SK Hynix unveiled a broad portfolio of memory and storage products for AI infrastructure at Dell Technologies Forum 2026 in Seoul, deepening its long-term collaboration with Dell Technologies. The lineup included sixth-generation HBM4, seventh-generation HBM4E, AI server DRAM, enterprise and client SSDs, and memory technologies such as 3D stacking, processing-in-memory and Compute Express Link-based memory pooling. SK Hynix Vice President of System Architecture Joo Young-pyo said agentic AI workloads may outgrow the performance gains available from conventional HBM improvements alone, citing a GPT-4-class mixture-of-experts model that requires 3.5 terabytes of memory for weights and reads 4.4 terabytes per second to generate 20 tokens per second. Dell Technologies executives said companies must redesign workflows and treat AI spending as a variable cost, with a hybrid architecture distributing workloads among cloud services, corporate data centers and local devices. Dell Technologies Korea President Yoo Sang-mo said 61% of enterprise leaders feel they need to integrate AI into the workplace, while 64% lack a clear execution roadmap. Matt Dunphy, Dell's senior vice president, said cost efficiency is central as agentic AI increases token use and management requirements. The forum also covered Dell's AI Factory, data platforms, local AI tools, security and infrastructure readiness, alongside reported growth in Dell's client business and a broader technology-sector rise in Dell shares.