• South Korea’s SK Telecom is very focused on being an AI leader in its home market and beyond
• It is preparing to launch its AI datacentre testbed in South Korea this December and launch a GPU-as-a-service (GPUaaS) offering in the same month 
• It harbours aspirations to offer AI datacentre services in multiple markets

The AI drumbeat at South Korea’s SK Telecom (SKT) is getting louder by the month: The network operator’s CEO, Ryu Young-sang, put more flesh on the company’s AI strategy bones during a keynote speech at its own AI Summit in Seoul, where he presented SKT’s vision for an “AI infrastructure superhighway”. 

The plan is based on three pillars – build out a national network of regional AI datacentres; enhance that with edge AI infrastructure; and launch a GPU-as-a-service (GPUaaS) offering to enterprise and government users before expanding the reach of the resulting portfolio of AI infrastructure services into other geographic markets via high-capacity submarine network links in collaboration with partners. 

SKT will start by opening an AI datacentre testbed in Pangyo, which is just south of Seoul, in December. That facility will be underpinned by technology, which has been developed within the SK group of companies, including GPU (graphics processing unit) virtualisation solutions, AI energy optimisation tools and SK Hynix’s HBM (high-bandwidth memory) chips – as well as products from key partners, such as Nvidia. The testbed will also use three types of liquid-cooling solutions, namely direct liquid cooling, immersion cooling, and precision liquid cooling, explained the CEO. 

The insights and experience from that facility will then be used in the construction of 100 MW (megawatts) hyperscale AI datacentres in multiple regions across the country, with plans to expand the IT capacity of those sites to GW (gigawatts) in an effort to develop an AI datacentre “hub in the Asia Pacific region”. SKT believes these AI datacentres will be able to source a “stable power supply through the utilisation of new renewable energy sources, such as hydrogen, solar and wind power, and easily expand to global markets through submarine cables”. 

One of the earliest AI datacentre sites will be in Gasan, on the outskirts of Seoul, where an existing datacentre facility will be upgraded through the deployment of a cluster of Nvidia H100 Tensor Core GPUs that, in partnership with San Francisco-based AI infrastructure developer Lambda, will be used to launch a GPUaaS offering. SKT and Lambda announced a partnership in August – see SK Telecom to deploy Nvidia GPU cluster with Lambda.

Then in March next year, SKT plans to deploy clusters of Nvidia H200 Tensor Core GPUs to further meet user demand. SKT also plans to use its experience in planning, building and running these AI datacentre sites to develop an AI infrastructure solution “that combines AI semiconductors, datacentres, and energy solutions through collaboration with AI companies in Korea and abroad, with the aim of entering the global market.” 

But it’s not all about hyperscale datacentres: SKT also plans to integrate AI infrastructure with its mobile network infrastructure to enable “edge AI” that can offer “reduced latency, enhanced security, and improved privacy” compared with large hyperscale datacentres. “Additionally, it enables large-scale AI computing, complementing the existing AI infrastructure, compared to on-device AI,” noted SKT in this announcement. 

The company says it is “carrying out various proof-of-concept (PoC) projects across six areas, including healthcare, AI robots, and AI CCTV, to discover specialised edge AI services,” and collaborating with global partners in edge AI infrastructure. As TelecomTV reported recently, SKT is one of five telcos to have submitted an application to join the AI-RAN Alliance, which is focused on the integration of AI and radio access network infrastructure – to find out more, watch our exclusive video interview with the AI-RAN Alliance chair, Alex Jinsung Choi – see Under the hood of the AI-RAN Alliance.

“So far, the competition in telecommunications infrastructure has been all about connectivity, namely speed and capacity, but now the paradigm of network evolution should be changed,” stated Ryu Young-sang in his keynote speech. “The upcoming 6G will evolve into a next-generation AI infrastructure where communication and AI are integrated,” added SKT’s CEO. 

- Ray Le Maistre, Editorial Director, TelecomTV