Sk hynix forecasts 20-fold growth in AI infrastructure performance by 2030

At the SEDEX 2026 semiconductor exhibition in Seoul, SK Hynix's Chief Development Officer Ahn Hyun predicted that AI infrastructure performance will grow by approximately 20 times by 2030. He highlighted the rapid expansion of the AI inference market across three dimensions: increased task complexity, larger context lengths, and a surge in inference requests. These factors are driving significant growth in demand for AI computing infrastructure. Ahn emphasized that storage is becoming a critical bottleneck due to the expansion of key-value cache (KV Cache) capacity, forcing data to be distributed across high-bandwidth memory (HBM), DRAM, and SSDs. This data migration is now a major constraint on performance and energy consumption. To address this, SK Hynix is advancing next-generation storage technologies, including 3D stacked DRAM and high-bandwidth flash (HBF). HBF aims to achieve 50% of HBM's bandwidth with 10 times the capacity.