JPMorgan Initiates Coverage on SK Hynix ADR with $245 Price Target
nashnova research
J.P. Morgan initiated coverage of SK Hynix's ADR at overweight with a $245 target, implying 32% upside; the core thesis is that the stock trades at an unjustifiable discount to Micron while an AI-driven memory super-cycle is stretching far beyond historical norms.
Why does J.P. Morgan think SK Hynix ADR is undervalued?
SK Hynix ADR trades at 5.8× forward P/E versus Micron's 6.5× — an ~11% discount. This means → the market is pricing two fundamentally comparable companies differently, and J.P. Morgan sees no good reason for it.
In plain terms = SK Hynix matches or beats Micron on DRAM margins, absolute scale, and HBM — high-bandwidth memory, the ultra-fast data store sitting next to AI chips — execution, yet its stock trades at roughly a 20% discount on broader measures.
J.P. Morgan expects the gap to narrow as the ADR listing improves global investor access, liquidity, and disclosure frequency.
How is the $245 target calculated?
The target is set for June 2027, based on 7× average 2026-2027 EPS for SK Hynix's local shares, plus a 20% ADR premium.
The 20% premium benchmarks TSMC's ADR, which has sustained a similar premium over its local shares since the AI capex-driven rally began in 2024.
SK Hynix ADR currently trades at an ~30% premium — above J.P. Morgan's assumed 20%. This reflects a structural dynamic: only 2.5% of total shares are in ADR float, and Korean regulations restrict local-share-to-ADR conversions, keeping the premium elevated.
How long can this memory upcycle last?
J.P. Morgan's call: DRAM ASP growth started in Q1 2024 and will extend past Q4 2028 — over 20 quarters of positive pricing, versus the historical norm of seven to eight. In plain terms = this cycle could run more than twice as long as any before it.
Demand side: hyperscaler bit-demand for DRAM and NAND is forecast to grow 60% and 58% YoY in 2027, respectively. Supply side: HBM's share of DRAM wafer capacity is projected to hit 31% by 2028, and bit output per unit of capex keeps falling — both cap supply growth.
J.P. Morgan projects global memory market revenue rising from $214 billion in 2025 to $1.44 trillion in 2027. This means → nearly a six-fold expansion in two years.
How deep is the HBM moat?
SK Hynix is the world's largest HBM supplier, holding an ~60% market share in 2025. Nvidia is its top customer, accounting for ~74% of HBM sales.
J.P. Morgan expects Samsung's improving execution to pull SK Hynix's share down to 40-46% from 2026 onward — still the market leader by a wide margin.
Long-term agreements (LTAs) — contracts locking in price and volume — already cover more than 50% of capacity, with upfront payments at 20-25% of total LTA value. This means → even in a downturn, a large portion of revenue is contractually secured. AI-related demand under LTAs accounts for over 85% of revenue, which J.P. Morgan sees as evidence that memory is shifting from a cyclical business to a structural one.
Can shareholder returns catalyze a re-rating?
In August 2026, SK Hynix raised its return policy from "up to 50% of cumulative free cash flow" to "more than 50%," and announced a KRW 40 trillion buyback-and-cancellation program — equivalent to 63% of H1 2026 free cash flow.
J.P. Morgan forecasts total shareholder return yields of 7.4%, 14.0%, and 20.4% for 2026-2028, a cumulative ~42% over three years.
The nearest hard catalyst: the Q3 earnings call in late October 2026. In plain terms = the market wants more than a promise of "we'll return more cash" — it wants a concrete timetable and dollar figure. Clearer capital-allocation guidance at that event would directly support the stock.
How fast will SK Hynix earnings grow over the next three years?
J.P. Morgan projects a 34% EPS CAGR for 2026-2028: +510% YoY in 2026, +27% in 2027, and +42% in 2028.
This means → 2026 is the breakout year — a low base meets surging AI demand — and growth moderates afterward but stays well above the industry average.
This reflects how AI-driven HBM demand is reshaping SK Hynix's earnings profile: no longer the boom-bust of traditional memory, but a rising staircase.
市场有风险，内容仅供研究参考，不构成投资建议。