News

announced it will be launching its Graphics Processing Unit (GPU)-as-a-Service (GPUaaS) in Singapore and Southeast Asia in the third quarter of this year. This service will provide enterprises with access to AI computing power, to drive greater efficiencies and accelerate growth and innovation. At launch, Singtel’s GPUaaS will be powered by NVIDIA H100 Tensor Core GPU-powered clusters, which are already being utilised in some of the upgraded data centres in Singapore. In addition to , Singtel will be among the world’s first to deploy NVIDIA’s next-generation GB200 Grace Blackwell Superchips. These chips deliver 30X faster real-time large language model inference than its predecessor, and Singtel will be one of the first NVIDIA Partner Network Cloud Partners to receive them early next year. This will give Singtel’s enterprise customers the option of different types of accelerators for their advanced computing and AI needs. The GPUaaS will eventually be expanded to run in three new sustainable, hyper-connected, high-density AI data centres by Singtel’s regional data centre business in Singapore, Thailand and Indonesia when operations begin. GPUaaS is like any other service offerings, in this case, GPUs are being offered as a service. Some other common examples of services are , . GPUaaS offers a convenient way to access high-performance computing resources for machine learning, deep learning, and other data-intensive applications. By utilizing the power of GPUs, GPUaaS allows users to leverage advanced computational capabilities without the need for expensive hardware or complex infrastructure management. An example of IaaS would be cloud storage services, which most, if not all, of us utilise in some form to store our data online. Singtel says that they are seeing keen interest from both private and public sectors which are raring to deploy AI at scale quickly and cost-effectively. Singtel’s GPUaaS will run in AI-ready data centres specifically tailored for intense compute environments, with purpose-built liquid-cooling technologies for maximum efficiency and lowest PUE. This allows customers the flexibility to deploy AI without having to invest and manage expensive data centre infrastructure. has its own data center platform and claims to be the world’s leading accelerated computing solution, deployed by the largest supercomputing centers and enterprises. For many companies, it would just not be cost effective to build their own infrastructure, and this is where Singtel and Nvidia are looking at an additional source of revenue. Just taking a very layman example of a physical server for the purpose of a network drive, a company would require an IT professional to manage and install the relevant software, an environment that maintains the right temperature around the clock, schedule backups and ensure that a disaster recovery plan can be executed smoothly when required. The company would also have to upgrade its infrastructure as and when it starts to become obsolete. With AI’s growing significance, companies are looking to harness the full potential of AI through readily available computing power to accelerate their development of generative AI, large language models and other AI workloads. Singtel would also likely benefit from any potential government projects with its next-generation digital infrastructure, which comprises of AI , fixed networks, 5G network and GPUaaS. This initiative supports Singapore’s AI ambitions and standing as a digital and AI hub, while also supporting digital transformation in the region. We think it is not financially significant at the onset, but significant in terms of direction. Singtel is a massive telecommunications company, and while the significance justified a public announcement, it is definitely still in a nascent stage. The benefit for Singtel is also that it can increase its range of offerings to its existing customers as well as potentially attract new customers and enable future cross-selling. This also aligns with Singtel’s strategic objective to seize opportunities created by macrotrends such as technology proliferation and large-scale digitalisation in this region. Singtel has made it clear that it is looking to unlock the value of its portfolio of quality infrastructure assets to invest in infrastructure crucial for the digital economy. It was recently disclosed that Singtel unlocked S$0.95 billion with the sale of a 0.8% direct stake in regional associate Airtel to US-based investment firm GQG Partners. The resultant gain from the sale is estimated to be S$0.7 billion. Singtel still has an effective stake of 29.0% in Airtel, worth an estimated S$33 billion. There was an article in the Australian press stating that Singtel was in advanced talks to offload Optus for A$16 billion (S$14 billion). In comparison to Singtel’s market cap of S$42 billion, an asset sale of this size would no doubt be very significant. Singtel previously acquired Optus for $11 billion in 2001 and Optus recorded an EBIT in 1H23 of $141 million. Both of these assets combined already exceed the value of Singtel’s market capitalisation. Singtel still has other regional associates such as Telkomsel in Indonesia, AIS and Intouch in Thailand and Globe in the Philippines, which in total contributed more than $800 million in operating income just in 1H23. These four regional associates are also listed, and their value itself is around $40 billion in total. Singtel also owns nearly 25% in NetLink Trust (SGX: CJLU) and 22% in Singpost (SGX: S08), worth about $1.1 billion in total. In conclusion, Singtel will need to continue to sell its mature assets and invest in upcoming, trendy and potentially higher yielding and higher growth assets with its vast amount of capital . READ MORE READ MORE