SK hynix Charts Its Business and Technology Direction at the 2026 Future Forum—”Now Is the Golden Time to Leap Forward”

SK hynix held its 2026 Future Forum on September 8 at the SUPEX Center on its Icheon Campus, where CEO Kwak Noh-Jung said the AI era presents a 'golden time' for the company to leap forward, emphasizing the need to shift perspective 'from the outside in' and adapt to a 'winding road' market. The forum featured keynote presentations and a panel discussion, including Thariq Shihipar, Head of Product at Anthropic, who discussed how AI collaboration is reshaping work and the potential for a 'Software Factory' in the semiconductor industry.

On September 8, SK hynix held the 2026 SK hynix Future Forum Future Forum at the SUPEX Center on its Icheon Campus under the theme, “The Golden Time of the AI Era: The Moment AI Memory Solution Creators Change the World,” sharing new trends and insights on AI Transformation AX . The Future Forum is designed to define the path SK hynix should take in the rapidly changing AI era and to help enhance the insight of its employees. Following the 2024 forum on the proactive exploration of AI technologies and the 2025 forum on innovation prospects across the AI value chain, this year’s forum took a broad view of the AI era in which work and environments are being transformed and the boundaries between business and technology are dissolving. In particular, it focused on examining ‘what we should do’ and on defining ‘the direction in which memory should advance.’ The event was made up of three keynote presentations and a panel discussion. Industry figures from Korea and abroad, along with professors from leading universities, took part in the presentations and discussion, adding to their weight. Key SK hynix executives also shared their memory insights, including Vice President Kyunghoon Kim, head of the Design Research Lab; Vice President Hyungsoo Kim, head of the DRAM Design department; Vice President Hoyoung Son, head of the Advanced Package Development department; Vice President Euicheol Lim, head of the Solution AT department; and Vice President Jaewuk Ju, head of the Smart Manufacturing Technology/AX Advance department. The forum opened with welcoming remarks from SK hynix CEO Kwak Noh-Jung. At the event, CEO Kwak stressed the importance of shifting our perspective “from the outside in,” looking anew at the things we have long taken for granted through the eyes of our customers and partners, and seeking possibilities we have not seen before amid the changes in the AI ecosystem. He drew an analogy: while the memory market of the past was a “straight road” where the race was about who could run faster, today it is like driving on a “winding road” that shifts from moment to moment. On this point, he emphasized that what matters is not only internal capabilities but also the speed with which the company reads the pace and direction of external change and becomes flexible. In closing, CEO Kwak said he hopes that, drawing on the insights gained through this year’s SK hynix Future Forum, “We can reflect together on how our company should move forward and change accordingly.” AI Collaboration Reshapes How We Work, Calling for New Directions in Memory The keynote presentations began by diagnosing how AI has changed the way we work. Thariq Shihipar, Head of Product at Anthropic; author Gilyoung Song; and SK hynix Vice President Jaewuk Ju took the stage in turn to open the discussion under the theme, “Human-AI Collaboration: ‘How Will Our Work Change?’” Drawing on real-world usage data from ‘Claude,’ Thariq Shihipar, Anthropic’s Head of Product, explained how AI is changing the way we work. He said that more experienced professionals tend to achieve better results because they know what to delegate and how to judge the outcomes, and he laid out adoption trends among global companies along with their implications for the semiconductor industry. He added that the semiconductor industry, too, can build a ‘Software Factory’ that applies AI across a range of tasks—such as circuit simulation, power measurement, and system optimization—and stressed the importance of putting in place a framework in which AI can verify its own results. He also predicted that once AI agents are connected to hardware such as robots, their scope of use will widen across the manufacturing floor. As a result, he said, AI will handle repetitive, verifiable tasks, while people concentrate on higher-value roles—drawing on their expertise to generate new ideas and make decisions. Author Gilyoung Song pointed to the ‘redesign of organizations and systems’ as the key change brought by AI adoption. As the use of AI moves beyond the stage of individuals simply using it as a tool, he noted, organizations have entered a phase in which the roles of people and AI must be redefined. Noting that there are limits to merely speeding people up while leaving the system unchanged, he stressed that individuals must be able to define their own expertise and skills. Vice President Jaewuk Ju walked step by step through SK hynix’s preparations for AI collaboration and the resulting changes to work, from redefining tasks to applying them in the field. Defining ‘AI as both a partner in collaboration and another member of the team,’ he presented a working environment in which people can concentrate on judgment. He then introduced examples of the work innovations SK hynix is pursuing to make this a reality. On the production floor, he explained, SK hynix is redesigning how people work alongside AI based on standard operating procedures , and is building a foundation AI can draw on by defining the meaning and relationships within process and equipment data. On this basis, he anticipated that work will shift toward a model in which, when a line anomaly occurs, AI analyzes the relevant data to suggest causes and countermeasures, while people concentrate on final judgment and decision-making. Vice President Ju noted, “We are preparing to run production within the fab through the GaiA Platform and are using AI across the board—including deploying safety-supervision robots and VLM drones for work at height.” He added, “I believe these efforts strengthen the company’s capabilities and will contribute greatly to our leap forward as a Full-Stack AI memory creator.” VLM Vision-Language Model : An AI model that can understand and analyze image and video information together with text Such advancement of AI is now demanding new changes in memory solutions as well. In the second batch of presentations, Professor Sungroh Yoon of Seoul National University, Jooyoung Kim, CEO of HyperAccel, and SK hynix Vice President Euicheol Lim continued the discussion under the theme “The Memory Solutions the AI Era Demands: ‘Beyond Commodity’” Professor Yoon first explained that as AI workloads evolve from reactive AI to agentic AI , the role of memory is changing as well. As the state information AI State that AI must remember and manage to carry its work forward grows in volume—and becomes more varied in lifespan and access characteristics—a single general-purpose memory is no longer enough to keep up. Designing a memory hierarchy suited to the characteristics of each workload, he predicted, will become a core competitive advantage for future AI systems. Reactive AI: The most basic form of AI, which responds only to current inputs and produces outputs according to fixed rules Agentic AI: An AI system that sets its own goals, formulates plans, and executes tasks without human intervention AI State: The state information—such as conversation context, task progress, and intermediate results—that AI must remember and manage to continue its work Kim then diagnosed that as multi-agent operations and long-running tasks expand, new bottlenecks are emerging in areas such as memory capacity and bandwidth. To address this, he stressed, HBM, DRAM, CXL, SSD, and new memory tiers should be deployed to suit each workload and optimized at the system level together with a range of accelerators, thereby improving both the performance and the total cost of ownership of AI infrastructure. Vice President Euicheol Lim introduced SK hynix’s ‘Full-Stack AI Memory’ strategy for responding to these changes. The core idea is to combine a range of memory—including 3D stacked DRAM, HBM, and HBF—to suit the characteristics of each workload, and to build optimal solutions through ‘system-level co-design’ with ecosystem partners across AI services, software, and accelerators. Through this, Vice President Lim set out how SK hynix will evolve beyond a supplier of memory devices into a Full-Stack AI memory creator that understands AI systems and designs them together with its partners. Memory Evolves Through 3D Technology — Exploring Opportunities With Outside Experts The third batch of keynote presentations was delivered under the theme of “Expanding the Memory Boundary Through 3D Technology” Professor Changhwan Shin of Korea University along with SK hynix Vice Presidents Kyunghoon Kim and Hoyoung Son took the stage. The three experts pointed to 3D memory as the ‘future memory technology’ to prepare for in the AI era and carried the discussion forward. Professor Shin described 3D not simply as a technology for stacking chips vertically, but as ‘a technology that removes the distinction between memory and computing chips.’ He argued that device scaling has reached its limits and that, with rising data transfer times and power consumption between systems and memory, the shift to 3D is unavoidable. He proposed a 3D integrated design framework that links four domains—materials, devices, packaging, and architecture—through AI. Next, Vice Presidents Kyunghoon Kim and Hoyoung Son shared the directions and challenges of the 3D technology SK hynix is preparing. They divided 3D technology into device integration and heterogeneous integration , and presented as its main technical directions ▲maximizing data bus width, ▲ultra short reach transmission, and ▲turning DRAM peripheral circuits Peri. into a logic foundry. Heterogeneous Integration: Attaching separately manufactured, different chips together afterward They explained that implementing this technology will require solving not only design and heat-dissipation issues but also technical challenges in the packaging domain, such as fine interconnects, bonding, and process integration. They also anticipated that, as the boundaries grow closer—not only the technical boundary between front-end Fab and back-end Packaging processes, but even the technical and business boundaries between foundry and memory IDMs—cross-domain co-optimization and new collaboration frameworks will become increasingly important. On this point, Vice President Kim said, “Bringing 3D stacked DRAM into an actual product will require solving not just design optimization but also technical challenges unlike those of the past, such as heat dissipation and process complexity.” “3D technology is even reshaping the technical boundary between fab and packaging,” emphasizing that “along with innovation in advanced packaging, it is important to build cross-domain co-optimization and new collaboration frameworks that go beyond existing boundaries.” Vice President Son added. The panel discussion, “Memory Outside-In: ‘New Perspectives, New Opportunities’,” offered a chance to look at where SK hynix stands today and the direction it should take next through the lens of the broader ecosystem. The panel brought together April Li, Director of AI & HPC Business Development at TSMC; Jooyoung Kim, CEO of HyperAccel; Hyungjun Kim, CEO of SqueezeBits; Sunghyun Park, CEO of Rebellions; and Jungsoo Kang, Head of the Research Center at Bluedot AI. Serving as moderator, SK hynix Vice President Hyungsoo Kim opened the discussion by examining how today’s memory boom differs from those of the past. He then sought the panelists’ insights on what conditions the memory that opens the next market must meet, and on what more the company needs to leap forward as a Full-Stack AI memory creator. This year’s Future Forum also stood out for weaving AI throughout its entire preparation process and for being conducted in both Korean and English, with an SK hynix global intern serving as host. At the venue, AI-generated introduction videos for each agenda item were screened, while an AI interpretation and captioning program handled real-time communication with overseas speakers. A humanoid robot served as an assistant host, helping to guide the proceedings. Filled with in-depth keynote presentations and a rich array of AI technologies, the event drew to a close with the remarks of SKHU President Hyeongsoo Kim. He said, “The true meaning of the SK hynix Future Forum, I believe, is to create a moment where the insights we gained today transform the way we work and the frameworks of our thinking.” “As we have already experienced with HBM, the power to expand the boundaries of semiconductor technology and to create new memory solutions lies in the curiosity and creativity of everyone here,” the SKHU president added. SK hynix will reprocess the insights drawn from the 2026 Future Forum into learning content for employees, share them internally, and connect them to the tasks of each organization to carry them through into action. In doing so, the company plans to respond in a timely manner to the ‘golden time’ of the AI era and to build a solid foundation needed to leap forward as a Full-Stack AI memory creator.