EWY

Current Thesis

Last week's dossier called this SATURATED and said SKIP at [entry redacted] (5/23). The tape ripped +18% straight through that read to $214.53 (6/3), ~1.5% under the $217.76 52-wk high — a clean, documented alpha leak of exactly the kind this playbook exists to prevent ("extended = skip" on an ACCELERATING, cluster-confirmed name). I'm correcting the stance. The memory super-cycle is PROVEN and still TIGHTENING, not rolling over: TrendForce has DRAM contract prices +58–63% QoQ and NAND +70–75% QoQ in Q2 2026 (after ~95% in Q1), SK Hynix is sold out through 2026, and no new fab capacity lands in volume before late 2027/2028. Relative strength is the tell — the EWY/QQQ ratio is +71.8% YTD (5/27), the widest annual outperformance since the 2001 dot-com bust. Jensen Huang personally courting Samsung + SK Hynix at a Taipei dinner (6/2) is fresh narrative fuel, not exhaustion. Theme registry reads ACCELERATING (6/4). Per the momentum-realignment rule, a fresh entry here is a ride-the-leg, not a chase — but I size it MEDIUM, not max, because (a) EWY is a country-ETF wrapper that dilutes pure memory beta (~34% Samsung+Hynix, rest financials/autos/Value-Up), (b) the entry is at maximum extension (+109% YTD, +5%/day candles), and (c) the redistribution political tail is live.

Bull Case

• Memory cycle still ACCELERATING, not peaking: DRAM contract +58–63% QoQ, NAND +70–75% QoQ in Q2 2026 (TrendForce, 3/31); SK Hynix HBM/DRAM/NAND "essentially sold out" for 2026; no volume capacity relief until late 2027/2028. Fundamentals are tightening, not topping.
• Samsung Q1 2026 OP 57.2T won ($39.3B), +750% YoY, chips = 94% of profit (~4/29) — the HBM-qual binary we deferred for printed massively bullish; thesis confirmed, archetype flips 7→2.
• Relative-strength leadership: EWY/QQQ ratio +71.8% YTD, widest since 2001 (Benzinga 5/27). Korea is beating the Nasdaq 100 by the most in 25 years — institutional + foreign flow, not just retail froth.
• Fresh narrative re-fire: Jensen Huang hosts Korean tech-power dinner in Taipei, deepening Nvidia ties with Samsung + SK Hynix for "next AI wave" (Benzinga 6/2). News is NOT drying up — a structural-cracks trim signal is absent.
• Samsung Q2 guidance points to sequential improvement: HBM >3x YoY in 2026, HBM4 to exceed ~50% of HBM sales from Q3, HBM4E samples in Q2 — the mix shift to high-margin HBM is the next leg.
• Google to build its first non-US AI campus in South Korea (Nikkei 4/27) — durable AI-infra anchor beyond memory; the strike (45k workers, Fortune 5/17) is paradoxically ASP-bullish via tighter supply.

Bear Case

• Maximum extension: +109.36% YTD NAV (as of 5/29), trading at $214.53 vs $217.76 52-wk high. A fresh buyer here is long the parabolic leg with a wide stop and poor near-term risk/reward if the move mean-reverts.
• Korea political tail is live: an official floated redistributing Samsung/SK Hynix AI profits to citizens (5/12) — directly anti-shareholder, guts the Value-Up re-rate leg. It triggered the Roundhill DRAM ETF's worst day ever (-11.8%), SanDisk -11.1%, Micron -10.3%. A single Facebook post wiped >10% off the leveraged complex; legislation would do worse.
• Peak-retail saturation tells are textbook late-cycle: "$1000 invested 5 years ago" look-back (5/22) + "$100 invested 5 years ago" (5/08) + "Forget Nvidia, +70% memory ETF in a month" (5/06). When the look-back-glory pieces hit, the easy money is behind.
• Country-ETF wrapper dilutes the trade — you're paying 0.59% to own ~34% memory beta plus chaebol financials/autos that don't share the AI tailwind. A single-name Samsung or SK Hynix is a cleaner expression if you want the cycle.
• Samsung Q2 earnings 7/23 (preliminary guidance historically ~7/7–7/8) is a binary just beyond the window — DX-division cost pressure already flagged; any HBM mix disappointment unwinds the multiple.

Setup & Price Structure

Parabolic, leadership-grade, and at the highs. EWY $214.53 (6/3), NAV $216.67 (+1.06% on the day); +5.28% on 6/1 ($205.83 → $216.70). 52-wk range $63.49–$217.76 — i.e. the fund has more than tripled off its low and sits ~1.5% under the high. The launch base for the current leg is ~$205 (6/1 breakout); the prior shelf is ~$182 (5/23). RSI is unquestionably stretched (multi-week parabola), but for archetype 2 a high RSI is only a trim trigger above 88 AND with structural cracks — and the structure shows none: cluster intact (Samsung/Hynix/Taiwan ripping together), news fresh (Jensen 6/2), no peer divergence, no distribution. So strength = confirmation here, not a fade. Fresh entry: MEDIUM/probe at market is playbook-valid (do NOT defer an ACCELERATING cluster-confirmed name for "extension"). Better R:R upgrade to HIGH on any controlled pullback that holds ~$200 round number / reclaims the 6/1 base on volume. First warning sign: daily close below $200; thesis-break: weekly close below ~$190.

Catalyst Calendar (next 30 days)

• ~2026-07-01 (est.) — Korea June trade/export data. Memory exports have printed 12 consecutive 30%+ YoY months; this is the cleanest in-window read on whether the cycle is still accelerating. CATALYST_DATE.
• ~2026-06-30 (est.) — TrendForce month-end / Q2-final DRAM-NAND contract pricing updates. Confirmation (or first deceleration signal) for the +58–63% / +70–75% QoQ thesis.
• Rolling — Samsung union/strike resolution; talks resumed 5/19 (Labour Commission), one issue unbridged. A settlement is a sentiment clear; prolonged strike is ASP-bullish but headline-noisy.
• ~2026-07-07–08 (est.) — JUST OUTSIDE WINDOW — Samsung Q2 2026 preliminary earnings guidance (full report 7/23). The real binary; do NOT initiate fresh size into the prelim print.
• Rolling — Nvidia/Korea AI-partnership follow-through after the 6/2 Huang dinner; any supply-agreement or HBM4 qual headline is incremental fuel.

What Would Change Our Mind

• Invalidation (exit/avoid): Weekly close below ~$190 (the early-June breakout base, -11% from [entry redacted]) — the parabolic leg is broken; stand aside, re-enter only on a fresh higher-low setup.
• Fundamental break: TrendForce flags Q3 2026 DRAM contract prices decelerating to flat/negative QoQ — the super-cycle has peaked; the whole memory leg deflates.
• Political break: the Samsung/SK Hynix profit-redistribution proposal advances from a Facebook post to actual draft legislation — the idiosyncratic Korea-specific tail that kills the Value-Up re-rate and re-rates the ETF's chaebol weight lower.
• Upgrade to HIGH: a controlled pullback that holds ~$200 and reclaims the 6/1 base on volume, OR a clean Samsung Q2 prelim beat (~7/7–8) with HBM4 ahead of plan — that converts late-extension friction into a fresh higher-low entry.
• If held (not currently): for archetype 2, trim only on RSI>88 WITH structural cracks (peer underperformance, distribution flow, news drying up). None present as of 6/4 → default HOLD/ride. EXIT_FULL only on the weekly-$190 break or the redistribution-legislation event.

Correlation Notes

• ~34% of NAV is the AI-memory complex (Samsung ~23% + SK Hynix ~11%) — EWY trades as a memory proxy with a chaebol overlay. Highest correlation to SK Hynix, Micron (MU), SanDisk, and the Roundhill DRAM ETF (which fell -11.8% on the 5/12 political shock — EWY is the lower-beta wrapper of that same trade).
• Strongly correlated to the broader AI-chip-infra cluster: Taiwan (EWT/TSM), the HBM supply chain, and Nvidia demand signals (6/2 Huang dinner). A Nvidia/AI-capex air-pocket would hit EWY through the memory weight.
• USD/KRW is a live drag: KRW breaking >1,400/USD imposes a 5–8% USD translation drag on EWY vs local KOSPI — watch BoK policy and the dollar. Iran-deal/oil-down risk-on (5/25) is a mild tailwind for EM/Korea beta.
• Lower correlation to the non-semi sleeve (financials, autos, Value-Up names) — this is the dilution that caps conviction vs a pure single-name memory expression.