SINGTEL : Z74 0% is developing a new generation of artificial intelligence (AI) data centres (DCs) with high power density and sustainable features.

The DC at Tuas under construction will be the first of the fourth-generation DCs tailored for intense compute environments. They can handle higher power density AI workloads with a concentration of graphic processor units (GPUs), said Singtel on Tuesday (Dec 5).

GPUs, which were originally designed to process images and visual data, are now a critical application to enhance the process of dealing with big data, and are particularly useful in areas such as machine learning and creative productions.

However, the price of GPUs’ great computational power is high energy consumption and heat.

“DC Tuas is designed to mitigate carbon impact through the use of lower carbon embedded materials in construction, while driving greater energy efficiency by incorporating liquid cooling capabilities,” said Bill Chang, chief executive officer of Singtel’s digital InfraCo unit.

The fourth generation of DCs will adopt more efficient cooling solutions to prevent overheating of the GPUs, Singtel said.

A NEWSLETTER FOR YOU

Friday, 12.30 pm

ESG Insights

An exclusive weekly report on the latest environmental, social and governance issues.

Sign Up

Chang noted that the DC Tuas, offering 58 megawatts (MW) capacity, the largest among Singtel’s current data centres, will be one of the industry’s most efficient with a power usage effectiveness (PUE) ratio of around 1.23 at full load. A ratio of one indicates full power efficiency.

In a 2023 survey by the Uptime Institute of data centre owners and operators across more than 800 locations globally, the average annualised PUE ratio stood at 1.58 for their data centres.

“We are also working to transition the energy for our data centres with green and renewable sources to reduce the overall carbon footprint,” Chang added.

Besides DC Tuas, Singtel is also developing two other projects of the fourth generation DCs in Thailand and Indonesia. These will increase Singtel’s total pipeline capacity beyond 200 MW, from its current operational capacity of 62 MW in Singapore.

While aiming for more sustainable practices, the new generation of DCs will incorporate real-time server monitoring platforms to ensure smooth management of the GPU clusters.

The DCs will also serve enterprises’ AI needs via connecting the AI clusters to smart devices, such as autonomous systems, “allowing AI to be brought to the edge at high bandwidth and low latencies to enable various industry use cases”, said Singtel.

Chang added: “Within these AI DCs, we will offer a host of core services including cloud co-location services, managed hosting services and value-added GPU-as-a-Service (GPUaaS).”

Enterprises can leverage GPUaaS to accelerate their development of generative AI, large language models and other AI workloads without having to incur high capital expenditures in acquiring the AI DC space and GPUs, said the data centre operator.

According to Fortune Business Insights, the global GPUaaS market size is expected to grow to US$25.53 billion by 2030 from US$3.16 billion in 2023, while the Asia-Pacific market could see the highest growth rate during this period, given the region’s significant interest in AI, Singtel noted.

Singtel is scaling up its regional data centre business. On Monday, it announced its first five-year green loan of S$535 million for data centres.

In September, the largest telco in Singapore also revealed KKR investment of up to S$1.1 billion for an enhanced stake in its regional data centre business.

Shares of Singtel were trading flat at S$2.28 as at midday market break on Tuesday.