AI is set to revolutionise the data centre sector, driving demand, energy-efficiencies and making developers think differently about where to locate their facilities.

Generative AI has become the darling of the tech sector, best known through applications such as Chat GPT. The term “generative” refers to its ability to create new data. Platforms learn patterns and structure from training data, and can create text, images and other media.

It has been one of the most heavily invested tech sectors in recent years. A study from Juniper Research found spend on Generative AI by banks alone will reach $85 billion in 2030, from $6 billion globally in 2024. It predicts leading banks will adopt generative AI services to offer more personalised user experiences at reduced cost.

Jack Harkness, Director, Regional Industrial & Logistics Services for Asia Pacific at Savills, says: “Generative AI requires a huge amount of computing power and this means it is going to become an increasingly important demand driver for data centres all around the world.

“Asian companies such as Alibaba, Tata and Baidu, as well as international companies such as AWS, Google and Microsoft, are developing AI solutions which will require data centre space in this region.”

A January report from the International Energy Agency predicted that data centres will use nearly twice as much electricity in 2026 as in 2022, due to growth of AI and cryptocurrencies. This forecast is much higher than previous estimates and implies huge growth in the number data centres over the next few years.

Schroders Capital, in a report published last year, said: “Following an extensive deep-dive into key Asian cities, we believe that the region presents some of the most exciting growth opportunities.”

As well as adding to demand in key cities, AI may also open up other locations for data centres. Generative AI has two phases, training (where the platform learns) and inference (where it puts this into practice). Inference requires low latency – the time it takes for data to be transferred – so users get quick responses. This is especially important in robotics and automation and means the data centres hosting the AI need to be close to where it is used.

This is not the case for AI training, which opens up an opportunity to locate data centres in more remote locations. This means they could be located where land is cheaper or renewable energy more available.

The energy used by data centres is one the prime concerns expressed about them. However, while AI is adding to energy demand, it can also be used to make data centres more efficient, either by running processes more efficiently or optimising cooling. For example, Google used AI to cut the cooling costs in its data centres by 40%.

Reducing the power demands of data centres may also widen the range of locations where they can be operated and better energy-efficiency will appeal to investors who are on a pathway to net zero carbon in their portfolios.

“We are in the early stages of the AI transformation, but it is already exciting to see what it can do for the prospects of the data centre sector in this region,” says Harkness.

Further reading:
Asian Cities Report H2/2023

Contact Us:
Jack Harkness