Analysis
SK Hynix priced its US share offering at $149 per American depositary share on July 9 and began trading on Nasdaq under the temporary ticker SKHYV on July 10, raising $26.5 billion in what is now the largest-ever US listing by a foreign company -- surpassing Alibaba's $25 billion IPO from 2014, a record that had stood for more than a decade. The South Korean memory-chip maker sold 177.9 million ADRs, structured so US investors could buy in at roughly a tenth of what a full share costs on the Korea Exchange in Seoul.
Demand was extraordinary by any measure: the offering was more than seven times oversubscribed, and the stock opened 14% above its IPO price before continuing to climb through early trading. That kind of reception for a foreign primary listing of this size is rare, and it reflects how central SK Hynix has become to the physical infrastructure of the AI boom rather than investor enthusiasm for Korean equities broadly.
SK Hynix's regulatory filing specifies that proceeds will fund new production facilities in Korea dedicated to expanding manufacturing of high-bandwidth memory, or HBM -- the specialized, stacked memory chips that sit alongside Nvidia's GPUs in AI accelerators and have become one of the tightest supply bottlenecks in the entire AI hardware stack. SK Hynix already supplies HBM to Nvidia at scale, competing primarily against Samsung and Micron in a three-way race that's arguably more consequential to near-term AI compute availability than the GPU race itself, since HBM shortages can bottleneck accelerator output even when GPU die supply is sufficient.
“The timing of the listing is notable against a backdrop of growing skepticism about AI infrastructure utilization.”
The timing of the listing is notable against a backdrop of growing skepticism about AI infrastructure utilization. The same week SK Hynix debuted, VentureBeat published survey data showing 86% of enterprises that run their own GPUs report utilization of 50% or less -- a genuine disconnect between how aggressively hyperscalers and memory suppliers are building capacity and how much of the deployed compute is actually earning a return. SK Hynix's own memory business benefits regardless of that utilization gap, since HBM demand is driven by hyperscaler capex commitments made well in advance of actual workload deployment.
For founders and investors, SK Hynix's oversubscribed debut is a signal that public-market appetite for AI-infrastructure exposure remains genuinely strong even as skepticism about buildout pace grows in parallel -- the two aren't mutually exclusive, and sophisticated capital is willing to buy the picks-and-shovels layer even while debating whether the mine itself is being overbuilt. For LPs benchmarking IPO market health, this listing alongside SpaceX's earlier $75 billion debut confirms that 2026's mega-IPO wave is real, even if it remains concentrated in a handful of infrastructure-critical names rather than broad-based issuance.
The bear case: SK Hynix's valuation now embeds continued explosive HBM demand growth, and any slowdown in hyperscaler AI capex -- or evidence that the 86% underutilization VentureBeat documented starts constraining future GPU and memory orders -- would hit the stock directly. What to watch next: how SKHYV trades once its temporary ticker converts and lockups begin expiring, and whether Samsung or Micron respond with their own US listing ambitions now that SK Hynix has proven the market's appetite.
