Seoul, March 3, 2025 — SK Hynix CEO Kwak Noh-Jung officially announced a new vision, 'Full Stack AI Memory Creator,' at the ‘SK AI Summit 2025’ held at COEX in Seoul’s Samseong-dong.
During his keynote speech, Kwak emphasized SK Hynix's evolving role, stating, “Until now, SK Hynix has served as a ‘Full Stack AI Memory Provider,’ delivering products to meet customer demands in a timely manner. Moving forward, we aim to solve problems collaboratively with our clients, actively engaging with the ecosystem to create value beyond expectations.” He reaffirmed his commitment to becoming a "Co-Architect, Partner, and Ecosystem Contributor" as a creator of full-stack AI memory, reflecting a strong determination for the next era of AI.
Reflecting on the company's rapid growth over the past year—rising to become the global leader in the memory market and earning the title of 'Top Employer to Work For'—Kwak expressed his pride, noting, “These figures genuinely mean a lot.”
The Growing Importance of Memory Semiconductors in the AI Era
Kwak highlighted the increasing importance of memory semiconductors amid the acceleration of AI adoption. “Data traffic is exploding, and hardware technology is rapidly evolving to support this,” he explained. “However, the ‘Memory Wall’—the performance gap between processors and memory—remains a significant challenge,” he added. He stressed that memory has transitioned from a mere component to a core value product vital for AI performance, underlining that existing methods are reaching their performance limits.
SK Hynix’s New Vision: ‘Full Stack AI Memory Creator’
Kwak announced that SK Hynix’s role must expand in the AI era beyond simply delivering products as a ‘Full Stack AI Memory Provider.’ The new vision focuses on becoming a ‘Full Stack AI Memory Creator,’ actively collaborating with clients to tackle their challenges and deliver exceeding value. The company intends to play a crucial role as a co-architect of AI computing and as a partner and contributor within the broader ecosystem.
Future-Ready Full Stack AI Memory Lineup
Kwak outlined plans for diversified memory solutions, including upcoming products like ‘Custom HBM,’ ‘AI DRAM (AI-D),’ and ‘AI NAND (AI-N).’ The Custom HBM enables tailored functionality by transferring specific GPU and ASIC features onto HBM dies, optimizing system performance and energy efficiency. Development is shifting from general-purpose to inference efficiency and optimization, with sample shipments expected in 2026.
AI-D aims to optimize data centers with low-power, high-performance DRAM, including variants for flexible, high-capacity memory expansion. AI-N solutions, such as ‘AI-N P,’ ‘AI-N B,’ and ‘AI-N D,’ will facilitate next-generation storage for large-scale AI inference environments, offering high performance, bandwidth expansion, and large storage capacity.
Global Collaboration and Future Strategies
Kwak concluded by emphasizing SK Hynix’s extensive collaboration with global partners—including NVIDIA, OpenAI, TSMC, and Omniverse—as well as its work with Naver Cloud to advance AI memory and storage standardization. “We are working closely with various partners worldwide to lead the next wave of AI memory innovation,” he said.
