SK Group has declared itself as the 'national representative' in the AI infrastructure sector. SK Telecom (SKT) unveiled its AI infrastructure roadmap at the 'SK AI Summit 2025' held on November 3-4. The core of the plan is to secure over 50,000 GPUs in collaboration with NVIDIA and establish a nationwide AI data center (AI DC) network covering the metropolitan area, Gyeongnam, and the southwest region.
Jung Jae-heon, SKT CEO / Source=SKT
Jung Jae-heon, CEO of SKT, stated in his keynote speech, "AI infrastructure is the source of national competitiveness," and emphasized, "We will collaborate with the government to lead the evolution of AI infrastructure as a national representative AI operator and strive to become one of the top three AI powerhouses." He introduced SKT's key achievements related to AI DC and announced strategies such as ▲considering a large-scale expansion of the Ulsan AI DC ▲global expansion of energy-specialized AI DC solutions ▲pursuing 'edge AI' (processing data at the base station without sending it to the cloud to enhance ultra-low latency, security, and network efficiency) in collaboration with global big tech ▲establishing a manufacturing AI cloud ▲and advancing as a comprehensive AI DC operator.
SK AI DC Ulsan, Targeting Completion of Phase 1 by 2027
Groundbreaking ceremony of SK AI DC Ulsan / Source=SKT
AI infrastructure includes the hardware (GPUs, servers, etc.), software, and networking resources necessary for developing and executing AI. Recently, its importance has grown due to its close association with AI sovereignty. The government also announced plans through the Ministry of Science and ICT (MSIT) at the APEC summit last month to deploy around 50,000 of the latest NVIDIA GPUs in the national AI infrastructure by 2030, positioning AI as a core of national competitiveness.
SK Group's AI infrastructure strategy is clear: establishing a triangular hub system connecting the metropolitan area, Gyeongnam, and the southwest region. The key plan is to expand the Ulsan AI DC, currently being built in collaboration with AWS, to a total capacity of over 1GW, equivalent to the power output of one nuclear power plant. The Ulsan AI DC held its groundbreaking ceremony last August and aims to complete Phase 1 by 2027.
Notably, a strategic partnership with NVIDIA has been added to this project. This is part of the AI cooperation plan with Korea announced by NVIDIA CEO Jensen Huang during his visit to Korea on October 31. Jensen Huang announced cooperation with the Korean government, Samsung Electronics, SK Group, Hyundai Motor Group, Naver, and others, proposing support for a total of 260,000 GPUs in Korea. Of these, up to 50,000 units are allocated to SK Group, which plans to prioritize deploying this quantity in the Ulsan AI DC to establish large-scale AI infrastructure.
SK Hynix and NVIDIA are expected to collaborate across a wide range of fields / Source=IT Donga
Furthermore, SK Group and NVIDIA plan to collaborate across a wide range of fields, including the development of SK Hynix's high-bandwidth memory (HBM) essential for GPU operation, next-generation advanced memory solutions, semiconductor manufacturing, and communication infrastructure. NVIDIA stated, "We will accelerate semiconductor research, development, and production in collaboration with SK Group and establish cloud infrastructure to support the development of digital twins and AI agents," adding, "We will create an AI factory that activates the Korean AI ecosystem with NVIDIA's GPUs and software as AI infrastructure."
Upon completion, the Ulsan AI DC is expected to become the largest 'AI factory' in the country. An AI factory is a new manufacturing plant concept for the AI era, providing infrastructure to drive next-generation memory, robotics, digital twins, and intelligent AI agents. Particularly, SK Group's AI factory will be offered to developers participating in the Sovereign AI Foundation Model Project.
SKT plans to build a manufacturing AI cloud based on NVIDIA GPUs / Source=NVIDIA
The 'manufacturing AI cloud' announced by SKT is also noteworthy. SKT plans to establish a manufacturing AI cloud by adopting NVIDIA's simulation platform Omniverse and approximately 2,000 RTX PRO 6000 GPUs. This will be utilized for the AI transformation of major manufacturing companies within SK Group, including SK Hynix, and support the domestic manufacturing industry in advancing physical AI and robotics development. Particularly, SK Hynix aims to enhance productivity and energy efficiency by simulating semiconductor processes in the form of digital twins using this cloud.
Meanwhile, SK Group announced last October that it had signed a memorandum of understanding (MOU) with OpenAI and would establish a third hub in the southwest region. Although the specific location is undecided, it is expected to become an "AI belt connecting east and west along with the SK AI Data Center Ulsan." In his keynote speech, CEO Jung Jae-heon emphasized, "We will create the second and third models of the Ulsan AI DC to attract global capital investment in Korea and make Korea the largest AI hub in Asia."
Completion of AI Infrastructure Vertical IntegrationSK Group's rapid emergence as a 'national representative AI infrastructure company' is not solely due to the quantity of GPUs secured. It is because the group's overall structural strengths, vertically integrated business portfolio, and global cooperation network are organically intertwined. The value chain connecting semiconductors (SK Hynix), telecommunications (SKT), and energy (SK Innovation) is a key competitive advantage.
Jung Jae-heon, SKT CEO / Source=SKT
First, SKT is expanding technological cooperation in the network area by leveraging its expertise in telecommunications infrastructure. CEO Jung Jae-heon stated, "With the increase in AI services, the network infrastructure secured by telecommunications companies is being re-evaluated," emphasizing, "Edge AI and AI-based intelligent communication network technology, 'AI-RAN,' which bridges the gap between AI DC and on-device AI using nationwide telecommunications infrastructure, is an area only telecommunications companies can handle."
SKT has established a foundation for R&D cooperation with AWS to implement edge AI. By combining SKT's understanding of the domestic market and AI infrastructure technology with AWS's cloud and AI technology, they are strengthening a mid- to long-term cooperation system, including commercial edge AI testing. Additionally, SKT has signed agreements with NVIDIA, the government, academia, and others for joint research and demonstration of AI-RAN technology. AI-RAN, an intelligent network with AI technology applied from the design stage, is gaining attention as a key infrastructure for the spread of manufacturing AI.
Meanwhile, the massive power consumption of AI DCs is a chronic challenge directly linked to operating costs and environmental issues. SK Group is proactively addressing this issue by introducing innovative energy solutions. SK Innovation is expanding into an energy-saving data center model by combining LNG power generation and cold energy systems. This model is planned to be offered as an energy-specialized AI infrastructure export model in global markets beyond Korea, such as Vietnam, Malaysia, and Singapore, to establish a standard for efficient, low-carbon AI infrastructure overseas.
SK AI Summit site / Source=SKT
Ultimately, SK Group's strategy is to consolidate its core competencies under the single goal of AI to secure future growth engines. The goal is to expand existing roles and advance as a 'comprehensive AI DC operator' overseeing the entire project from design to construction and operation. SKT plans to internalize the core technology areas of AI infrastructure by collaborating with global big tech in each field and productize a cost-effective 'AI DC solution package.'
To achieve the ultimate goal of advancing to AI G3, it is crucial to make balanced strategic decisions encompassing hardware acquisition, network infrastructure acquisition, sustainable energy solution development and application, alliance management with global big tech, and ensuring openness and diversity in the domestic AI ecosystem. To this end, it will be essential to establish an organic cooperation system among the government, private sector, and academia, particularly securing the professional workforce necessary for AI infrastructure operation and development.
IT Donga Reporter Kim Ye-ji (yj@itdonga.com)
ⓒ dongA.com. All rights reserved. Reproduction, redistribution, or use for AI training prohibited.
Popular News