Menu

sPEG Index · watchlist

SK hynix

SDS 13/20 · [3 · 3 · 3 · 2 · 2] · ordinal

HBM design, manufacturing, and packaging capability inside a broader DRAM and NAND business.

Reviewed 2026-09-17 · Next review 2026-12-16 · released speg-index-2026-09-17-v1

Five judgments

Constraint · Moderate

3/4

HBM capability and packaging qualification create meaningful supply barriers with active qualified rivals.

Counterevidence: Samsung reports HBM4 sales and HBM4E sampling, while announced new capacity is not current qualified output.

Substitution resistance · Moderate

3/4

Customer qualification imposes material burden, though alternative supply exists.

Counterevidence: Samsung reports HBM4 sales and HBM4E sampling, while announced new capacity is not current qualified output.

Economic capture · Moderate

3/4

Current operating capture is visible but partly coincides with memory-price increases.

Counterevidence: Samsung reports HBM4 sales and HBM4E sampling, while announced new capacity is not current qualified output.

Persistence · Moderate

2/4

Contract resets and rival ramps leave three-to-five-year retention unresolved.

Counterevidence: Samsung reports HBM4 sales and HBM4E sampling, while announced new capacity is not current qualified output.

Cost of defense · Moderate

2/4

Continued capital investment makes residual economics plausible but sensitive.

Counterevidence: Samsung reports HBM4 sales and HBM4E sampling, while announced new capacity is not current qualified output.

Admission gates

dimension floors: fail — At least one dimension is unknown or below its floor.
materiality: fail — Company-level materiality is unassessable.
decisive confidence: pass — Every decisive dimension is at least Moderate confidence.
source packet: fail — Minimum admission packet is incomplete.
current review: pass — Structural review is current at the research cutoff.
material event: pass — No unresolved material eligibility event is recorded at the research cutoff.

Rationale, falsifiers, and next observation

13/20; watchlist Q2 release reports HBM4 mass shipments, approximately ten customer LTAs and capacity expansion; it attributes strong economics partly to memory-price increases. The Indiana project targets production in the second half of 2029. Samsung reports HBM4 sales and HBM4E sampling, an active rival response. [H1–H2, X2] **Judgment:** C3/S3 reflect qualified memory/packaging capability; E3 reflects operating capture rather than net-income windfalls; P2 reflects unresolved pricing and competition through renewals; D2 recognizes continued investment. The narrower HBM thesis cannot automatically cover all DRAM/NAND economics. Moderate confidence does not mean LTAs fix prices or margins. **Admission evidence needed:** contract coverage and resets, customer retention across generations, and retained economics after rival ramps/defense spending sufficient for P3. Next results or material qualification update/by December 16. Do not import the old internal symbol “SKH” as a verified current US security identifier; the issuer record and selected share class require separate verification before price/PEG ingestion.

Rival qualification erodes allocation or economics
Contract resets fail to retain price-cost benefit
Required investment absorbs retained economics

Next: Verify contract coverage and resets, customer retention across generations, company materiality, and economics after rival ramps.

Valuation context

PEG: unavailable (qualified reference path not assembled). Economic sPEG: unavailable (paired valuation uncalibrated). SDS is not a valuation input.

Source packet

X2 · Samsung Electronics · Samsung Q2 results — Rival HBM/foundry activity; plans not equivalent qualified scale H1 · SK hynix · SK hynix Q2 results — Preliminary K-IFRS, current capture not guaranteed future rents H2 · SK hynix · SK hynix Indiana project — Future construction/ramp, not today's qualified output