The Rise of Liquid Cooling in Data Centres: A Guide to High-Density Thermal Management
Data centres are essential to the digital economy and enable services such as online banking and e-commerce. Their energy consumption is rising rapidly – 7.4 gigawatts in 2023, accounting for 4% of global consumption – with demand in Europe expected to more than double by 2030 due to AI and digitalisation. An uninterrupted power supply and cooling are essential to prevent outages and data loss.
Optimising HVAC systems for the sustainability of data centres
Heating, ventilation and air-conditioning (HVAC) systems play a crucial role in the reliable operation of data centres. However, it is estimated that cooling alone accounts for between 20% and 50% of a data centre’s total energy consumption.
Given this significant proportion of energy spent on thermal management, today’s HVAC solutions must do more than simply regulate temperature. They must be highly efficient and scalable, and capable of supporting increasingly power-intensive IT environments – all in line with CO₂ reduction targets and sustainability requirements.
As global demand for digital infrastructure grows, data centres are under increasing pressure to deliver more computing power with a lower environmental footprint. One development supporting this trend is liquid cooling – a technology that, whilst not new, is now gaining significant traction as a practical and scalable option.
The benefits of liquid cooling
The emergence of liquid cooling comes at just the right time. Although the technology has been around for years, it is now experiencing a phase of rapid adoption, driven by the limitations of air cooling and the rising demands of high-performance computing (HPC), machine learning and hyperscale data operations.
Conventional air-based systems are gradually reaching their physical and economic limits. Server racks with a power consumption of over 30 kW are becoming increasingly common, yet the upper limit of air cooling is close to this figure. Liquid cooling , on the other hand , enables far more efficient heat transfer and supports rack densities of 50, 80 or even 100 kW without requiring a massive expansion of infrastructure or power consumption.
Cooling distribution units (CDUs) circulate a glycol-water mixture through closed circuits and, via cooling plates and distributors, extract heat directly from the most critical and thermally intensive parts of the server, such as chipsets.
Some CDU systems can provide up to four times as much cooling capacity as air-based alternatives. Typically, air can cool a rack density of 25–30 kW at acceptable temperatures. However, as racks now reach 100 kW and above, CDUs are designed to operate at these densities.
Air cooling vs. liquid cooling: complementary, not competing
Although liquid cooling represents a transformative leap in thermal management, it does not yet fully replace air-based systems. Instead, a hybrid approach is increasingly gaining ground. Many data centre operators employ a mix of cooling strategies depending on workload requirements, with liquid cooling handling high-density or GPU-intensive racks and air cooling supporting more conventional equipment.
These integrated systems can be dynamically controlled via intelligent building management systems and software platforms, enabling real-time adjustment of CDU flow rates, cooling capacity and airflow. This level of coordination can be efficient and help operators reduce their energy efficiency (PUE) to around 1.2, with some aiming for even more ambitious figures.
Heat recovery and reuse: turning waste into energy
One of the unique advantages of liquid cooling is its potential for heat recovery and reuse. The warm fluid leaving the server chipsets, which typically has a temperature of around 30 °C, can be passed through a heat pump and raised to a temperature suitable for district heating networks or industrial applications.
This reuse can help to:
- reduce overall energy consumption
- improve the PUE
- generate added value from what was once considered waste
Pushing the boundaries of cooling efficiency
As liquid cooling matures, innovation is shifting towards improving performance within the system itself. A key area is the reduction of the approach temperature – the temperature difference caused by inefficiencies in heat exchange. Whilst some CDUs may have a temperature spread of 4 °C, Carrier’s advanced dual heat exchanger design reduces this to just 2 °C. The integrated, state-of-the-art controls also enable a rapid response to fluctuating IT loads whilst maintaining acceptable temperature variations.
Looking to the future, further reductions to 1 °C are on the horizon, offering even greater efficiency and laying the groundwork for supporting multi-megawatt installations with minimal thermal overhead. As the scale and density of data centres continue to grow, such refinements will be crucial to meeting future demand without proportionally increasing the environmental impact.
Future trends in data centre cooling
Direct-to-chip liquid cooling is expected to remain the dominant cooling architecture for high-performance environments over the next five to ten years. However, immersion cooling, in which entire servers are submerged in a thermally conductive, electrically non-conductive liquid, could become viable on a large scale in the long term.
In the meantime, Carrier continues to lead the way in the development, testing and practical implementation of next-generation data centre cooling technologies, including its QuantumLeap™ solution suite and the certification of its products through Eurovent Certified Performance.
Please contact one of Carrier’s experts to find out more about Carrier’s thermal lifecycle management solutions.
