Liquid cooling enables very high heat loads to be dissipated and also facilitates the use of heat recovery in the data centre. In this interview, Valeria Mercante, the Team Lead of STULZ’s Liquid Cooling division, explains the fundamental issues that need to be considered before liquid cooling solutions are integrated, what technical and regulatory constraints also need to be factored in, and what trends are currently preoccupying the data centre market.
What fundamental issues should data centre operators consider before they decide to integrate liquid cooling infrastructure?
Several key factors need to be taken into consideration when integrating liquid cooling systems. Firstly, the actual motivation for converting to liquid cooling is key. As well as increasing power density, it can reduce the amount of space required while simultaneously increasing cooling capacity – and sustainability management can also be an overriding factor. First of all, they need to consider whether this is a greenfield or a brownfield project. If it’s a retrofit, this generally puts major constraints on the choices available. In this case, basic compatibility with the existing cooling infrastructure and the specific requirements for the IT equipment must be clarified as the first step. Furthermore, structural aspects such as load-bearing capacity, ceiling height and the transmission of vibrations also play a part.
What key prerequisites have to be met in order to successfully install a liquid cooling infrastructure?
For successful planning, a lot of detailed information is needed. This includes, first and foremost, precise requirements concerning the IT load, such as the expected heat load per rack. The specifications of the server hardware and associated cold plates are also important, and include the refrigerant temperatures and flow rates required for operation. Where the cooling liquid is concerned, decisions also need to be made about whether to use water or dielectric fluid. With an existing infrastructure, parameters relating to the cooling water circuit, such as temperature windows and the liquid used, must also be incorporated in planning. When choosing cooling units, data on the required cooling capacity per unit, the desired level of redundancy for pumps and cooling units, and the necessary external loss of pressure are also relevant considerations.
What are the main advantages of using liquid cooling as compared to conventional air cooling?
The use of liquid cooling can bring considerable benefits, especially in terms of managing high power densities of over 50 kW per rack. Moreover, much higher cooling capacities of up to 250 kW are achievable with liquid cooling. Another advantage is that the cooling process itself fundamentally becomes more energy efficient, because cooling with liquid is around 50 to 1000 times more efficient than cooling with air. This is a tangible factor that can also contribute to a much lower PUE value. What’s more, liquid cooling enables servers to run at higher operating temperatures, which can potentially facilitate heat recovery and help more efficient use to be made of free cooling. The reduced use of air cooling in high-density areas can also lead to a reduction in space requirements in the data centre. In some cases, such as with direct-to-chip cooling, liquid cooling is actually essential for certain AI high-performance components, such as Nvidia Blackwell 200.
What challenges are currently facing the implementation of liquid cooling systems?
Despite its many advantages, there are a few hurdles to overcome when setting up liquid cooling systems. As an example, one major challenge is the lack of uniform market standards. At present, we can see a trend towards operating with lower refrigerant temperatures, as is often demanded by hardware and server manufacturers. However, that is often in conflict with the desire for greater energy efficiency and the effective deployment of heat recovery. Finding a balance between energy efficiency at low inlet temperatures, high reliability during peaks in demand, and the requirements of the IT hardware itself is a vital part of the decision-making process. Pump redundancy and the necessary pump pressure on the TCS side impact on energy efficiency and operating costs. Economic factors also have to be examined, as well as purely technical considerations. If we are not expecting any extreme power densities of over 100 kW for an application, there are cheaper alternatives to direct-to-chip cooling, such as active rear door heat exchangers, for instance. These are suitable for power densities of up to about 30 kW per rack.
What are the most important developments and trends that are currently emerging in the field of liquid cooling?
In general, we expect to see healthy market growth in liquid cooling in the data centre sector, particularly in the hyperscale segment. Also, direct liquid cooling solutions are expected to gain a bigger share of the market than immersion cooling models, with a ratio of roughly 80 to 20. Here at Stulz, for example, we are currently further developing our CDU product range to create more powerful versions for use in gray space with higher pump pressure. In view of ever more stringent sustainability requirements, water-cooled chillers combined with liquid cooling are gaining in importance, especially in urban environments with stricter noise protection standards. Heat recovery from liquid cooling systems and its use in heating networks or other fields such as plant cultivation remains a vital trend, as does the continuing development of general standards and guidelines for liquid cooling.
So what role do reliable standards play in the implementation of liquid cooling?
Today, the market for liquid cooling is still in the development phase, and as such there are as yet no comprehensive market standards covering redundancy, security, or other aspects. The consequence of this is that many of the systems currently being implemented are very customer-specific and demand close collaboration between data centre operators and suppliers of the technology. This individualisation is also driven by different customer and equipment requirements, as well as the specific cold plate requirements of the server manufacturers themselves. On the other hand, we are also seeing a clear move toward greater standardisation. In addition, the development of service corridor versions of CDUs aims to create more flexible opportunities for use overall. As the market grows in maturity, we can also expect to see greater standardisation as regards interfaces, security protocols and redundancy concepts in future. This would make the planning and implementation of liquid cooling systems easier in future, and it’s highly likely that it would also lower the cost of implementation. Until then, however, a combination of standardised components and individual adaptations will remain common practice.
To what extent do regulatory constraints influence the decision for or against the use of liquid cooling?
Regulatory constraints are playing an increasingly important role in the choice of a particular cooling system. Indirectly, the F-gas Regulation, with its clear timetable for restricting refrigerants with a high global warming potential (GWP), can favor the use of liquid cooling, because no or less F-gas is employed in processes using water or other liquids. However, requirements such as those contained in the German Energy Efficiency Act have a more direct impact. This limits the PUE values of new data centres to 1.2 maximum from the year 2026. If these values are to be achieved – especially alongside increasing power densities – liquid cooling is set to grow in importance, as it enables more efficient heat dissipation than conventional air cooling.
And even though the new refrigerant regulation has been defined for the European market, it also has an impact on exports to Africa. The regulation stipulates that many devices exported from 2027 onwards must have a GWP of < 1000. Refrigerants such as R134a, R410a, and R407c will then no longer be permitted for export to Africa.
And last but not least, ambitious corporate sustainability targets and international commitments such as the Paris Agreement on climate change are also factors that will induce data centre operators to lend greater consideration to more energy-efficient liquid cooling-based solutions in future.
