Jonathan Blake, VP and Director, TMT EMEA, HDR, shares five key factors that are readying the region to challenge North America and prepare the next generation of data centre activities.
The total volume of data is growing exponentially, driving significant appetite in the global data centre market – which is forecast to reach €235 billion by 2026. While the US remains a leader, the number of data centres across Europe has seen a considerable jump. Rapidly advancing technologies, such as AI, Machine Learning (ML) and the Internet of Things (IoT), are increasing demand, which is likely to continue outstripping supply for the next three to five years.
These and other factors, including the race to decarbonisation and a shortage of power, are shaping the future of the data centre industry, with several of the trends interlinked.
A growing focus on AI
The prevalence of AI, which requires much higher computational power, is growing hand-in-hand with the increasing pressure for data storage capacity. This is transforming data centre design in different ways, from introducing just a few AI-based racks to entire AI-enabled centres. In response to this trend, hyperscalers will increase their rack density to an estimated 50kW per rack by 2027 yet with greater densities already being possible.
We have yet to see a move to AI in Europe in the same way as in the US, largely because of Europe’s power supply issues. Making data centres AI-ready means ensuring they have enough capacity and can cater to the larger power and cooling needs of AI. This was a priority when HDR worked on the AI-ready IC3 Super West data centre in Sydney for Macquarie Data Centres. The facility will use liquid cooling, as well as air cooling, harnessing advanced monitoring systems and holding a total of 45MW.
Data storage requirements will continue to grow, and AI will also be applied to automation, monitoring and optimising energy efficiencies in data centres, as well as quickly identifying faults.
Water use in cooling systems
For data centres, the cooling process alone can take between 20% and 40% of their overall energy consumption, depending on location, and there is a concerted push to reduce water usage (WUE) within that. Historically, the drive was to cut power usage (PUE). Direct air with Adiabatic cooling helps drive power efficiencies yet requires significant water usage during the warmest months. Due to increased rack densities, air cooling alone is insufficient to meet the cooling needs of the denser racks.
As the industry moves away from air-only cooling solutions, the challenge is to choose the most efficient, cleanest versions of liquid cooling – and this is likely a Direct-to-Chip solution when building at scale. Immersion cooling may have a place; however this is likely to be restricted to smaller scale specialist facilities. All liquid cooling solutions can remove heat more effectively while decreasing power consumption. Another benefit is that, although such systems utilise water, they are closed loop, thereby reducing the need for Adiabatic cooling and thus reducing water usage overall.
Reducing embodied carbon
One of the biggest causes are the structures themselves, so tackling embodied carbon by improving the design and construction process is the best place to start. This includes making smart materials choices and embracing a whole lifecycle approach. Steel and concrete are two major offenders when it comes to embodied carbon. Reusing steel where possible is far more environmentally friendly and some of the steel fabricators are already facilitating this approach to material reuse as opposed to recycling.
Timber data centres are an interesting proposition, and HDR has already been working on live projects where this is being proactively developed as an option. HDR has utilised mass timber extensively in other facility types including laboratories, offices and logistics buildings.
From a compliance perspective, regulation and industry commitments are putting embodied carbon front and centre, such as the Structural Engineers 2050 Commitment Programme, the Architecture 2030 Challenge for Embodied Carbon, as well as MEP 2040. Europe’s Climate Neutral Data Centre Pact, meanwhile, is an agreement by major operators to be carbon-free by 2030, water conservation being one of the goals.
We will see greater focus on reducing the carbon footprint of data centres and supporting the circular economy, by exploring how facilities are built and how they can be effectively dismantled and adapted going forward.
On-site energy and backup power
The demand for more data centre space and capacity is far exceeding the existing grid capacities in many European countries. In certain locations, therefore, attention is shifting to on-site power generation initially from ‘transitional’ fuels such a natural gas. However, the hope, eventually, is that this will transition to ‘green’ hydrogen supplemented by ever increasing capacities from renewables with the potential for greater battery storage to provide continuity.
Reusing waste heat from data centres is another key consideration. Air-cooled data centres do not produce good quality waste heat so, again, as more data centres inevitably transition to closed loop liquid cooling this brings advantages for the use of high-grade waste heat. Infrastructure investment outside of the data centre site is therefore key to allowing local communities to benefit from this source of energy.
As an example, our team partnered with Swansea Council, DST Innovations and Batri on the delivery of the large-scale green energy and transport hub and port development in Swansea. The project brings together eco-homes with a tidal lagoon (producing tidal energy), solar farm, battery manufacturing plant (storing power generated on-site) and hyperscale data centre, whose waste heat can be integrated into a new local district heating system.
Size and location changes
As AI data centres come into their own, this will influence locations. While cloud services need to be near large urban areas for latency reasons, there is currently no such concern with AI, so these data centres can potentially be built in areas where the climate is more favourable, or where there is more land and power availability. This will kick-start a move away from urban locations and companies will explore locations that have not been considered previously.
Interestingly, looking even further ahead, this trend may well reverse once more as growing technologies like autonomous vehicles will mean that more Edge-type data centre hubs will be needed closer to urban environments to ensure strong connectivity. Moreover, due to land shortages and high costs, we will see more multi-storey centres that maximise available space.
