Singtel to host Vultr’s GPU workloads in AI datacentres

Singtel is hosting the graphics processing unit (GPU) workloads of specialist cloud providers in its new generation Nxera artificial intelligence (AI) datacentres in a bid to provide core AI infrastructure for enterprises.

It recently inked a partnership with Vultr, a privately held cloud computing provider, to expand the latter’s capacity of Nvidia GPUs in Singapore and across Southeast Asia.

As part of the deal, Vultr’s Nvidia-powered cloud GPU platform, will integrate with Singtel’s Paragon platform and leverage Singtel’s Nvidia H100 Tensor Core GPU capacity in Singapore for Vultr’s customer workloads.

Vultr currently operates 32 datacentre locations globally, including nine across the Asia-Pacific region where it is providing Nvidia infrastructure for AI training and inference workloads across Singapore, India, Australia, Japan, and South Korea.

In the third quarter of 2024, Vultr users, primarily cloud engineers looking for a lab environment, will be able to deploy large-scale clusters of Nvidia H100 GPUs in Vultr’s Singapore region hosted by Singtel.

“Both Singtel and Vultr aim to make high-performance computing accessible and affordable for enterprises,” said Bill Chang, CEO of Singtel's Digital InfraCo unit and Nxera regional datacentre business.

“This partnership will help foster productivity and innovation at scale without customers having to worry about procuring and managing their own infrastructure of GPUs, datacentres, and networks,” he added.

Earlier this year, Singtel launched the Nxera datacentre brand for a new generation of datacentres built to cater to AI workloads.

The first such datacentre, the 58MW DC Tuas, will be the first of its kind to incorporate more efficient cooling services, such as direct-to-chip liquid and immersive cooling, enabling it to achieve a power usage effectiveness (PUE) of around 1.23 at full load.

Through a partnership with Nvidia, Singtel will also become an Nvidia cloud partner in the region and offer GPU clusters to support its customers’ cloud and AI needs through a GPU-as-a-service (GPUaaS) offering.