Why does SK hynix expect AI infrastructure performance to grow 20-fold by 2030?

SK hynix attributed the projected growth to expanding AI inference demand across three areas: increasingly complex tasks, larger amounts of information processed at once and a rising volume of inference requests.

How are key-value caches changing memory requirements for AI inference?

Key-value caches store and reuse information previously processed by AI models. Their growing capacity requirements are prompting data to be distributed among HBM, DRAM and SSDs rather than relying on a single memory type.

What next-generation memory technologies is SK hynix pursuing for AI infrastructure?

SK hynix is pursuing 3D stacked DRAM and high-bandwidth flash while continuing to improve existing HBM products. The company is also positioning different memory types to handle the storage and data-transfer demands of AI inference.

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

Listen

Summary

SK hynix projected that AI infrastructure performance will grow roughly 20-fold by 2030 in southern Seoul. Ahn Hyun, head of development at SK hynix, said rising AI inference demand is being driven by greater task complexity, larger information volumes and more inference requests. The company is developing 3D stacked DRAM and high-bandwidth flash while improving existing high-bandwidth memory products.

Key Facts

Ahn Hyun made the projection during a keynote speech at the 2026 Semiconductor Exhibition, which opened Wednesday for a three-day run.
Growing key-value cache capacity is requiring AI data to be distributed across high-bandwidth memory, dynamic random-access memory and solid-state drives.
SK hynix said AI inference demand is expanding simultaneously across task complexity, the amount of information processed at a time and the volume of inference requests.

By Yonhap

Published Oct 14, 2026 3:18 pm KST

Ahn Hyun, head of development at the Korean chip giant SK hynix, speaks at the 2026 Semiconductor Exhibition (SEDEX) in Seoul, Wednesday. Yonhap

SK hynix projected Wednesday that the performance scale of artificial intelligence (AI) infrastructure will grow roughly 20-fold by 2030, driven by rapidly rising demand for AI inference.

Ahn Hyun, head of development at the Korean chip giant, made the projection during his keynote speech at the 2026 Semiconductor Exhibition (SEDEX), which opened earlier in the day for a three-day run in southern Seoul.

Ahn said the AI inference market is expanding simultaneously in three areas — task complexity, the amount of information processed at a time, and the volume of inference requests.

The interaction of the three factors is rapidly increasing demand for AI infrastructure, he said.

As demand for AI inference surges, memory is becoming increasingly important for storing and transferring large volumes of data.

In particular, the growing capacity requirements of key-value (KV) caches, which store and reuse information previously processed by AI models, are requiring data to be distributed across different types of memory, including high-bandwidth memory (HBM), dynamic random-access memory (DRAM) and solid-state drives (SSDs).

SK hynix is pursuing next-generation memory technologies, including 3D stacked DRAM and high-bandwidth flash (HBF), while continuing to enhance the performance of existing HBM products, the executive added.

Explore More