Introduction
In the previous blog “Why Do AI Data Centres Need So Much Power?“, we have seen that the modern data centres do not just need electricity to run computers. They also need electricity to remove heat.
Every server, processor, memory chip, storage device, and network switch inside a data centre consumes electrical energy. A large part of that energy eventually becomes heat. If this heat is not removed, the temperature inside the data centre rises, and the equipment can slow down, become unreliable, or get damaged.
This is why cooling is one of the most important parts of data centre design.
Why Do Data Centres Get Hot?
When electric current flows through electronic circuits, some electrical energy is converted into heat. This happens because real materials have resistance.
In a laptop or mobile phone, we can feel this as warmth. In a data centre, thousands of computers are working together. The heat produced is much larger.
AI data centres can generate even more heat because they use powerful processors, such as GPUs, to perform large numbers of calculations. More computation usually means more electrical power, and more electrical power means more heat.
So the basic Physics idea is simple:
More computing work means more electrical energy used.
More electrical energy used means more heat produced.
More heat produced means more cooling is needed.
Heat Must Be Removed
Cooling a data centre is not about making the room cold for comfort. It is about keeping electronic equipment within a safe temperature range.
Heat naturally flows from a hotter object to a colder place. Cooling systems use this principle to move heat away from servers.
The heat may be carried away by:
- air,
- water,
- special cooling liquids,
- or a combination of these.
The goal is always the same: move heat away from the computing equipment and release it somewhere else.
Air Cooling
Many data centres use air cooling.
In air cooling, fans push cool air toward the servers. The air absorbs heat from the equipment and becomes warmer. This warm air is then moved away and cooled again.
This method depends on good air flow.
If hot and cold air mix badly, cooling becomes inefficient. The system may need stronger fans or more cooling power. That means more electricity is used.
This is why data centres often separate cool air and hot air. Cool air is sent to the front of server racks, while hot air is removed from the back.
Liquid Cooling
As AI systems become more powerful, air cooling alone may not always be enough.
Liquid cooling can remove heat more effectively than air because liquids can carry more heat. In liquid cooling systems, a cooling liquid flows near hot components and carries heat away.
This does not mean water is poured directly over electronics. The liquid usually moves through sealed tubes, cold plates, or other controlled systems.
Liquid cooling is becoming more important for high-performance computing and AI workloads because these systems produce large amounts of heat in a small space.
Cooling Also Uses Energy
Cooling systems themselves need electricity.
Fans, pumps, chillers, cooling towers, and control systems all consume power. This means a data centre uses electricity in two ways:
- To run the computers.
- To remove the heat produced by the computers.
This is why efficient cooling matters.
If cooling is inefficient, the data centre needs even more electricity. If cooling is efficient, more of the total electricity can be used for useful computing work instead of support systems.
The Physics Connection
Cooling data centres uses many basic Physics ideas:
Heat
Servers produce heat when they use electrical energy.
Temperature
Cooling systems keep equipment within safe operating limits.
Energy transfer
Heat is transferred from hot servers to air, water, or cooling liquids.
Convection
Moving air or liquid carries heat away.
Electric power
Fans, pumps, and cooling equipment need electrical power to operate.
Efficiency
Better cooling design reduces wasted energy.
This is why data centre cooling is not just an engineering problem. It is a Physics problem too.
Why This Matters
AI is growing rapidly. More AI systems mean more computing, more power demand, and more heat.
If we only think about processors and electricity, we miss half the story. Every watt used by a server eventually becomes heat, and that heat must be removed.
So when we ask why AI data centres need so much power, cooling is one of the key answers.
A data centre is not just a room full of computers. It is also a carefully designed heat-removal system.
Next Blog
In the next blog, we will look at:
Power vs Energy: MW and MWh Explained