Jeong Seok-geun, 50, CEO of SK Hyper, who is leading SK Group's artificial intelligence data center (AIDC) business development, named "securing capacity" without hesitation when asked about the top priority that will determine the company's success or failure. He said that meeting SK Hyper's target capacity would be proof that it had solved multiple challenges at once. Jeong said, "Securing capacity means we have brought in global customers beyond Korea, solved the technical challenges for building, and reached common ground with global investors to obtain funding."
Jeong graduated from the computer science department at KAIST and earned an MBA from Georgetown University in the United States. As head of Naver CLOVA CIC (company-in-company), he led the development of HyperCLOVA, and in 2023 he joined SK Telecom to oversee the company's AI business. In Jul. this year, SK Hyper was launched and he became its inaugural CEO. SK Hyper is a corporation dedicated to SK Group's AIDC business and is a 100% subsidiary of SK Telecom. SK Hyper aims to build AIDC at the gigawatt (GW) level, including 5 GW by 2029 and 15 GW by 2035.
Jeong cited U.S. big tech companies as the customers that will fill the capacity of the AIDC built by SK Hyper. Jeong said, "With infrastructure challenges in the United States and a surge in AI demand in Asia coinciding, the need for big tech to transfer globally is growing." Jeong, who delivered a keynote speech at "SMARTCLOUD SHOW 2026," the country's largest tech conference hosted by ChosunBiz on the 26th of last month, was interviewed. The following is a Q&A.
─What was the reason for launching SK Hyper as a separate corporation.
"The paradigm of the data center business has changed. As we enter the AIDC era, the scale has grown from the megawatt (MW) level to the GW level, so it was necessary to play a role in concentrating the capabilities of the entire group. Because large-scale funds are involved, I thought we needed an organization specializing in financing and a specialized organization to sell to and negotiate with big tech."
─What has been the most challenging task since taking office.
"Overall, it's not easy. Not only is the scale unprecedented in Korea, but competition is also fierce globally. Recently, Nvidia announced the establishment of a $500 billion (about 671 trillion won) AI infrastructure funding network, and the money changing hands is not in trillions of won but in the hundreds of trillions. The market is changing very fast. It's not just the technology that is changing—because such large amounts of capital are being invested, the methods of financing are also changing."
─What is the reaction from the big tech companies you are in talks with.
"Big tech prefers the United States, which is nearby, as the most favored AIDC location. However, there are various infrastructure-related challenges within the United States, so they feel the need to go global. Also, as AI demand in Asia is rapidly increasing, there is demand to establish footholds within Asia. But to attract GW-level capacity to Korea, we must be a top-tier location in Asia. That is our biggest concern these days."
─What are Korea's strengths and weaknesses in the location competition.
"In the past, data centers were built close to where demand existed, so they were spread evenly across countries. In contrast, now that capacity has grown, factors such as power stability, government support, and geopolitical stability have become far more important. Korea's power infrastructure is a clear strength, and at a time when AI models are becoming strategic assets, trust in geopolitical stability is also high. On the other hand, land is scarce, so buildings must be taller, and there is a burden from electricity rates. If the AIDC-specific tariff system under government review becomes tangible, our competitiveness will be further strengthened."
─What differentiates SK Hyper.
"When big tech chooses a partner, it puts weight on 'whether there is conviction about the future of AI and a willingness to invest' and 'how quickly one can keep pace with a rapidly changing world.' Led by Chairman Chey Tae-won, SK Group is making large-scale investments in AI at the group level, making that commitment clear. We also have a track record of working with Amazon Web Services (AWS) on sizable projects in Korea."
─SK Telecom is investing in SK hynix's U.S. AI investment corporation and joining its board. Is this also a move to pool group capabilities.
"It is a decision to secure the innovation happening in the United States. Data centers are not simply a construction business; they are a business where technological innovation is active. As AI Semiconductor evolves, the way data centers are built and operated will also change. Therefore, exchanges on data center technology and with related corporations are important for both SK hynix and SK Telecom. It is also important for SK hynix to ensure that its memory is well used in data centers."
─Will contracts with big tech be signed directly by SK Hyper, or by a special purpose company (SPC) to be established as the AIDC operator.
"We are still deliberating. Because the scale of financing is so large, this is not a matter we can decide on our own. We need to work with investors and big tech, our customers, to flesh out the structure."
─What is the ratio of internal to external funding for the 15 GW build-out.
"It has not been decided yet."
─In the United States, there are warnings that hyperscalers' investment scale is at historic levels and may be excessive.
"I understand the concerns because the scale is so large. But considering that a single coding agent has completely changed software development methodologies, we are only at the beginning of AI-driven change. It's not easy to implement the same AI performance with less infrastructure; in the end, it inevitably becomes a battle of scale. In fact, when you meet big tech, while there is anxiety about pressure from shareholders and the market, there is a firm consensus that 'we cannot stop investing.'"
─SK Hyper's target scale is also significant.
"It is an unprecedented scale. Over the past 20-plus years, the data centers built in Korea amount to 1 GW, while SK Hyper is saying it will build 5 GW in just the next three years, by 2029. It is an unprecedented challenge across technology, sales, and finance. The government is supporting this, but if more attention and help are given, we can build computing infrastructure of meaningful scale in the country. Once big tech establishes a foothold, it expands outward from there, and from the outset it moves with a 20- to 30-year horizon. Now is the perfect time to grow the related industries."
─What indicators would you like SK Hyper's future success to be measured by.
"Capacity is the most important. That is because securing capacity means we have brought in global customers beyond Korea, solved the technical challenges for building, and reached common ground with global investors to obtain funding. Contributing to Korea's "AI top 3" goal is also an important success metric. If AIDC and computing infrastructure are built at scale domestically and many big tech firms come in, I believe it will ultimately have a major positive effect on Korea's overall industries. It will also help countless startups and the public use AI."
─At Naver you built AI models, and at SK Group you are leading the development of the infrastructure that runs those models. For the development of Korea's AI industry, which is more important, models or infrastructure.
"Models are still the most critical element. After all, infrastructure exists to run models well. But if you look at where the bottleneck is now, it is infrastructure. No matter how outstanding a model's performance is, without infrastructure it cannot meet customers.
If only models mattered in the AI business, we would have to compete with the world's best models just to be in the game. But even those who build the best models are being hamstrung by infrastructure, and we can solve the infrastructure problem. As the importance of infrastructure rises, an opportunity has opened for us to influence the game."
―SK Telecom is continuing model development alongside infrastructure. What is the long-term direction.
"The approach is to build a frontier-scale model first and then make it lighter. After delivering top performance, we plan to move on to slimming and optimization. Even looking only at Chinese open-source models, the advantage of scale is clear. It's not that slimming is less important; rather, it means that for now, it is right to first tackle the bigger and harder task of scaling model size."
─What differentiation will SK Telecom present in the "Independent Foundation Model (Dokpamo)" competition, which will narrow from three teams to two early next year.
"AI cannot be evaluated by benchmark scores alone. Doing well on a college entrance exam does not necessarily mean doing the job well. We can tout being at the top of the leaderboard publicly, but customers do not buy based on leaderboard scores. We aim to put emphasis on real-world usable performance and ease of use."