South Korean chip giant SK Hynix predicted on Wednesday that, driven by rapid growth in AI inference demand, the performance scale of artificial intelligence infrastructure will increase by about 20 times by 2030. Ahn Hyun, head of R&D at SK Hynix, said at the 2026 Korea Semiconductor Show that the AI inference market is expanding simultaneously from three dimensions: task complexity, amount of information processed at a time, and total number of inference requests. The combination of these three factors is rapidly boosting the demand for AI infrastructure. As the demand for AI inference explodes, the importance of memory for storing and transmitting massive amounts of data continues to increase. In particular, demand for key-value cache capacity continues to rise. KvCache is used to store and reuse information previously processed by the AI model. This requires data to be distributed in different types of storage media, including high-bandwidth memory HBM, DRAM, and solid-state drive SSDs. The executive added that SK Hynix is deploying next-generation storage technologies, including 3D stacked DRAM and high-bandwidth flash HBF, while continuing to iterate on the performance of existing HBM products.

South Korean chip giant SK Hynix predicted on Wednesday that, driven by rapid growth in AI inference demand, the performance scale of artificial intelligence infrastructure will increase by about 20 times by 2030. Ahn Hyun, head of R&D at SK Hynix, said at the 2026 Korea Semiconductor Show that the AI inference market is expanding simultaneously from three dimensions: task complexity, amount of information processed at a time, and total number of inference requests. The combination of these three factors is rapidly boosting the demand for AI infrastructure. As the demand for AI inference explodes, the importance of memory for storing and transmitting massive amounts of data continues to increase. In particular, demand for key-value cache capacity continues to rise. KvCache is used to store and reuse information previously processed by the AI model. This requires data to be distributed in different types of storage media, including high-bandwidth memory HBM, DRAM, and solid-state drive SSDs. The executive added that SK Hynix is deploying next-generation storage technologies, including 3D stacked DRAM and high-bandwidth flash HBF, while continuing to iterate on the performance of existing HBM products.