SK Hynix has officially entered the US equity market, raising $26.5 billion through a secondary ADR listing on the Nasdaq under the ticker SKHY. Priced at $149 per ADR, the offering was 7x oversubscribed, eclipsing the $25 billion Alibaba IPO of 2014 to become the largest US listing by a foreign company. While the firm has maintained a presence on the Korea Exchange since 1996, this Nasdaq debut provides the liquidity necessary to fund the massive infrastructure requirements of the ongoing AI memory supercycle.
The capital raise, managed by Bank of America, Citigroup, Goldman Sachs, and J.P. Morgan, is earmarked for new fabs and equipment to expand High Bandwidth Memory (HBM) production. This investment is a direct response to a fundamental shift in the semiconductor industry. According to Wedbush, the sector has transitioned from a compute-limited era to a bandwidth-limited one, where the speed at which data moves to the processor has become the primary constraint on AI performance.
SK Hynix currently commands a 56-62% share of the HBM market, with over 50% of the HBM3 and HBM3E segments. Their strategic position is bolstered by a multi-year partnership with NVIDIA, which includes co-developing next-generation memory like HBM4 and the Vera Rubin architecture. According to UBS, SK Hynix holds approximately 70% of the HBM4 allocation for NVIDIA’s Rubin chips. With the global HBM market projected to reach $54.6 billion in 2026 — a 58% year-over-year increase — the company is leveraging its market dominance to secure long-term supply agreements.
Despite the influx of capital, the immediate supply landscape remains constrained. Wedbush reports that all HBM4 production from the three major suppliers — SK Hynix, Samsung, and Micron — is sold out through 2026. While Samsung is projected to increase its HBM share above 30% and Micron is positioning itself as a primary challenger for the second spot in HBM4, the current reality is one of total capacity utilization. The $26.5 billion raised by SK Hynix is not a solution to current scarcity, but rather a down payment on future capacity.
For AI labs and agent builders, the bottleneck is not just in the silicon, but in the memory architecture surrounding it. As labs push for larger models and more complex agentic workflows, the reliance on HBM4 becomes absolute. With production for this critical component fully committed through the end of the year, developers must navigate a landscape where hardware availability is dictated by multi-year supply contracts rather than spot market access.
A significant portion of this capital will flow into the $4 billion advanced packaging facility in West Lafayette, Indiana. However, the timeline for this expansion highlights the gap between capital deployment and operational output. Construction began in April 2026, but production is not scheduled to come online until the second half of 2028. This two-year lag between the current capital raise and the realization of new domestic capacity underscores the structural rigidity of the AI supply chain.
The combined market capitalization of the three primary HBM suppliers — SK Hynix, Samsung, and Micron — has reached approximately $4.1 trillion, according to market data. This valuation reflects the market’s belief in the longevity of the AI memory supercycle. Yet, for the industry, the challenge remains the same: the transition from a capital-intensive phase to a capacity-constrained reality. While the Nasdaq listing provides SK Hynix with the necessary liquidity to scale, the physical limitations of fab construction and the pre-sold nature of HBM4 capacity suggest that the bandwidth bottleneck will persist well beyond the current fiscal year.