Singtel detailed plans to introduce a GPU-as-a-service (GPUaaS) for enterprises in its home market Singapore and across Southeast Asia in Q3, with intentions to offer different types of accelerators for customers’ advanced computing and AI needs.

The operator stated the service will initially be powered by Nvidia H100 Tensor core GPU clusters running at its local data centres and early in 2025 Singtel will be one of the first to deploy the chipmaker’s next-generation GB200 Grace Blackwell superchips, which deliver 30-times faster real-time large language model inference than its predecessor.

Singtel’s GPUaaS will be expanded to run in three high-density AI data centres operated by its regional data centre business Nxera in Singapore, Thailand and Indonesia when they begin operations.

Bill Chang, CEO of Singtel’s Digital InfraCo and Nxera units, said it is seeing “keen interest from the private and public sectors which are raring to deploy AI at scale quickly and cost-effectively”, adding the service will “run in AI-ready data centres…tailored for intense compute environments with purpose-built liquid-cooling technologies for maximum efficiency”.

Chang added enterprises of all sizes would be able “to harness the full potential of AI through readily available compute power to accelerate their development of generative AI, large language models and other AI workloads”. 

Singtel signed a deal with Nvidia in January to make AI more accessible in the city state and across the region by providing access to the chip company’s its GPU clusters in compatible data centres.