With the likes of ChatGPT, AI is becoming much more powerful. That calls for data centre providers to move with the times to both accommodate surging demands and remain sustainable, and Data4 for one is seeking to get ahead of the game.
Over the past year, artificial intelligence has moved to the next level. The entry onto the scene of generative AI via the likes of the now year-old ChatGPT is forecast to have the potential to add trillions of dollars in value annually to the global economy.
That’s exciting news for all industry sectors, providing the ability to drive creativity and innovation, and automate multiple work tasks. At the same time, the new wave of AI creates sizeable challenges for the data centre market, not least the additional demand for facilities, resources like server racks and better networking infrastructure.
As a consequence of rising resource use, the power consumption of generative AI is estimated to be several times that of traditional workloads, while cooling technologies will be needed to counteract the heat generated by the servers.
With such technology in its early days, now is a crucial time for data centre players to think carefully about how to accommodate such demands in the long term. Doing that will help ensure that systems and buildings are well set up to deal with the expected future surge in adoption of AI.
“We expect AI categories like natural language processing, computer vision and machine learning to really ramp up the demand for data centre solutions,” explains Adam Levine, chief commercial officer at Paris-headquartered data-centre provider Data4.
Bigger and higher-density
Data4 is one player making strides to address these future requirements. For a start, the company is seeking to procure larger campuses with higher levels of power. As examples, it owns sites with capacities of 120MW and 250MW in Paris, and a 180MW site in Frankfurt, Germany.
Inside the data centres themselves, Data4 is putting in place measures to optimise space for servers, maximising efficiency by creating higher-density facilities that support larger AI workloads.
To then provide the necessary cooling for this density of servers, Data4 is using innovative liquid and immersive technologies. The company has, for instance, widely employed free cooling at its sites. Via this technique, hot indoor air is replaced with lower-temperature outdoor air rather than using a regular cooling system, significantly improving power usage effectiveness (PUE).
There is also a need to take a two-pronged approach to locating and running data centres in a way that both supports initial AI training – which requires a central focus on PUE, cost and general efficiency – and putting that training into practice, which requires a stronger focus on reducing latency by ensuring proximity to metropolitan hubs.
Green pursuits
These methods are pathways to making data centres ready to accommodate an AI future, which will demand a new level of high-performance computing. At the same time, another critical element will, of course, be incorporating methods and technologies to improve sustainability as AI places greater demands on data-centre resources.
“If we do not start making better decisions that will prolong the longevity of data centre products, such as switching to liquid and immersive cooling techniques, our efforts to accommodate AI-dense infrastructure will essentially become futile,” says Levine.
AI itself offers ways of doing this, such as through its ability to capture millions of data points from thousands of sensors inside facilities that inform how to optimise operations. In its early stages, Data4’s AI technology, which is currently being deployed and trained in the company’s data centres, will make some recommendations for the optimisation of cooling systems. Meanwhile, the potential of other liquid cooling systems is also being investigated.
In addition, Data4 is looking into how to more widely integrate data centres into the local area in a more sustainable way, recognising that such facilities cannot be designed in isolation but need to be viewed as part of the wider industrial landscape.
One consideration that’s becoming a prerequisite for the development of new sites is the reuse of waste water. The likes of district heating, which is widely used in the Nordic countries, are potential solutions. Forging ahead for the future, Data4 plans to make this available in towns and cities where it establishes new sites or for the likes of agricultural applications.
Holistic view
To reduce the environmental impact of facilities themselves, Data4 is using low-carbon concrete, after finding that concrete contributes to 40% of the carbon footprint of data-centre buildings.
In general, the company’s strategy is to focus on the full end-to-end environmental impact of data centres through the use of lifecycle analysis. This comprehensive approach goes beyond the simple analysis of greenhouse gas emissions alone, examining the whole range of effects in relation to key indicators such as water use, ecotoxicity, use of mineral resources and the impact of construction.
Only by taking this type of holistic approach can all the necessary means for sustainability be taken into account. And that will be ever more important to get right in advance as the data centre industry seeks to incorporate the big next wave of AI growth.
