News Categories

Singtel designs its upcoming data centres to better serve AI workloads

By Ken Wong - on 5 Dec 2023, 1:05pm

Singtel designs its upcoming data centres to better serve AI workloads

image source: Singtel.

With the rise of AI workloads, Singtel is ensuring that its data centres are ready to handle higher power density AI workloads that will come from a concentration of Graphic Processor Units (GPUs) that consume a lot of power and generate a lot of heat.

Bill Chang, CEO of Singtel's Digital InfraCo unit said:

With the fast-growing adoption of AI across various industries, we see an inflection point in the data centre industry for high power density, yet sustainable and hyper-connected AI DCs that are optimised to manage GPU clusters for AI workloads. This is why we are building our fourth-generation data centre platform specifically tailored for such intense compute environments.

Singtel is building three fourth-generation data centres, in Tuas, Thailand, and Indonesia which will increase its total pipeline capacity beyond 200MW from its current operational capacity of 62MW in Singapore.

Of the three, DC Tuas will be the largest at 58MW capacity and will be Singtel’s first fourth generation, high power density, and sustainable AI data centre. It is designed to mitigate its carbon impact by using lower carbon embedded materials in its construction and driving greater energy efficiency by incorporating more efficient cooling solutions like direct-to-chip liquid or immersive cooling to avert overheating. DC Tuas is also an integrated submarine cable landing station making it a key platform for global enterprises and cloud companies to host their AI workloads in Singapore as a key hub in Asia.

DC Tuas will be one of the industry’s most efficient with its Power Usage Effectiveness (PUE) of around 1.23 at full load. Power Usage Effectiveness is a ratio that describes how efficiently a computer data centre uses energy, i.e. how much energy is used by the computing equipment. An ideal PUE is 1.0. Anything that isn't considered a computing device in a data centre (e.g. lighting, cooling, etc.) falls into the category of facility energy consumption.

These AI data centres will be connected to Singtel's various fixed and 5G mobile networks that will be orchestrated by Singtel’s Paragon platform to serve enterprises' AI needs.

This will enable smart devices like robotics and autonomous systems to be connected to the AI clusters within these data centres, allowing AI to be brought to the edge at high bandwidth and low latencies to enable various industry use cases.

According to Chang, Singtel plans to offer a host of core services including cloud co-location services, managed hosting services and value-added GPU-as-a-Service (GPUaaS) out of the new data centres.

We reached out to Singtel to find out what type of servers the company planned to deploy but all a Singtel spokesperson would say is that they will be using a combination of various AI server vendors. We will update this article if we receive more information.

Join HWZ's Telegram channel here and catch all the latest tech news!
Our articles may contain affiliate links. If you buy through these links, we may earn a small commission.