According to a recent survey conducted by the Uptime Institute, only one in four respondents think that air cooling will remain the dominant form of data centre cooling infrastructure beyond the next decade in data centres larger than one megawatt (MW) in size. While some see immersion cooling as a solution looking for a problem, it is increasingly being lauded as the next advance in cooling, able to meet demand from an industry which is constantly striving for greater capacity.
We spoke to Maikel Bouricius, chief commercial officer at Asperitas, to find out why he believes an immersion cooling revolution is inevitable.
Hi Maikel, nice to meet you. Let’s kick off by asking about how immersion cooling compares to air cooling in terms of energy efficiency?
Hello, nice to meet you too. Well, there are a number of great advantages to immersion cooling. One of them is that it eliminates the need for fans. When we first started out, estimates suggested that up to 20 per cent of server power consumption was dedicated to the fans. Obviously, Open Compute has driven improvements, but that energy consumption is still probably somewhere between five and 15 per cent. So, to remove that energy consumption entirely via immersion cooling is a great win, and, overall, immersion cooling can lead to up to 95 per cent of energy reduction on the cooling side of a facility’s PUE.
We like to be clear and transparent at Asperitas, however, and it must be said that, while we can position immersion cooling in a way where those numbers are really eye-catching, the reality is that is mostly being applied in existing data centres. Retrofitting immersion cooling in a facility which already has the investment and infrastructure in place probably means that the impact that immersion cooling makes is lower overall.
For me, however, it is best to start with the quick wins. When the average global PUE is roughly 1.6, and immersion cooling can reduce that to below 1.1 with ease, then I think there is no better efficiency approach than immersion cooling in today’s market.
How do the CapEx and OpEx for immersion cooling compare to those of air cooling?
It depends on the case. If you were comparing immersion cooling in a small-scale environment to a large-scale data centre primarily run-on air-cooled systems, then the case will be very different. Generally, we tell people that a realistic return on investment for immersion cooling is below three years, based purely on operational energy reduction. Considering what is happening within the energy sector and through global supply chains at present, that estimate is only really improving and will continue to improve over time, I think.
Of course, that also does not take into account potential savings from greenfield data centres that have been designed and optimised for immersion cooling; the potential upside for their CapEx and OpEx is huge, and we have seen examples where they have been slashed in half.
We are not quite at the stage where there are a large number of users that have optimised their facility for the technology yet, but I think we will get there for sure.
What other aspects does immersion cooling excel at?
I think sustainability is one of the major positive factors for immersion cooling. We already covered reducing energy consumption, but reducing concrete usage is another big factor, as concrete can have a huge impact on a facility’s carbon footprint when you take into account embodied carbon and Scope 3 emissions. Optimised and dedicated immersion-cooled facilities can be much more minimalistic in design, so that is a big advantage.
Additionally, water usage is becoming a major topic of conversation in the industry, particularly in certain regions of the world where water is becoming a scarce resource. Immersion cooling can contribute to reducing water usage because it is a closed-loop system.
What areas are you working on to improve immersion cooling?
I think sustainability, despite the obvious benefits which we just talked about, is something that we hope to continuously improve on. There are also challenges to do with large-scale adoption, making this a scalable and flexible solution, as well as integration, and so a lot of what we do is focused on the user experience – how users bring immersion cooling into a data centre from a logistical point of view, how is it serviced and maintained, etcetera.
How realistic is it that data centres rely solely on immersion cooling? Or, is a combined implementation alongside air cooling more realistic?
I think it is more realistic to see a hybrid form of cooling at present, for the simple reason that colocation providers, the largest part of the market, are not quite at the stage where they are willing to make big bets on immersion cooling. Dedicated data centres, however, are doing their research and figuring out that, for their facilities, immersion cooling is a sensible choice, taking into account their sustainability roadmaps, and with one eye toward next-generation hardware – CPUs and GPUs – which are set to increase the need for cooling even further.
How do you see the future of data centre cooling?
When we started out in 2017, there was not much in the way of research or tracking of the market for immersion cooling, but we are seeing a big shift toward that environment now. I think, within the next five to ten years, we will continue to see a shift from air cooling to liquid cooling. Of course, there are other innovations out there – cold plates, for example, are great for static environments, supercomputers, and so on – but I think single-phase immersion cooling is probably the solution of choice when looking for environmentally friendly, scalable solutions.
One of the big drivers for this is an emphasis on new standards and regulations, particularly in Europe. The Sustainable Digital Infrastructure Alliance (SDIA) have done some great work in laying down a roadmap for the environmental side of the industry, and immersion cooling can help hit many of the markers on that roadmap. Within the next two years, as more facilities opt to adopt immersion cooling, we will see more evidence of this, I am sure, and I am looking forward to seeing how things progress.
