**AI Strategy Announced at ‘SK AI Summit 2024’**
(SEOUL = Yonhap News) Hyun-young Cho reports that Yoo Young-sang, CEO of SK Telecom, revealed plans to build an ‘AI Infrastructure Super Highway’ to help South Korea leap forward as an AI G3 country, focusing on three main pillars: Artificial Intelligence Data Centers (AI DC), Graphics Processing Unit cloud services (GPUaaS), and Edge AI.
During the ‘SK AI Summit 2024’ held at COEX in Seoul, Yoo stressed that the formula for success which started with infrastructure during the ICT era, making South Korea one of the world’s top ICT powerhouses, could also be applied in the AI era. Yoo also serves as the chairman of the SK SUPEX Council’s ICT Committee.
Regarding AI DC, Yoo mentioned plans to commence with a hyperscale level requiring more than 100MW (megawatts) of power at local hubs, with goals to expand this to a GW (gigawatt) level, positioning the country as an ‘AI DC Hub’ in the Asia-Pacific region.
SK Telecom explained that establishing AI DCs at regional hubs would leverage local renewable energy sources like hydrogen, solar, and wind power, and also facilitate global expansion through submarine cables.
Moreover, by combining SK’s highly efficient next-generation semiconductors and immersion cooling energy solutions with AI cluster management capabilities, the company aims to build cost-efficient AI DCs.
Before this, the company plans to open an AI DC testbed next month in Pangyo, implementing GPU virtualization solutions including NVIDIA’s latest chips, SK Hynix’s HBM, and three next-generation liquid cooling solutions.
This initiative expects to attract more than 50 trillion won in new investments, create over 550,000 jobs, generate more than 175 trillion won in economic impact, and foster the development of AI advanced industries in the region.
To address GPU supply shortages in the domestic market, SK Telecom will convert the Gasan Data Center in the Seoul metropolitan area into an AI DC and launch a GPUaaS offering cloud-based GPU services.
Collaborating with U.S. company Lambda, the GPUaaS service will start next month with the H100 GPU platform, with plans to introduce the latest H200 GPUs, a first in the country, by March next year incrementally increasing the supply.
Simultaneously, from next year, SK Telecom plans to invest a total of 100 billion won in creating a ‘Korean Sovereign AI’ by integrating Rebellion’s Neural Processing Unit (NPU), SK Hynix’s HBM, and various AI DC solutions from SK Telecom and its partners.
The company plans to build the largest NPU farm in the country, working with the government to strengthen AI technology independence and ecosystem expansion, not just in the public sector where data sovereignty is crucial, but also with major domestic companies, cloud providers, and K-AI startups.
For the introduction of Edge AI, nationwide communication infrastructure will be utilized.
Edge AI, an infrastructure combining mobile networks and AI computing, is expected to function complementarily with existing AI infrastructures, offering advantages in terms of low latency, enhanced security, and privacy over large-scale AI DCs and enabling large-scale AI calculations compared to on-device AI without server intervention.
Yoo commented, “So far, the competition in communication infrastructure was about speed and capacity, but now the paradigm of network evolution needs to change,” adding, “In the future, 6G will evolve into next-generation AI infrastructure that merges communication and AI.”
Additionally, SK Telecom plans to offer AI DC total solutions and GPUaaS in the global market by integrating AI semiconductors, data centers, and energy solutions through collaboration with domestic and international AI companies.