Liquid Cooling for Growing AI Platforms

Female on laptop with icons
19 Sep 2025

AI is driving transformative change across both technology and the wider world.

According to the International Data Corporation (IDC), global spending on AI is expected to more than double by 2028, reaching USD 632 billion. This rapid growth, driven by strong investments in both AI software and supporting infrastructure, reflects a compound annual growth rate (CAGR) of 29% over the 2024–2028 period.

As AI adoption expands, so too does the thermal output of the powerful hardware it relies on. High-performance computing components, particularly GPUs and other AI accelerators, generate significantly more heat than conventional IT equipment. This poses a growing challenge for data centre operators, who must ensure efficient thermal management to protect infrastructure and maintain performance. 
GPU chip

Liquid Cooling Offers a Highly Effective Solution 

Liquid cooling is emerging as a leading solution, offering clear advantages over traditional air-based systems in handling dense, high-power platforms. 
 
Unlike conventional cooling methods that use air to dissipate heat, liquid cooling uses fluids such as water or specialised coolants to absorb and transfer heat away from critical components. Water’s specific heat capacity (~4.186 kJ/kg·K) is more than four times that of air (~1.005 kJ/kg·K), allowing it to move heat far more efficiently.

Several liquid cooling technologies are available today: 

  • Rack-based liquid cooling - that circulates coolant through the racks, extracting heat directly from the servers. 
  • Liquid-to-chip cooling - which delivers coolant directly to the heat-generating components, such as GPUs, via cold plates mounted on the hardware. 
  • Immersion cooling - where entire servers are submerged in a thermally conductive, non-electrically conductive liquid, enabling heat management across all components. 

By targeting heat at its source, liquid cooling enables data centres to support greater compute density and efficiency, which is essential for AI applications.

 

Hybrid Approaches for Balanced Efficiency 

Despite its advantages, liquid cooling is not yet a complete replacement for air cooling. Even with liquid systems in place, residual heat will escape into the surrounding environment. As a result, some form of air cooling is still required to maintain optimal conditions within the data hall. 

For many data centres, a hybrid cooling strategy which combines liquid and air systems delivers the best results. This approach ensures high thermal performance while also optimising Power Usage Effectiveness (PUE) and reducing energy consumption across the facility. 
 
 

Looking Ahead 

As AI continues to evolve and demand for compute power increases, so too does the need for more efficient, scalable cooling solutions. Liquid cooling technologies offer data centres the performance and energy efficiency required to support next-generation workloads, while hybrid strategies ensure practicality and adaptability for existing infrastructure. 

With careful planning and the right cooling architecture, data centres can future-proof their operations and stay ahead in the AI-driven era. 
 
At Colt DCS, we are AI-Ready. All our new hyperscale data centres in development have been designed to use a hybrid cooling approach using both liquid-to-chip cooling and traditional air. For more information on how our facilities are designed to handle AI applications, click here

Contact us

If you would like to know more about Colt Data Centre Services, please get in touch with our team. We would be delighted to support your digital infrastructure requirements.
Contact us