Key facts
- SK Hynix plans to triple its wafer capacity by 2034 to meet AI chip demand.
- Chairman Chey Tae-won forecasts memory shortages, including HBM, will continue until 2030.
- Nvidia CEO Jensen Huang publicly urged SK Hynix to increase production.
- SK Hynix's 2026 HBM output is already sold out.
- Japan is considered an ideal location for SK Hynix's future overseas expansion.
SK Group Chairman Chey Tae-won announced that SK Hynix intends to triple its wafer capacity by 2034, a significant expansion driven by the escalating demand for high-bandwidth memory (HBM) chips essential for artificial intelligence computing. This ambitious plan represents a substantial increase from earlier projections to double capacity within the next five years.
Chey reiterated his forecast that a structural memory supply shortage, particularly for HBM, is likely to persist through 2030, with industry-wide wafer supply potentially falling more than 20% below demand. He highlighted that the production of HBM consumes considerably more wafer capacity per bit compared to conventional DRAM, and that building new fabrication facilities is a lengthy and costly process, taking at least three years from an existing site and over five years for a greenfield location.
The urgency for increased production was underscored by Nvidia CEO Jensen Huang, who visited the SK Hynix booth at Computex 2026 and publicly inscribed "Please Make More" on an HBM4E wafer. This direct appeal from a key customer, whose platforms are heavily reliant on HBM, emphasizes the critical supply constraints. SK Hynix's entire HBM production for 2026 is reportedly already sold out, and some customers have reportedly offered to prefund new production lines due to the near-zero available capacity.
Chey also indicated that Japan is a strong candidate for SK Hynix's future overseas manufacturing operations. He noted ongoing collaborations with partners such as TSMC, Foxconn, and Acer, emphasizing the need for expanded partnerships as the AI business grows. Chey suggested that South Korea could learn from Taiwan's proactive embrace of the AI era and accelerate its own response.