SK Group showcased its capabilities for assembly the fast-growing market demand for synthetic intelligence (AI) throughout its 2025 SK AI Summit Monday, the place Chairman Chey Tae-won emphasised “effectivity over scale” within the group’s strategy to offering reminiscence chips, infrastructure and companies.
SK Group associates and associate firms throughout the semiconductor, vitality, AI information heart and agent companies sectors joined the annual discussion board at Coex in southern Seoul, sharing the most recent updates for his or her services to business and educational leaders.
“The exponential progress in demand for computing energy is much extra severe, as provide can not sustain,” Chey mentioned throughout his keynote speech. “It’s time to shift the paradigm from a contest of scale to a contest of effectivity.”
Chey mentioned demand for AI computing energy maintains a pointy progress trajectory with the rise of AI inferencing, which requires extra tokens than present training-based fashions. He added that the development is accelerating: As business-to-business AI adoption expands, agentic AI turns into extra widespread and nations pursue their very own sovereign AI initiatives.
SK Group Chairman Chey Tae-won, heart, shakes palms with a robotic exhibited by RLWRLD, a humanoid basis mannequin developer, on the 2025 SK AI Summit at Coex in southern Seoul, Monday. Korea Instances picture by Nam Hyun-woo
He mentioned that such “exponential progress in demand” is creating challenges for chipmakers, together with SK hynix, to fulfill the market demand for AI-specific high-bandwidth reminiscence (HBM), as a result of lengthy lag time required to increase provide and the problem of forecasting demand precisely.
He additionally identified that geopolitical components, reminiscent of balancing between totally different nations’ calls for and figuring out the place to speculate and the place to provide, are making it even tougher to handle manufacturing and distribution.
To handle these “bottlenecks,” Chey mentioned SK Group is specializing in “effectivity over scale.”
SK hynix is ready to start full-fledged operation of its new HBM fab subsequent yr in Cheongju, North Chungcheong Province. In 2027 it should full the primary fab in a cluster in Yongin, Gyeonggi Province, that can have 24 instances the capability of the Cheongju plant’s M15X fab when it’s absolutely operational. Chey mentioned that in finishing up these initiatives, the group is looking for to introduce breakthrough applied sciences reminiscent of ultra-high-capacity reminiscence and cost-efficient, large-scale NAND flash reminiscence merchandise to be extra environment friendly in assembly market demand. SK can be engaged on optimizing its information heart operations by designing and managing them from the bottom as much as finest help AI workloads.
“What we’re specializing in now’s fixing AI issues with AI,” Chey mentioned. “We’re making use of AI to each reminiscence chip fabrication and information heart operations to enhance effectivity and pace by way of course of optimization and virtualization. Particularly, we now have begun utilizing AI in reminiscence manufacturing by way of our collaboration with Nvidia, using its Omniverse platform to dramatically improve manufacturing effectivity.”
SK Group Chairman Chey Tae-won offers a speech in the course of the 2025 SK AI Summit at Coex in southern Seoul, Monday. Yonhap
Throughout Nvidia CEO Jensen Huang’s go to to Korea final week, SK Group and Nvidia agreed to increase their partnerships in bodily AI, by organising a producing AI system powered by greater than 50,000 Nvidia graphics processing items in addition to organising an precise AI manufacturing facility.
“SK is working with Nvidia to construct an AI manufacturing facility specialised for SK hynix by introducing superior digital options,” Chey mentioned. “Our purpose is to develop manufacturing AI that can absolutely automate the chip manufacturing course of.”
SK hynix CEO Kwak Noh-jung delivers a speech on the 2025 SK AI Summit at Coex in southern Seoul, Monday. Courtesy of SK hynix
Throughout a gathering with reporters after the keynote, Chey mentioned SK Group was in a position to make a quick transition to AI as a result of it “already had a portfolio tailor-made to AI.”
“To any extent further, every SK firm will develop its personal technique on the right way to leverage AI in manufacturing to strengthen its competitiveness,” he mentioned.
Relating to claims of an “AI bubble” that query the round capital circulate between chipmakers, GPU suppliers and AI service suppliers, Chey mentioned: “A transaction is a transaction, and calling {that a} bubble goes too far.”
He additionally famous that he needed to miss the informal assembly at a hen restaurant between Huang, Samsung Electronics Govt Chairman Lee Jae-yong and Hyundai Motor Group Govt Chair Chung Euisun, which went viral on-line and have become an web meme, as a result of he was internet hosting the Asia-Pacific Financial Cooperation CEO Summit final week. He added that Huang understood his absence.
In the course of the discussion board, OpenAI CEO Sam Altman and Amazon CEO Andy Jassy despatched separate video messages sharing their dedication to AI enterprise partnerships with SK Group. In a latest assembly between Chey and Altman, OpenAI requested 900,000 HBM wafers per 30 days for the Stargate information heart venture, whereas Amazon Net Companies (AWS) has teamed up with SK Group to construct a mega-sized AI information heart in Ulsan.
Following Chey, SK hynix CEO Kwak Noh-jung launched the corporate’s plan to reposition itself as “full-stack AI reminiscence creator.”
“Whereas previous reminiscence options centered on integration round computing, future reminiscence options will diversify and increase the function of reminiscence itself, enabling a lot higher use of computing assets and structurally resolving inference bottlenecks,” Kwak mentioned.
SK Telecom CEO Jung Jai-hun delivers a speech on the 2025 SK AI Summit at Coex in southern Seoul, Monday. Courtesy of SK Telecom
New SK Telecom CEO Jung Jai-hun additionally delivered a speech on the discussion board, saying the telecom firm’s initiatives to speed up the enlargement of its AI information heart enterprise, in addition to plans to cooperate with SK associates to introduce extra energy-efficient information heart options.
The summit was attended by not solely SK Group firms but additionally its peer firms, together with AWS, Nvidia, Schneider Electrical, Kakao and plenty of Korean AI startups. Earlier than his keynote, Chey toured cubicles of SK associates and collaborating firms, exhibiting curiosity in a number of startups together with Clify, an AI-powered mentality caring service agency, and RLWRLD (Realworld), a bodily AI firm growing basis fashions for humanoids.
SK Group Chairman Chey Tae-won visits SK Group’s sales space for the 2025 SK AI Summit at Coex in southern Seoul, Monday. Korea Instances picture by Nam Hyun-woo
Additionally on the discussion board, Tim Costa, Nvidia’s common supervisor of commercial & computational engineering, quantum, gave a presentation on AI supercomputing for next-generation semiconductor design and manufacturing. Kakao CEO Chung Shin-a additionally delivered a presentation on the way forward for AI brokers for sustainable private and social improvement.
