LONGActiveMed1Y$000660.KS1,647,000.00 KRWFri 4 Sep

SK Hynix: the largest HBM rent, valued like a Korean cyclical

By MenFem Editorial·Semiconductors·9 June 2026·Methodology·
ai-infrastructuresemiconductorsmemory
Share
SK Hynix: the largest HBM rent, valued like a Korean cyclical

Key Points

  • ~62% HBM share and ~two-thirds of NVIDIA's HBM4 — the largest rent-capture in the cluster.
  • 72% operating margin and ~61% ROE, yet a single-digit forward multiple — rent in the earnings, not the multiple.
  • A confidential ~$14B Nasdaq listing is management's own explicit re-rating catalyst.
  • Memory bandwidth scales ~1.6x per two years against compute’s ~3x — every accelerator generation needs proportionally more HBM, so the bottleneck widens rather than clears.
Price at Call
$2215000.00

SK Hynix captures the single largest slice of the AI-memory scarcity rent — roughly 62% of high-bandwidth memory, about two-thirds of NVIDIA's HBM4, sold out and pre-paid into 2027 — and earns it at a 72% operating margin and a ~61% return on equity. And yet it trades like a Korean memory-cyclical at a single-digit forward multiple. The rent shows up in the earnings, not the multiple. Management is so convinced of the mispricing that it is taking the company to Nasdaq specifically to force a re-rating against Micron — a rare case where the best fundamentals in the cluster come with the most explicit, dated catalyst. BULLISH, conviction MEDIUM: the lead is yield-and-trust rather than a monopoly — Samsung out-qualified it on HBM4 timing yet won only a quarter of the allocation — and two-thirds of HBM riding one customer is the risk that caps conviction. The structural driver underneath the share numbers is a physics gap: compute scales roughly 3x every two years while memory bandwidth scales only about 1.6x, so each accelerator generation needs proportionally more HBM than the last — 80GB on H100, 192GB on Blackwell, more again on Rubin. That is what makes this a rent rather than a cycle. The bottleneck widens with every generation instead of clearing. (Merged in from the retired "HBM Suppliers Are the Picks-and-Shovels of AI" call, 2026-08-03.)

Research Log

SKHY tape reference, researched from primary sources (catalogue II item 71). The ADS exists: SK hynix listed 177,900,000 ADS on Nasdaq as SKHY on 2026-07-10 at $149.00/ADS (~$26.5B raised); each ADS represents one-tenth of one ordinary share (10 ADS : 1 ordinary share) — SEC EDGAR Form 424B4, https://www.sec.gov/Archives/edgar/data/0002120882/000119312526299963/d32785d424b4.htm. Tape print: SKHY closed at $177.00 on 2026-09-04 (Yahoo Finance / stockanalysis.com historical data). At that ratio, $177.00/ADS implies $1770.00 per ordinary share. Converting at 1359.30 KRW/USD — the Seoul market close for 2026-09-04 per KB Kookmin Bank's daily FX-desk report (https://kbthink.com/investment/fx/daily/260904.html); the Fed's own H.10 release for that week had not published as of 2026-09-07 and BOK's basic-rate page was unreachable this pass, so this is a near-primary bank market-data source, not BOK/Fed itself — that implies ≈2,405,961 KRW per ordinary share on the SKHY tape, against the 1,647,000 KRW close already recorded for 000660.KS on the same date: a gap of +46.1%, i.e. the USD-tape SKHY print implies a materially richer ordinary-share value than the KRX print the same day (likely thin-liquidity/arbitrage dislocation in a newly-listed ADS rather than a reconciled cross-rate). The rule stands regardless: the call was made on 000660.KS at 2,215,000 KRW (the 2026-06-09 publish-date close) and that entry price does not change; SKHY is a tape reference only, never a substitute entry or exit price for this call.

source: docs/plans/markets-work-catalogue-2026-09-07.md

AMENDED 2026-09-06 (Connor's ruling on the review). The catalyst this call named — the US listing — completed on 10 Jul 2026: ~177.9m ADS at $149, ≈$26.5B, Nasdaq ticker SKHY. The falsifier's second leg ("listing shelved or withdrawn") is therefore permanently unreachable; the KRW line is −25.64% from the 2,215,000 entry (1,647,000 on 4 Sep), worst weekly close −35.8%. The row's keyPoints still describe "a confidential ~$14B" listing — that is stale and is corrected here rather than rewritten silently in the thesis body: the listing is done, it was ~$26.5B, and the re-rating the thesis expected has not happened. The rent thesis itself is NOT FIRED; the catalyst is spent, so the call now rests on the HBM4 share and the multiple alone. A USD instrument (SKHY) now exists; the weekly-close spine quotes 000660.KS in KRW, % only. KEEP, conviction MEDIUM, next print derived 27 Oct.

source: docs/plans/call-reviews-2026-09-06-tickered.md

Sources filled from the shelf and the 2026-09-06 review docs (catalogue item 3). 3 entries.

source: docs/plans/markets-work-catalogue-2026-09-06.md

Entry price filled: 2,215,000 KRW, the 000660.KS close on the publish date 2026-06-09 (Yahoo daily chart, read 2026-09-05). Was null because the retired Finnhub cron could not quote KRX. Connor's ruling 2026-09-06.

source: docs/plans/market-call-targets-2026-09-05.md

Q2 2026 printed (29 Jul) — records across the board, stock −9.6% on the day.

Revenue ₩79.32T. Operating profit ₩60.54T, +557% YoY and +61% QoQ. Gross margin 83%, operating margin 76.3%. Cash and equivalents ₩88T, up ₩33.6T on the quarter. HBM4 entered mass production during Q2 with a significant ramp guided for H2.

The shares fell anyway, because the print missed expectations (the street looked for roughly ₩64T operating profit on ₩84T revenue) rather than missing the thesis. Analysts pinned the shortfall on HBM4 shipments landing light, pushing revenue recognition into the second half.

This call argued the largest HBM rent in the world was being valued like a Korean cyclical. Q2 says the rent is real and still compounding at a 76% operating margin. What has changed is the nature of the disagreement: the market has stopped arguing about whether the rent exists and started arguing about how long it lasts. That is a duration argument, and duration is what the next two prints will settle.

Bull Case

Captures the most HBM scarcity rent of anyone — 62% share, two-thirds of NVIDIA's HBM4, 72% operating margin — at a single-digit forward multiple, with management taking it to Nasdaq to force the re-rating.

Bear Case

A leveraged single-customer (NVIDIA) HBM bet at peak margins, where Samsung has already out-qualified on HBM4 and the chaebol structure discounts the shares.

What would prove this wrong

The reason for holding this stops being true if the rent is competed away — Samsung's yields converging and NVIDIA rebalancing HBM4 allocation, which turns a rent into a share fight and makes the low multiple correct rather than mistaken. It also stops being true if the Nasdaq listing is shelved or withdrawn: the argument here is that management will force the re-rating itself, and without that catalyst this is a cheap cyclical that stays cheap — which is precisely what the market already believes.

Catalysts

Nasdaq US listingRegulatory

~$14B raise explicitly aimed at re-rating SK Hynix against Micron; the cleanest dated catalyst in the cluster.

HBM4 16-Hi volumeProduct Launch

Allocation share in the HBM4 generation as Samsung pushes for parity.

Risk factors

Samsung yield convergenceHigh

Samsung out-qualified on HBM4 timing; if yields converge, NVIDIA rebalances allocation and the ~2/3 share erodes.

Single-customer concentrationMedium

Two-thirds of HBM rides NVIDIA's roadmap — a Broadcom wobble alone took the stock down ~10% in a session.

Related Calls

LONG$TSMActive

TSMC: own the bottleneck, not the brand

Own the bottleneck, not the brand: the chokepoint every AI chip depends on, at a discount to all of them.

Fabricates and CoWoS-packages every leading-edge AI accelerator — NVIDIA, AMD, Broadcom, Apple alike.

~22x forward earnings: a discount to NVIDIA, AMD and Broadcom, the names that cannot ship without it.

ai-infrastructuresemiconductorsfoundry
TSMC
High3YSemiconductorsUpdated 4d ago9 Jun
WATCH$AVGOActive

Broadcom: the right custom-silicon thesis at the wrong price

The right custom-silicon thesis at the wrong price — watching the most expensive, lowest-quality way to play it.

The named winner of 'custom silicon eats commodity inference' — AI revenue +143%, FY27 AI guided >$100B.

Trades at a premium to NVIDIA on every earnings multiple while earning a third of the ROE.

ai-infrastructuresemiconductorscompute
Broadcom
Med1YSemiconductorsUpdated 4d ago9 Jun
LONG$MUActive

Micron: priced like a cyclical at the top, contracted like a utility

Priced like a peak-cycle memory-cyclical, contracted like a utility — a re-rating bet, not an earnings bet.

Entire CY2026 HBM output sold under multi-year fixed price-and-volume contracts.

Overtook Samsung into the #2 HBM slot; power-efficiency edge keeps it in every NVIDIA generation.

ai-infrastructuresemiconductorsmemory
Micron
Med1YSemiconductorsUpdated 4d ago9 Jun

Conviction

Conviction History

MedMed

3 Aug 2026

Reviewed against the Q2 2026 earnings wave (SK Hynix 29 Jul, Samsung 30 Jul, and the late-July hyperscaler prints). Records on every line at a 76.3% operating margin; the miss was against expectations, not the thesis. The argument moved from whether the rent exists to how long it lasts. Held — no conviction change.

Key Metrics

SK Hynix HBM Market Share
62%
Operating Margin (Q2 2026)
76.3%
Memory Bandwidth Gap
1.6x vs 3x/2yr
HBM4 Bandwidth
2TB/s
Blackwell HBM per Chip
192GB

Weekly closes

entry 2215000.00
See all Semiconductors calls →