Korea +16% in a month: reading the breakout in the AI value chain
The headline number that structures this entry: EWY is up +15.94% over 30 days while ASML's proxy ETF (EWN) is -0.53% and Japan equipment (EWJ) is flat at +1.07%. That is a very unusual distribution across the AI supply chain. Foundry and memory are running hard; equipment and design support are not. Understanding that divergence is the job this week.
What I'm watching
- EWY (Korea): $174.02, 1m +15.94%, 1y +202.65% -- the strongest move in
the panel by a large margin; 52w-pos at 86.1% confirms this is not a dead-cat bounce.
- EWT (Taiwan): $89.83, 1m +7.47%, 1y +73.96% -- continued strength, now
at 84.4% of 52-week range. The year-over-year number is extraordinary.
- EWN (Netherlands): $63.77, 1m -0.53% -- ASML accounts for most of EWN's
equity beta. Flat when the rest of the chain is running is a meaningful signal.
- Copper: 6.1875 USD/lb, 30d +7.65% -- AI data center buildout is a
copper-intensive proposition (power, cooling, interconnect). Continued copper strength confirms the physical AI infrastructure capex cycle is still live.
What changed / what matters
Korea's move: three candidate explanations
An ETF that goes up +16% in a month when broad EM is mixed needs a reason. Three candidates, roughly in order of explanatory weight.
First, Samsung HBM3e qualification progress. The April 2026 entry on this theme flagged Samsung's failure to qualify HBM3e at Nvidia as the key binary for EWY -- if Samsung qualifies, Hynix margins compress; if not, the supply bottleneck persists. Recent market behaviour looks more consistent with a qualification milestone than a miss. If Samsung is now shipping qualified HBM3e at meaningful volumes, that expands the overall HBM-capable supply base and is a positive for the whole ecosystem, including Samsung itself. EWY is the beneficiary of both Hynix (oligopoly pricing) and Samsung (volume recovery).
Second, US-Korea strategic trade alignment reducing risk premium. The May 2026 KUSPI MOU (IPTM 2026-05-08, Korea-US Shipbuilding Partnership Initiative) is the visible piece of a broader US-ROK strategic trade and investment framework worth roughly $350 bn. The market may be pricing a reduction in tariff and export-control risk for Korean tech exports to the US. BIS's advanced IC due-diligence framework (IFR, January 2025) exempts qualified foundry-aligned producers; Korea's policy alignment reduces the probability of punitive export controls hitting DRAM or NAND exports.
Third, valuation catch-up vs Taiwan. Taiwan ran first and faster; Korea's 1y number was already +202% as of today, but much of that was from a lower base. The relative-value argument -- Korea memory is cheaper than Taiwan foundry on most earnings multiples -- may be attracting systematic rotation.
Why is ASML (EWN) flat?
Equipment cycles lag foundry cycles by 12-18 months. ASML books orders when fabs commit to new capacity; revenue and earnings follow when the tools ship. The current EWT and EWY strength reflects utilisation of already-installed capacity, not new node expansions. New high-NA EUV orders -- the equipment event that would re-rate EWN -- would need to coincide with TSMC's next-node capacity commitment, which is not priced as an imminent announcement. EWN at -0.53% 1m is not weakness; it is the market correctly distinguishing between current-capacity yield and forward-capacity investment.
India tech: wrong exposure
INDA -6.45% 1m is not a chip story. India's ETF beta to the AI value chain is almost entirely software services (Infosys, TCS, Wipro, HCL), not semiconductor fab or memory. Software services is running into margin compression from AI automation displacing entry-level coding and BPO tasks -- the opposite of the hardware story. India's weakness here is structurally coherent, not a buying signal for the tech-ai theme.
Regime context
DXY at 99.4 (flat) and VIX at 18.12 (cautious) means the USD headwind is neutral and volatility is elevated enough to justify some discount on high-beta positions. A regime shift to "risk-on" would require VIX below 15 sustained. At 18 we are in the zone where tech-AI plays are viable but position sizing should stay measured.
Candidate picks within this theme
- EWY (Korea) -- HBM oligopoly plus Samsung recovery optionality; the 1m move
has momentum but the 1y run is not obviously overdone given earnings delivery.
- EWT (Taiwan) -- TSMC node leadership is the irreplaceable asset in AI compute;
geopolitical tail risk is real but the 1y re-rating reflects genuine earnings power.
- EWJ (Japan) -- equipment diversification (TEL, Disco, Advantest), low beta,
a quieter way to stay in the chain without the single-name concentration of EWN.
- EWN (Netherlands) -- flat now, but high-NA EUV is the next step-change; a
long-term position for anyone with a 3-5 year horizon on the next node transition.
- INDA -- omit for tech-ai theme; wrong exposure, wrong momentum.
What I'd revise if I saw
- A confirmed Samsung HBM3e volume-ramp announcement would change the EWY
thesis from "oligopoly pricing" to "competitive normalisation" -- reduces conviction, does not eliminate it.
- ASML booking guidance cut at next earnings -- signals hyperscaler capex
pulling back more than expected, which would reprice EWT and EWY together.
- VIX sustained above 22 would shift the regime to risk-off and compress
the case for high-beta EM tech across the board.
Cross-references
- Previous entry on this theme:
docs/thinking/2026-04-22-tech-ai-value-chain.md - Korea strategic context:
docs/iptm/actions/2026-05-08-us-korea-shipbuilding-partnership-initiative-kuspi.md - BIS IC due-diligence framework:
docs/iptm/actions/2025-01-16-us-bis-advanced-ic-due-diligence-ifr.md - Relevant macro route:
/api/country/KR,/api/country/TW,/api/country/JP