Hook

Friday’s listing of SK hynix on U.S. exchanges isn’t just another Korean chaebol debut. The data table that matters: HBM3E market share is 55%+ for 2024 Q3, yet the company’s trailing PE sits at 15-20x—half that of Micron and a third of Samsung’s semiconductor division. This is a metric anomaly screaming for a forensic examination. The market still prices hynix like a cyclical memory maker when its AI exposure is structural. But the real signal is buried in the concentration risk.

Context

SK hynix is the world’s second-largest DRAM maker and the undisputed leader in High Bandwidth Memory (HBM), the critical component powering NVIDIA’s Hopper and Blackwell GPUs. Unlike its rivals, hynix pioneered the MR-MUF packaging process, which gives it a 1-1.5 year lead in HBM yield and thermal performance. The U.S. IPO—selling Korean-listed shares via American Depositary Receipts—comes at the peak of the AI capex cycle. Yet the pricing suggests investors still see it as a DRAM/NAND cyclical play, not an AI infrastructure core. That gap is either an opportunity or a trap.

Core

Let the on-chain data speak for itself—though here “on-chain” means the supply chain. Three facts form the evidence chain:

1. Revenue composition shift: HBM and high-value DRAM now account for ~40%-50% of total revenue, up from <10% two years ago. This is not a blip; NVIDIA’s H100 and B200 demand locked in multi-year contracts. Gross margins hit 55%-60% in 2024, comparable to TSMC.

1. Capex deployment timeline: The Cheongju M15X HBM-dedicated fab, with billions of dollars in investment, targets equipment move-in by mid-2025 and mass production by late 2025. This is the only concrete expansion path for HBM capacity outside of Samsung. Any delay in equipment delivery—especially MR-MUF tools—becomes a hard constraint on NVIDIA’s GPU output. Investors should treat the fab’s progress as a critical metric.

1. Customer concentration: Over 80% of HBM sales flow to a single counterparty—NVIDIA. On-chain wallet analysis of NVIDIA’s procurement patterns would show that hynix supplies ~90% of its HBM3E. This is both a moat and a landmine. A second-source qualification by Samsung or Micron would instantly erode hynix’s pricing power.

Contrarian

The “too good to be true” angle here is the assumption that HBM leadership is permanent. Correlation does not equal causation: hynix’s current lead is a function of time-to-market execution, not an unassailable IP moat. Samsung has deeper pockets, a broader customer base, and is investing aggressively in HBM4. The real battle isn’t technology—it’s the ability to scale yield under NVIDIA’s relentless timeline. If Samsung matches hynix’s MR-MUF yield by late 2025, hynix’s 55% market share could collapse to 30% within two quarters. The U.S. IPO may be the peak of its relative advantage, not the beginning of a growth super-cycle.

Takeaway

Watch the Cheongju fab’s first wafer-out date. If it slips past Q3 2025, the narrative flips. The next-week signal: monitor Samsung’s HBM4 joint development announcements with NVIDIA. If Samsung secures a co-development deal before hynix’s IPO closes, the offering’s valuation thesis breaks.

This is not investment advice. Follow the data, ignore the hype.