SK Hynix CEO predicts worst memory shortage in 2027, supply to lag demand beyond 2030
Global HBM and DRAM supply crunch could last years, warns SK Hynix CEO Kwak Noh-jung.
SK Hynix CEO Kwak Noh-jung told Reuters that the global memory industry is heading for its worst supply crunch ever in 2027, with demand for high-bandwidth memory (HBM) and DRAM expected to outstrip supply well beyond 2030. The comments came as the South Korean chipmaker began trading on Nasdaq under ticker SKHY, raising roughly $27 billion from its ADR offering. Kwak emphasized that even with aggressive capacity expansion — including a $4B advanced packaging plant in Indiana, a $10B U.S. AI solutions venture, and massive domestic investment in Yongin — customer demand continues to outpace the company's ability to produce. The shortage is driven by the relentless buildout of AI data centers requiring HBM, which SK Hynix leads as a key Nvidia supplier.
The memory chip oligopoly — SK Hynix, Samsung, and Micron — controls the bulk of global supply, but new Chinese entrants CXMT and YMTC are beginning to play a role. Kwak noted that the industry's cyclical nature makes this shortage unprecedented in duration, comparing it to past waves where supply eventually caught up. He cautioned that the current boom may be near a turning point, as markets start worrying about overinvestment. Both SK Hynix and Samsung are part of a South Korean government plan to double the country's memory production within five years, a 400 trillion won ($266 billion) investment that some investors fear could amplify downside risk in a downturn. The shortage is expected to keep memory prices elevated, impacting GPU costs and AI infrastructure buildout for the next several years.
- SK Hynix CEO forecasts worst memory supply crunch in 2027, with demand exceeding supply beyond 2030
- Company raised $27B via Nasdaq ADR listing (SKHY) to fund new factories in Indiana, US, and Yongin, South Korea
- HBM shortage threatens Nvidia GPU production; Samsung and Micron also racing to expand capacity
Why It Matters
Prolonged memory shortages mean higher AI infrastructure costs and potential delays in GPU supply for years.