• Singtel is one of a number of Asian telcos developing GPU-as-a-service (GPUaaS) offerings
• It has been forging partnerships and investing in its datacentre facilities ahead of its commercial GPUaaS launch
• Its latest partnership is with GMI Cloud, which has datacentres in Asia Pacific and plans to expand in the US

Singtel has further advanced its GPU-as-a-service (GPUaaS) strategy ahead of its planned commercial launch later this year by agreeing a partnership with GMI Cloud, a GPUaaS specialist that is already offering services from datacentres in Asia Pacific. 

In March, Singtel announced its plan to launch commercial GPUaaS in Singapore and South-east Asia during the third quarter of this year, and since then it has been investing in its own datacentre capabilities and forging strategic partnerships in order to offer an attractive and meaningful service to enterprise users from day one. 

Just last week it announced a collaboration with AI cloud platform company Nscale whereby Singtel will leverage Nscale’s AMD and Nvidia GPU capacity in Europe, while Nscale will be able to tap into Singtel’s Nvidia H100 Tensor Core GPU capacity in the South-east Asian region for its customers’ workloads through an integration with Singtel’s orchestration platform, Paragon – see this What’s Up With… article for further details. 

Singtel also recently announced a strategic partnership with network operator industry group Bridge Alliance that will enable it to offer its GPUaaS offerings to enterprises across South-east Asia via the alliance’s members – see this press release for further details. 

And earlier this year it announced a similar partnership with US-based Vultr, which operates 32 cloud datacentre locations globally, including nine across the Asia Pacific region (as well as locations throughout the Americas, Europe, Israel and South Africa), and which already offers Nvidia-based infrastructure for AI training and inference workloads in Singapore, India, Australia, Japan and South Korea.

Now it has struck a “strategic collaboration” with GMI Cloud to “expand GPU capacity by combining both companies’ GPU resources and infrastructure,” a move that “extends Singtel and GMI Cloud’s reach into previously untapped markets, enabling customers of both companies to gain access to GPUs across a wider footprint for their workloads.” 

GMI Cloud currently has five datacentre locations, including two in Taiwan and one in Thailand (Supernap), with plans to build and expand in the US. Its GPU platform will also be integrated with Singtel’s Paragon orchestration platform to leverage Singtel’s Nvidia H100 Tensor Core GPU capacity in Singapore. In addition, Singtel’s datacentre division, Nxera, is expanding its portfolio of facilities across Singapore, Indonesia, Thailand and Malaysia. “These state-of-the-art facilities featuring advanced Nvidia GPUs are designed to manage the most demanding AI workloads efficiently and sustainably,” noted Singtel. 

Bill Chang, CEO of Singtel’s Digital InfraCo and Nxera, stated that the GMI Cloud partnership “expands [our] capacity and availability zones in Asia Pacific, complementing our own GPUaaS points of presence. This will give enterprises the flexibility and scalability they need as they harness supercomputing power to accelerate innovation.”

GMI Cloud CEO Alex Yeh added: “Our collaboration with Singtel marks a significant milestone in democratising AI infrastructure across Asia Pacific. GMI Cloud’s expertise in providing scalable, telecom-optimised GPU solutions complements Singtel’s robust network perfectly. Our flexible deployment models and experience with high-volume, low-latency workloads position us uniquely to support the telecommunications industry’s AI initiatives. Together, we’re creating an ecosystem that will accelerate AI adoption, enabling telecom providers and enterprises to enhance their services and drive innovation in the region.” 

Singtel isn’t the only Asian telco with GPUaaS aspirations: South Korea’s SK Telecom has also been busy forging partnerships that will enable it to offer such services to the enterprise sector – see SK Telecom to deploy Nvidia GPU cluster with Lambda.

- Ray Le Maistre, Editorial Director, TelecomTV