SK hynix projects AI infrastructure performance to grow 20-fold by 2030

SEOUL, Oct. 14 (Yonhap) -- SK hynix Inc. forecasted on Wednesday that the scale of performance for AI infrastructure will increase nearly 20 times by 2030, propelled by escalating demand for AI inference. During his opening remarks at the 2026 Semiconductor Exhibition (SEDEX), Ahn Hyun, the head of development at the South Korean chip manufacturer, disclosed the projection.

Ahn highlighted that the AI inference market is expanding concurrently across three domains: task complexity, information processing capacity at once, and the volume of inference requests. He emphasized that the combined impact of these factors is swiftly heightening the need for AI infrastructure. As the demand for AI inference surges, memory plays an increasingly crucial role in managing and transferring vast amounts of data.

Specifically, the expanding capacity demands of key-value (KV) caches, which store and reuse previously processed data by AI models, necessitate the distribution of data across various memory types, such as high-bandwidth memory (HBM), dynamic random-access memory (DRAM), and solid-state drives (SSDs). SK hynix is actively pursuing advancements in next-generation memory technologies, including 3D stacked DRAM and high-bandwidth flash (HBF), while simultaneously improving the performance of their current HBM offerings.