🧠

SKHY

💡
Thesis. The HBM leader posting a 76% operating margin on KRW 79 trillion of revenue up 257%, with ~10 long-term customer agreements and NVIDIA confirming it cannot avoid the price — while simultaneously building the $51B NAND fab that is the whole sector's 2027 supply risk.
Reviewed 19 Sep 2026, after Q2 2026 results on 28 Jul. Next review after Q3 2026 earnings on 28 Oct 2026.

What it does

SK hynix is the world's HBM leader — high bandwidth memory, the stacked DRAM beside every AI accelerator — plus conventional DRAM and NAND through Solidigm. It listed a Nasdaq ADR on 10 July 2026, the largest ADR listing in US history at roughly $28–29B. The Korean listing (000660) remains primary, so the ADR is a secondary claim on the same company with its own liquidity and premium characteristics.
Like its peers, it has converted part of a commodity business into contracted revenue: long-term agreements with about ten customers, the SK hynix equivalent of Sandisk's NBMs and Micron's SCAs. Coverage percentages are not disclosed.
This company is both the best expression of the memory shortage and the single largest reason it will eventually end. Both facts belong in the same position.

Bull case

  • Q2 2026 revenue KRW 79.3 trillion, +51% QoQ and +257% YoY. Operating income KRW 60.5T, +557% YoY, at a 76% operating margin — an all-time high. EBITDA KRW 64.6T at an 81% margin
  • First-half revenue crossed KRW 100 trillion for the first time in company history
  • The strongest balance sheet in memory: cash KRW 88T, net cash KRW 69.4T (+KRW 33.6T QoQ), debt-to-equity 7%
  • HBM4 mass production shipments began in the quarter with a broader ramp in H2; HBM4E samples shipped to major customers. 2026 HBM capacity effectively sold out
  • It retains the leading HBM position with NVIDIA — the gap
    🧠
    MU    's own guidance acknowledges it is trying to close. SK Group and NVIDIA expanded their strategic partnership on 27 July
  • Enterprise SSD sales doubled sequentially; Solidigm high-capacity enterprise SSD revenue rose more than threefold. Both DRAM and NAND prices rose significantly
  • Management describes a "structural shift in demand" as AI moves from training to inference, with AI memory and conventional memory growing at the same time
  • SK Group Chairman Chey Tae-won made his first-ever personal purchase of SK hynix shares in early August. UBS initiated the ADR at Buy with a $204 target

Bear case

  • It is building the supply that ends the cycle. The ~$51B NAND fab in South Korea, due by 2029, is the single largest capacity addition in the industry — and it competes directly with
    🧠
    SNDK
  • Shares fell on all-time-record results — the same pattern as Sandisk and Micron. The market is looking past the peak, and has been for several quarters
  • ADR liquidity is thinner than the Korean listing, and the premium or discount to the primary line shifts
  • KOSPI correlation is real: on 3 August the index fell 5% and Samsung 8% in a session, dragging the name with it regardless of its own fundamentals
  • Samsung is a formidable HBM competitor with deeper pockets, and a three-player race is decided by share rather than market growth
  • Long-term agreement coverage is not quantified, so the contracted floor is asserted rather than sized

Major customers

  • 🏛️
    NVDA     — HBM3E and HBM4 for Blackwell and Vera Rubin. The leading supplier position, with an expanded strategic partnership announced 27 July
  • Hyperscalers — server DRAM, AI server DRAM and enterprise SSDs via Solidigm
  • ~10 long-term agreement counterparties — identities not disclosed
  • Smartphone and PC OEMs — conventional DRAM and NAND
🔗
Read-through. SK hynix,
🧠
MU and
🧠
SNDK are the three memory names and they move together on one variable — memory pricing — which
🏛️
NVDA has confirmed is still rising into 2027 by guiding its own gross margin down on it. Holding all three is one position taken three times. Within the group, SK hynix is uniquely two-sided: the HBM leader (bullish for itself) and the builder of the largest NAND capacity addition (bearish for Sandisk), and its Solidigm SSD business competes with Sandisk's datacentre drives directly. A 76% operating margin on a commodity is the same bottleneck signature as Sandisk's 84.6% gross margin — and
💾
P is on the paying end of it.

What would change the view

  1. HBM4 ramp through H2 2026 and HBM4E timing — the lead over Micron, and the thing that justifies the premium
  2. Q3 results in late October, and specifically whether operating margin holds above 70%
  3. The $51B NAND fab timeline — the supply wave, and the clearest dateable risk in the memory group
  4. Long-term agreement count and, if ever disclosed, coverage
  5. The US fab decision
  6. ADR volume and premium or discount to the Korean line

Update log

19 Sep 2026 — Converted to the standard template. Current through Q2 2026 and the post-print developments; no new disclosures since 4 September.
28 Jul 2026 — Q2 2026 reported. Revenue KRW 79.3T (+51% QoQ, +257% YoY); operating income KRW 60.5T (+557% YoY) at a 76% margin; EBITDA KRW 64.6T at 81%; net profit KRW 93.9T including non-operating gains. Cash KRW 88T, net cash KRW 69.4T, debt-to-equity 7%. HBM4 mass production shipments began; HBM4E samples shipped. Long-term agreements with about ten customers. Enterprise SSD sales doubled sequentially. Shares fell on the record.
27 Jul 2026 — SK Group and NVIDIA expanded their strategic partnership across AI factories and next-generation memory.
20 Jul 2026 — Exploring US fab options.
10 Jul 2026 — Nasdaq ADR listed, the largest ADR listing in US history at ~$28–29B.
Research and education only — not investment advice.