Liquid cooling in modern data centers serves as an alternative to traditional air-based cooling. The latter is indeed no longer sufficient to manage and dissipate the vast amounts of heat generated by chips, servers, and CPUs—heat resulting from the increasingly high computing densities required by cloud computing and artificial intelligence applications. As a cooling medium, liquid offers a superior heat transfer coefficient compared to air, allowing for more efficient and rapid heat removal, while consuming less energy than air-conditioning systems.
Another compelling aspect that further enhances the energy efficiency of liquid cooling solutions in data centers is the potential for implementing heat recovery systems. Because liquid absorbs heat more efficiently and in a more concentrated manner than air, the temperatures at the cooling system’s outlet are higher than those achieved with conventional air-based methods. This facilitates the adoption of highly advantageous energy recovery solutions, enabling the reused thermal energy to be applied to a variety of purposes—ranging from district heating and heat pumps to the heating of domestic hot water or the preheating of process fluids. In this way, the energy consumed by data centers—which would otherwise be dissipated as waste heat—is repurposed in line with circular economy principles, thanks to liquid cooling.
Tempco acts as a qualified system integrator for the implementation of liquid cooling skids for data centers, featuring brazed plate heat exchangers (BPHEs) as their technological core. There are various possible architectures for liquid cooling ecosystems in data centers (that often can also include chillers, free cooling and evaporative towers), and for each type, Tempco’s range of BPHEs offers the ideal solution to ensure maximum cooling efficiency and reduced energy consumption, thereby lowering the Power Usage Effectiveness (PUE) of computing facilities.
It is therefore worth examining the advantages—as well as some related drawbacks—associated with different liquid cooling options for servers and chips. Whether using cold plates (conductive metal plates placed in contact with electronic components to remove generated heat via a coolant circulating through microchannels) or direct immersion cooling (where servers are submerged in dielectric fluids), there are two primary operating modes for the thermal cycle: single-phase liquid cooling and two-phase liquid cooling. The former involves no phase change of the coolant—which could, for instance, be a 75% water-glycol mixture—offering the advantage of a simpler system design.
In two-phase liquid cooling, the coolant boils either within cold plates or upon direct contact with components submerged in cooling tanks. The transition to a vapor phase necessitates a more complex system to manage the vapor, requiring the inclusion of condensers as well as precise control over pressure and temperature levels. Conversely, two-phase liquid cooling enables the removal of large quantities of heat with significantly higher efficiency than single-phase methods.
Another aspect involves the comparison between cold plates and immersion liquid cooling. The use of cold plates requires pumping systems and piping to deliver coolant to the plates positioned on the components, which introduces a risk of leaks that must be carefully monitored. Immersion liquid cooling, by contrast, offers a more streamlined system architecture; it facilitates more uniform cooling and simplifies inspection and maintenance—essentially requiring only that the level of the coolant or dielectric fluid in the cooling tank be kept sufficient. On the other hand, despite its simpler system layout, immersion liquid cooling entails higher initial costs, driven by the high price of the specialized dielectric fluids required and the need for custom-built hardware, which must be constructed using materials compatible with immersion in special cooling liquids (requiring, for instance, careful selection of appropriate adhesives and gaskets).