SK Telecom (017670.KS), which has emerged as a key player in AI factories, has begun developing AI technology specialized for operating AI data centers (AIDC) in earnest. Linked to the AI factory construction project it is pursuing with Nvidia, the company is developing dedicated AI agents and foundation models that predict and manage power, cooling, and failure signs in real time. The combination of SK hynix's high-bandwidth memory (HBM) competitiveness with SK Telecom's AI infrastructure operation technology is drawing assessments that the building of an AI infrastructure value chain at the SK Group level is taking more concrete shape.
According to the information technology (IT) industry on Sunday, SK Telecom is developing AI agents and a Time Series Foundation Model (TSFM) specialized for AIDC operations. TSFM is an AI technology that learns from data accumulated over time, such as power usage, temperature, cooling equipment status, network traffic, and failure logs, to predict future conditions. SK Telecom plans to combine this with a large language model (LLM) to implement an automation system optimized for AIDC management.
Once this technology is applied, it can analyze in real time and respond preemptively to an AI data center's power consumption, cooling status, graphics processing unit (GPU) utilization rate, network bottlenecks, and failure signs. For example, it can predict in advance the possibility of a power load surge or server overheating during a specific time period to adjust cooling device output, and distribute training and inference tasks to other zones. While conventional data center control involved humans monitoring the current status and responding after the fact, the operational AI being pursued by SK Telecom is closer to a system that predicts future conditions and takes automatic measures.
SK Telecom's intent to use AI in AIDC is because the profitability of the AI factory business depends on operational efficiency. AI data centers, where GPUs are densely concentrated on a large scale, carry far greater power density and heat burdens than conventional data centers. While securing as many expensive GPUs as possible is important, how reliably they can be run at a high utilization rate and how much power costs can be reduced determine the competitiveness of the AI cloud business. The same reason is why global companies are pursuing operational automation, such as Google applying DeepMind AI to its data center cooling control systems.
Operational automation is also essential for overseas expansion in particular. Since power environments, cooling conditions, network quality, and workforce levels differ by country and region, there are limits to relying solely on humans directly transferring operational know-how. SK Telecom's effort to build an AI-based operating system is interpreted as a strategic move to secure the same level of operational quality and efficiency as in Korea, no matter where it builds an AI factory.
The recent cooperation between SK Group and Nvidia is further raising the importance of such operational AI. During the visit to Korea by Nvidia CEO Jensen Huang early this month, SK Group announced an AI factory partnership with Nvidia. The AI factory is a next-generation AI data center optimized for AI training and inference, based on Nvidia's AI infrastructure platform "DSX." SK Telecom plans to begin operating its first AI factory in Korea next year, and afterward push to expand overseas bases centered on the Asian region. SK Group Chairman Tae-won Choi also revealed in an interview with Japan's Nihon Keizai Shimbun a plan to build an AI factory in Japan, targeting 2028-2029.
The industry expects SK Telecom to accelerate the commercialization of its AIDC foundation model on the basis of its development cooperation with Nvidia. The two companies previously disclosed cooperation results applying a digital twin to SK hynix's manufacturing process. SK Telecom developed technology that uses Nvidia's agent technology to automatically process field data, such as manufacturing facilities and spatial structures, to fit a digital twin environment. This type of cooperation can extend to AI factory operations as well. "SK Telecom's role in the partnership with Nvidia will not be limited to an infrastructure investor or a system integration (SI) provider," an industry official said. "There is a possibility it will also take charge of developing the solutions needed to operate AI factories."
SK Telecom's AX application is spreading beyond data center operations to enterprise-wide work innovation. CEO Jung Jae-hun presented "AX Innovation 2.0" at the "2026 New Icheon Forum" held recently at the SKMS Research Institute in Icheon, Gyeonggi Province. The vision is to fundamentally redesign how employees work to achieve productivity gains and business model innovation. Accordingly, AI agents are assigned employee numbers and given affiliations, job functions, and authorities, managed in a manner similar to humans. "Through the routinization of AX, employees' time and capabilities will be converted into a growth engine that drives new challenges," CEO Jung said.
