cooling is interrupted, high-power GPUs can continue generating heat.

A chilled water storage tank can provide a certain amount of cooling buffer during:

• equipment switching;
• system adjustments;
• partial equipment failure;
• temporary chiller interruptions;
• short-term cooling load changes.

A chilled water storage tank does not replace a complete backup cooling system.

However, it can become an important supporting part of the overall cooling strategy.

Does Every AI Data Center Need a Chilled Water Storage Tank?

No.

Cooling designs vary significantly from one data center to another.

Some high-temperature liquid cooling systems can reject heat through dry coolers or other forms of free cooling, reducing dependence on conventional chilled water systems.

Therefore, it would be inaccurate to say that every AI data center requires chilled water storage.

The value of thermal storage is generally greater in projects that have:

• large AI computing loads;
• centralized chilled water systems;
• high cooling demand;
• noticeable peak load variation;
• peak electricity management requirements;
• high cooling continuity requirements;
• a need to optimize chiller operating schedules.

As AI data centers continue to increase in scale, these types of applications are likely to receive more attention.

Once a chilled water storage tank is operating, one important question is:

How much useful cooling capacity is still available inside the tank?

Large chilled water storage tanks often rely on natural thermal stratification.

Under normal operating conditions, the tank typically develops three main zones:

Upper section: warmer water

Middle section: thermocline

Lower section: colder water

Because water density changes with temperature, properly designed inlet and outlet systems can help maintain relatively stable separation between warmer and colder water.

This temperature stratification is important to the performance of the storage tank.

If warm and cold water mix excessively, the usable cooling capacity of the tank can decrease.

Why Is One Temperature Sensor Not Enough?

Consider a chilled water storage tank that is 20 meters high.

The water temperature at the bottom may be 5°C.

At the top, it may already be 12°C.

If only one temperature sensor is installed near the bottom, it can confirm that cold water is still present.

But it cannot show how high the cold water layer extends.

If only the upper section is monitored, the condition of the lower cold water layer remains unknown.

For this reason, a single temperature value provides limited information.

What really matters is:

the temperature distribution from the top of the tank to the bottom.

What Is the Thermocline in a Chilled Water Storage Tank?

The thermocline is the transition zone between the cold water layer and the warmer water layer.

Within this zone, temperature changes gradually over height.

For example:

This type of temperature profile clearly shows the warmer upper section, the colder lower section, and the transition zone between them.

During charging and discharging, the thermocline position moves through the tank.

Monitoring this movement is important for understanding how the storage system is operating.

Why Does a Chilled Water Storage Tank Need Multi-Point Temperature Monitoring?

The most direct purpose of multi-point monitoring is to create a vertical temperature profile of the storage tank.

By monitoring temperatures at different heights, operators can observe:

• cold water layer height;
• warm water layer height;
• thermocline position;
• thermocline thickness;
• charging process;
• discharging process;
• excessive mixing;
• abnormal temperature zones.

These temperature data can also be transmitted to PLC, BMS, or SCADA systems for further analysis and control.

Why Use a Multi-Point Temperature Sensor Cable?

If a 20-meter or 30-meter chilled water storage tank requires temperature monitoring at many different heights, installing separate temperature probes can create a large number of:

• sensors;
• cables;
• connectors;
• mounting points;
• wiring connections.

As the number of measuring points increases, installation and maintenance become more complicated.

A practical alternative is a:

multi-point temperature sensor cable.

Multiple temperature sensors are integrated into a single cable, which is installed vertically inside the tank.

For example, a 20-meter temperature cable can be designed with:

• 10 temperature points;
• 15 temperature points;
• 20 temperature points;
• or another quantity according to project requirements.

Sensor spacing can also be customized.

Typical options may include:

0.5 m / 1 m / 2 m

or other intervals according to the required monitoring resolution.

This makes it possible to collect temperature data from multiple tank heights using a single cable assembly.

1. Number of Temperature Points

The required number of sensors depends mainly on:

tank height;

expected temperature stratification;

control requirements;

required temperature resolution.

A taller tank will generally require more measuring points.

2. Sensor Spacing

If the project requires detailed thermocline monitoring, smaller sensor spacing may be more suitable.

For basic stratification monitoring, a larger spacing may be sufficient.

For this reason, chilled water storage tank temperature cables often need to support customized:

cable length and sensor quantity.

3. Waterproof Performance

The cable may remain immersed in water for long periods.

Important factors therefore include:

cable jacket waterproofing;

sensor sealing;

connector sealing;

long-term immersion stability.

4. Measurement Consistency

For chilled water tank monitoring, the accuracy of each individual sensor is important.

However, consistency between different sensors is also important.

The system often compares temperatures between adjacent heights to evaluate thermal stratification.

If individual sensors have large measurement differences, thermocline analysis may become less reliable.

5. Data Communication

Large data centers usually need to integrate tank temperature data into a central monitoring system.

Depending on the project, the temperature monitoring system may support:

RS485;

Modbus RTU;

data acquisition units;

PLC;

BMS;

SCADA.

This allows the tank temperature profile to become part of the overall cooling management system.

Temperature Monitoring?

The relationship is easier to understand when viewed as part of the data center cooling chain.

AI computing growth increases server power.

Higher server power creates more heat.

Higher heat density increases the need for liquid cooling and water-based cooling systems.

Larger chilled water systems create greater demand for efficient cooling production, distribution, and storage.

Some large data centers use chilled water storage tanks for:

• cooling energy storage;
• peak load management;
• energy optimization;
• cooling buffer capacity.

Once a chilled water storage tank is operating, it becomes necessary to understand the internal temperature stratification.

A single temperature point cannot fully describe the condition of the tank.

For this reason, temperature monitoring at multiple heights becomes increasingly valuable.

This is one of the main reasons why multi-point temperature sensor cables for chilled water storage tanks are becoming relevant to modern data center cooling systems.

The impact of AI is not limited to chips and servers.

As data center computing capacity increases, power supply, cooling, liquid cooling, chilled water systems, and thermal energy storage infrastructure must also develop.

For data centers using large centralized chilled water systems, chilled water storage tanks can support peak load management, cooling buffer capacity, and more flexible system operation.

However, a large storage tank does not have a uniform temperature.

To understand its actual operating condition, it is necessary to monitor the temperature profile from the top of the tank to the bottom.

A multi-point temperature sensor cable can provide temperature data at several heights using a single vertical cable assembly.

These data can help operators monitor thermal stratification, thermocline movement, charging and discharging conditions, and overall chilled water storage performance.

As AI data centers continue moving toward higher power densities, data center liquid cooling, chilled water thermal energy storage, and multi-point tank temperature monitoring are likely to become increasingly important parts of the supporting infrastructure.

FAQ

1. What Is a Chilled Water Storage Tank?

A chilled water storage tank stores cooling energy in the form of low-temperature water.

The system can produce chilled water during lower-load periods and use the stored cooling later when cooling demand increases.

2. Why Do AI Data Centers Use Water Cooling?

AI servers, particularly GPU servers, can have very high power density.

Water and other liquids can remove heat more efficiently than air, making liquid cooling more suitable for high-density computing environments.

3. Does Every AI Data Center Need a Chilled Water Storage Tank?

No.

The decision depends on the cooling architecture, data center size, cooling load, electricity pricing, redundancy requirements, and operating strategy.

Chilled water storage is generally more relevant to large facilities using centralized chilled water systems.

4. Why Is One Temperature Sensor Not Enough for a Chilled Water Storage Tank?

Large chilled water storage tanks usually have temperature stratification.

Water temperature at the top, middle, and bottom of the tank can be significantly different.

A single sensor cannot show the cold water layer height or thermocline position.

For this reason, multi-point temperature monitoring is usually more useful.

5. What Is a Thermocline?

A thermocline is the temperature transition zone between the cold water layer and the warmer water layer inside a stratified storage tank.

Its position and thickness can help indicate the operating condition of the storage tank.

6. Can a Chilled Water Storage Tank Temperature Cable Be Customized?

Yes.

Depending on the project, the following parameters can be customized:

total cable length;

number of temperature sensors;

sensor spacing;

signal output;

data acquisition method.

For example, different sensor layouts can be designed for 20-meter, 30-meter, or other tank heights.

7. Can Tank Temperature Data Be Connected to a PLC or BMS?

Yes.

Using a temperature acquisition unit and industrial communication interfaces such as RS485 or Modbus, multi-point temperature data can be transmitted to:

PLC systems;

BMS platforms;

SCADA systems;

other cooling management systems.