• • • DATA CENTRE MANAGEMENT • • • AI is here, but are our
data centres ready for it? Recent events have highlighted just how fast-paced and unpredictable the world of AI development can be
ne thing remains clear, however; we need bigger, better and more efficient data centres to manage this change. Amrik Sangha, a partner at Gateley Legal in Dubai, examines whether the current stock is up to the task.
O
When it comes to power, modern data centres are getting hungrier. The increasing use of power-intensive workloads such as artificial intelligence (AI) and cryptocurrency mining have driven up the power that data centres use, surpassing the efficiency gains that had, for many years, helped to keep data centre power use relatively stable.
AI has a big part to play in this. According to a 2024 report by Goldman Sachs, AI could potentially increase data centre power demand by 160 per cent by 2030, resulting in data centres consuming up to four per cent of the world’s power by the end of the decade. AI will represent 19 per cent of data centre power demand, according to the same report.
Energy is everything Small wonder, then, that many of the world’s biggest hyperscalers are heavily investing in alternative forms of energy, such as renewables, hydrogen and Small Modular Reactors, to keep data centres powered without sacrificing any commitments relating to carbon emissions reductions.
In many cases, these investments will also coincide with the construction of new, purpose-built data centres that are ready to handle these power demands efficiently and are more easily linked to alternative energy sources.
In the Middle East alone, for example, the data centre construction market is predicted to be worth US$4.39 billion by 2029, while the colocation market is expected to read US$2.14 billion by 2028. That is not to say, however, that new players in the AI market are not finding ways in which to make AI more efficient.
The Chinese AI chatbot DeepSeek, for example, has caused significant turmoil in the US market because its newly launched open- source model purports to be far cheaper and less power-hungry than those developed by the US. In addition to causing one of the biggest market value drops in US stock market history, the news has also seen many investors alter their positions in US energy companies.
Retrofit or rebuild? How these developments in AI software will affect the construction of new-build data centres is, as yet, unclear. What is clear, however, is that the current stock of data centres is in significant need of a reboot if they are to avoid obsolescence as AI, cryptocurrency mining and cloud computing become more widespread and advanced requiring more power in a confined rack. Retrofitting incumbent data centres is no easy task, however, and will often require a fundamental re-evaluation of the data centre’s design and infrastructure.
Data centre operators, for example, may face increasing pressure from customers to commit to using more carbon-neutral construction methods plus energy sources. This may prove difficult if the data centre’s existing licences, locations and designs do not allow for this.
Keeping it cool
One way in which incumbent data centres could explore efficiency gains is through cooling technologies.
Where traditional air-cooling systems are employed, these can account for 30-55 per cent of a data centre’s power consumption. They are also highly inefficient at cooling power-intensive workloads, all of which at first glance makes them a relatively low-hanging fruit for incumbent data centres to change. Retrofitting new cooling technologies, such as liquid-based immersion cooling, is not a simple matter of swapping out and plugging in, however. This comes with its own considerations, such as whether the structural floor loading of a data centre is sufficient to manage the greater weight of the cooling baths, or whether plumbing systems (traditionally designed to keep water out at all costs) can be reconfigured to allow water to flow into the interior as well.
Data centre operators also need to weigh up any retrofitting projects with their day-to-day obligations, particularly if they service third parties. What is the likelihood of any downtime during renovation, and how might this impact any existing service level agreements or contracts, for example.
While advancements in AI software will help to solve part of the problem, it is still important that incumbent data centres review their current infrastructure and decide where any energy efficiency gains can be made. A country’s very economic competitivity could well depend on it.
52 ELECTRICAL ENGINEERING • FEBRUARY 2025
electricalengineeringmagazine.co.uk
Page 1 |
Page 2 |
Page 3 |
Page 4 |
Page 5 |
Page 6 |
Page 7 |
Page 8 |
Page 9 |
Page 10 |
Page 11 |
Page 12 |
Page 13 |
Page 14 |
Page 15 |
Page 16 |
Page 17 |
Page 18 |
Page 19 |
Page 20 |
Page 21 |
Page 22 |
Page 23 |
Page 24 |
Page 25 |
Page 26 |
Page 27 |
Page 28 |
Page 29 |
Page 30 |
Page 31 |
Page 32 |
Page 33 |
Page 34 |
Page 35 |
Page 36 |
Page 37 |
Page 38 |
Page 39 |
Page 40 |
Page 41 |
Page 42 |
Page 43 |
Page 44 |
Page 45 |
Page 46 |
Page 47 |
Page 48 |
Page 49 |
Page 50 |
Page 51 |
Page 52 |
Page 53 |
Page 54 |
Page 55 |
Page 56 |
Page 57 |
Page 58 |
Page 59 |
Page 60 |
Page 61 |
Page 62 |
Page 63 |
Page 64 |
Page 65 |
Page 66
