brain/
← all entities
entityHXSCLstock-market

SK Hynix

Notes

SK Hynix

One-line summary: Dominant HBM supplier; Q2 2026 KRW 79.3T revenue / 76% operating margin (official, 2026-07-29); ~10 LTAs with deposits (amounts undisclosed); M15X mass-production pulled forward; named Big Tech fab-equity bankroll still unfiled.

What it is

South Korean integrated memory manufacturer, focused on DRAM and NAND. SK Hynix leads the HBM market by production volume and customer relationships — Nvidia is its primary HBM customer for GPU stacking. First to volume production of HBM4 (February 2026).

Why it matters to stock-market

SK Hynix is the load-bearing constraint in the hbm-supply-bottleneck thesis. Its 59% HBM market share and 3-year demand backlog mean that supply normalization cannot happen without either SK Hynix capacity expansion or a competing vendor (Samsung, Micron) closing the yield/volume gap. The Big Tech backstop financing dynamic — if confirmed — would be a structural shift in how supply bottlenecks in AI infrastructure are resolved.

Key facts

  • Q1 2026 results (April 22-23, 2026): Revenue KRW 52.6T (first time >KRW 50T quarterly), +198% YoY, +60% QoQ; operating profit KRW 37.6T; operating margin 72% (all-time high, higher than both Nvidia and TSMC); net margin 77%. From 2026-05-21-autoresearch-hbm-supply-update-micron-q2-sk-hynix-may-2026.
  • HBM market share: 59% (Q1 2026 context, NineScrolls) — vs. 43% earlier TrendForce full-year estimate. The gap likely reflects SK Hynix's faster HBM4 ramp vs Samsung's yield lag. From 2026-05-21-autoresearch-hbm-supply-update-micron-q2-sk-hynix-may-2026.
  • 3-year demand backlog: SK Hynix management stated customer HBM requests already exceed planned production capacity for the next 3 years. From 2026-05-21-autoresearch-hbm-supply-update-micron-q2-sk-hynix-may-2026.
  • M15X fab (Cheongju): May 21 source said mass production November 2026, +~350K wafers/month → 900K WPM. ⚠ Contested: later trade press and the originating May 8 bankroll rewrite put the +350K / ~900K total on Yongin Y1, and M15X at ~50–90K wpm. Official Q2 2026 (2026-07-29) says only that the company is pulling forward the M15X mass-production schedule, with Yongin Phase 1 cleanroom in early 2027 — no official wafer-start restatement. From 2026-05-21-autoresearch-hbm-supply-update-micron-q2-sk-hynix-may-2026; correction from 2026-08-26-autoresearch-sk-hynix-q2-direct-supply-financing.
  • Big Tech backstop financing (reported, still unfiled as fab equity): Alphabet, Meta, Microsoft were reported (ad-hoc-news.de / ChosunBiz) to have offered dedicated-line or EUV funding. Reuters (2026-05-08) left the proposers unnamed. Official Q2 confirmed LTAs with ~10 customers and deposits, not fab equity, and did not name those three. See sk-hynix-big-tech-direct-supply-financing. From 2026-05-21-autoresearch-hbm-supply-update-micron-q2-sk-hynix-may-2026; Q2 update from 2026-08-26-autoresearch-sk-hynix-q2-direct-supply-financing.
  • Q2 2026 results (official, 2026-07-29): Revenue KRW 79.3187T (+51% QoQ, +257% YoY); operating profit KRW 60.5426T (76% margin); net income KRW 93.9226T; cash KRW 88T; net cash KRW 69.4T. HBM4 mass shipments began in Q2. 2026 capex guided to the high-KRW-40T range. Nasdaq ADR listed 2026-07-10. From 2026-08-26-autoresearch-sk-hynix-q2-direct-supply-financing.
  • LTAs (official Q2 call): ~10 customers including key customers; typically ~five years; volume commitments plus deposits; pricing "not uniform" / designed to address volatility; take-or-pay not said; LTA share of sales and individual terms not disclosed; NVIDIA named as the HBM partnership example. From 2026-08-26-autoresearch-sk-hynix-q2-direct-supply-financing.
  • HBM4 production (February 2026): 160K WPM; HBM4E sampling H2 2026 (faster than expected, compressing the HBM4 cycle). Nvidia-qualified. From 2026-05-21-autoresearch-hbm-supply-update-micron-q2-sk-hynix-may-2026.
  • ASML High-NA EUV: 2 units in 2026 — investing at the leading edge of memory process tools. From 2026-05-11-autoresearch-picks-shovels-semicap-update-may-2026.
  • 2016 Hynix revenue context: The KRW 52.6T quarterly figure dwarfs historical annual revenues — this is the inflection point of AI memory economics.
  • Still #1, but no longer exclusive: Counterpoint Q3 2025 HBM revenue share — SK Hynix 57%, Samsung 22%, Micron 21%. SK Hynix remains Nvidia's lead HBM4 supplier and benefits from a sold-out market, but Samsung's clearance of Nvidia HBM3E qualification (Sept 2025) removes the "Samsung excluded" tailwind — the market is now genuinely three-way competitive. The bull case rests on SK Hynix's lead and margins, not on rivals being locked out. Fires the step-4 falsifier on hbm-cowos-as-binding-bottleneck. From 2026-06-01-autoresearch-hbm-cowos-bottleneck-2026-corroboration.
  • Vera Rubin HBM4 allocation confirmed (June 5, 2026): Jensen Huang certified SK Hynix for Vera Rubin HBM4 on June 5. Supply-chain analyst estimates allocate SK Hynix ~60-70% of Vera Rubin HBM4 volume — up from prior 57% Counterpoint share estimate, suggesting SK Hynix's lead widened in the HBM4 generation. Vera Rubin entered full production June 1, 2026; first hyperscaler (AWS, Google Cloud, Microsoft Azure, Oracle) shipments target summer 2026. From 2026-06-08-autoresearch-hbm-supply-sk-hynix-samsung-micron-helium-2026.
  • $3.9B Indiana 2.5D packaging plant: SK Hynix is building a 2.5D advanced packaging facility at the Purdue University campus in West Lafayette, Indiana, targeting operations by 2028. Funded by $3.9B investment + $458M CHIPS Act grants + $500M loans. The plant will offer "turnkey" HBM delivery — memory stacks already integrated with silicon interposers — bypassing TSMC CoWoS entirely. CoWoS is sold out through 2027; this plant decouples SK Hynix from TSMC packaging constraints for the next generation. From 2026-06-08-autoresearch-hbm-supply-sk-hynix-samsung-micron-helium-2026.
  • Helium buffer: 8-16 weeks (second-most exposed among major HBM suppliers; Samsung is most exposed at 6-12 weeks, Micron least exposed at 12-24 weeks). SK Hynix sourced 64% of its 2025 helium from Qatar. Korea's overall Qatar helium dependency was 65% in 2025. Advanced AI chips and HBM production receive triage priority for available helium — the constraint tightens supply margins rather than triggering a production collapse. From 2026-06-08-autoresearch-hbm-supply-sk-hynix-samsung-micron-helium-2026.

Strengths (from a thesis-input perspective)

  • First-mover on HBM4 at scale; 3-year customer backlog provides revenue visibility
  • 76% Q2 / 72% Q1 operating margin means extraordinary pricing power and capital-generation capacity (KRW 88T cash as of Q2)
  • LTA deposits (existence confirmed; size not filed) improve demand visibility; not the same as a named hyperscaler fab-equity backstop
  • HBM4E next generation already sampling in H2 2026, maintaining generational lead
  • Preferred Nvidia partner for current HBM generations

Weaknesses (from a thesis-input perspective)

  • OTC US liquidity (HXSCL) is thin; limited US investor access vs. SK Hynix's Korea-listed shares
  • Korean corporate governance (chaebol structure) adds geopolitical and regulatory risk
  • If deposit-LTAs tighten allocation without customer-funded capacity, non-depositor buyers stay short; if fab-equity later lands, customers gain leverage in future pricing
  • HBM4E early sampling compresses HBM4 ASPs faster than expected
  • Samsung catching up in HBM4 yield would erode SK Hynix's dominant share

Sources

Related

Referenced by
Mechanisms
HBM pre-committed + CoWoS half-owned by Nvidia → DRAM spot price +158% → memory/packaging bottleneck downstream of foundry
Concepts
Contract structure, not time, is the seam in the memory marketHBM supply bottleneck
Questions
Hypothesis: Big Tech direct financing of SK Hynix expansion → new structural supply-demand mechanic in HBM
Entities
Micron Technology