SK Hynix raises $26.5B in biggest foreign IPO
SK Hynix raised $26.5B on Nasdaq on July 10, the largest US listing ever by a foreign company. The HBM maker controls 60% of global AI memory.
SK Hynix priced its US ADR offering on July 10 at $149 per share, selling 177.9 million American Depositary Shares to raise $26.5 billion, the largest US listing ever by a non-American company, topping Alibaba’s $25 billion IPO from 2014. Shares opened at $170, a 14% premium to the IPO price, and closed the day up 12.8%. The company begins regular trading on Nasdaq under the ticker SKHY on July 13.
The offering was more than seven times oversubscribed. Proceeds will fund a new semiconductor fabrication plant in South Korea, a next-generation HBM packaging facility, and EUV scanning equipment.
Why this matters beyond the market story
SK Hynix supplies the high-bandwidth memory inside Nvidia’s H100 and H200 GPUs, the chips that train and run most large-scale AI models. The company controls approximately 60% of the global HBM market, with Samsung and Micron splitting the remainder.
HBM availability has been a real constraint on AI hardware procurement for the past two years. Nvidia has repeatedly cited memory packaging capacity, not chip fabrication, as the limiting factor in GPU supply. The $26.5 billion in fresh capital goes directly to expanding that capacity: a new fab addresses leading-edge DRAM production; the packaging facility addresses the HBM stacking step that has been the actual bottleneck.
Commerce Secretary Howard Lutnick used the IPO occasion to publicly push SK Hynix and Samsung to build US manufacturing capacity, pointing to Micron’s $250 billion US commitment as a benchmark. SK Hynix did not announce US plants.
What to watch
The IPO filing (covered when it was submitted) told builders that SK Hynix had US ambitions. The completion of the raise, at a premium that signals strong investor confidence in AI memory demand, tells you that the capital is now in hand and spending can start. Lead times on HBM-based hardware tend to be 6 to 18 months from factory commitment to customer delivery. The new capacity funded today shows up in supply in 2027 to 2028.
For how hardware availability fits into long-range AI build planning, see what the AI chip race means for you.