Data centres, these data processing and storage centres, have become an essential pillar of the digital ecosystem. The sheer volume of data exchanged around the world inevitably passes through one of these infrastructures. Among the trends of the industry – such as acceptability and sustainability -, the wave of AI and its impact on the density of data centres are an unmissable challenge for 2024 and beyond.
Power-hungry, creating few jobs and often perceived as impregnable fortresses, data centres have been the target of criticism for many years. The significant growth of this sector – which creates between 1,000 and 2,500 new direct jobs* every year – has brought to light a major challenge for the entire industry: the issue of acceptability to both local authorities and local communities.
In addition to requirements in terms of the energy performance of their infrastructures (PUE), data centre operators must now demonstrate other “UE” – Usage Effectiveness – type parameters concerning water (WUE), carbon (CUE), complete life cycle analyses (LCA), etc.
Furthermore, the majority of players in the sector highlight their carbon impact under scopes 1 and 2, whereas the majority of their environmental impact concerns scope 3, i.e. their indirect responsibility (construction of new buildings, waste management on sites, purchases of products and services, employee travel, etc.). Reducing this scope will be a major challenge in the years to come.
Integrating into the region and the economy
A data centre campus cannot be designed in isolation. It has to be part of an industrial landscape and of a local area to ensure a sustainable future with other sectors.
The reuse of the so-called ‘waste heat’ – that is produced by data centres – is thus becoming a prerequisite for the development of new sites. Solutions exist, such as district heating, which is widely used in the Nordic countries and which Data4 plans to make available to the towns and cities in which its new sites are located, or applications dedicated to the agri-food sector (hydroponics, greenhouses, fish farms, etc.). While there are still only a small number of use cases that are technically and financially unprofitable, it is now essential to forge multi-sector partnerships to move to an industrial scale.
Programs to raise awareness of digital issues among the younger generation, through partnerships with secondary schools and engineering schools, are becoming increasingly popular, and Data4 is stepping up initiatives in this area on its European campuses.
Riding the AI wave
The demand for computing and data storage capacity continues to grow. Generative AI, such as Chat GPT, is revolutionizing the internet and data centres. AI learning models, which are very resource-intensive, are also developing rapidly.
Therefore, CPUs and GPUs [hardware units that enable computers to function] now have TDPs [Thermal Design Power] over 500W, generating racks of over 40kW! The industry has never been closer to getting supercomputers, and this time on an industrial scale. Having the capacity to rapidly deploy “HPC-ready” or “IA-ready” data centres will hence be the biggest challenge for the market to meet growing customer demands.
“It can be said that we have never been closer to getting supercomputers than we are now – and on a massive scale,” stresses Adam Ponichtera, Country Director at Data4 in Poland. “The biggest challenge for data centre operators to meet market demands will be the ability to rapidly deploy HPC and AI-ready data centres.”
Optimising energy management at our sites
Looking closer to the demand generated by AI, another challenge consists of optimizing the energy management on sites. Several options can be considered: working closely with the electricity grid operator, to reduce consumption at peak times, or using generators or storing energy on-site. Common efforts will also need to be made with cloud operators to ensure that the load is adapted to the availability of renewable energy, something that data centres cannot do on their own.
Partnerships programs between data centres and networks need to be intensified, and the winter of 2022/2023, particularly in France, clearly demonstrated the critical nature of having common efforts. Moreover, these collaborations bring in return enormous potential for ‘intelligent’ management of incoming energy.
“While hydrogen-based technologies are also promising, they are still at the experimental stage. As with renewable energies, investment plans and subsidies from governments and Europe will be decisive in scaling up to an economically viable level. Hydrogen technologies, which are in the experimental phase, also seem promising,” says Adam Ponichtera. “To make them viable, like other renewable energy sources, investment plans and subsidies from the government or the EU are necessary,” he adds.
Developing new cooling systems
Numerous projects involving new cooling methods are emerging, such as using river currents, where possible, to cool data centres, with the promise of energy savings.
As CPUs and GPUs become ever more powerful, techniques based on liquid cooling are being developed, either directly via DLC [Direct Liquid Cooling] systems with direct heat exchange on critical elements via “water blocks”, or by completely submerging the servers. Both of these techniques are very effective in terms of PUE.
In addition, the digital approach based on machine learning (AI) also enables significant gains to be made on PUE, “simply” by using the millions of data points generated by the data centre’s critical systems.
Rely on life cycle analysis for all projects
Being aware of their impact and the huge environmental challenges, data centre operators, including Data4, are banking on life cycle analysis (LCA), which has established itself as THE most comprehensive assessment method for auditing all the flows and impacts involved in the design and production of a data centre.
This approach goes beyond a simple analysis of greenhouse gas (GHG) emissions and the parameters that influence resources and the quality of the environment, as it allows us to calculate the environmental impact associated with the construction phase of a data centre and that associated with its operation.
“The aim is to provide a clearer picture of what can and must be improved, to help prioritize actions and to create an ‘environmental truth’ for all technical and technological projects, thereby avoiding the transfer of environmental impacts. Obtaining the best possible visibility in terms of carbon impact is also prompting operators to explore new avenues in the eco-design of data centres, with new themes such as the re-use of industrial sites (particularly in connection with the ZAN law) and the use of low-carbon concrete, as we are doing at Data4,” says Adam Ponichtera.
“On the one hand, we are seeing growing demand generated by the development of cloud services or artificial intelligence, while on the other hand, we are trying to achieve a lower and lower environmental footprint and increase the energy efficiency of facilities. The environment requires us to implement new technologies, but we cannot forget the local communities near the campuses. The next years in the data centre industry will certainly be interesting and marked by intensive work of specialists from all areas,” concludes Adam Ponichtera.