HBM4 Memory Capex
Mass ProductionNext memory supercycle — SK Hynix, Samsung, Micron racing
Production Ramp Now (H1 2026) → Mass Production H2 2026
Thesis
The HBM4 cycle is beginning with Samsung starting HBM4 production, SK Hynix completing 12-high hybrid bonding validation, and Micron ramping. All three memory makers are simultaneously increasing capex. This capex flows to equipment makers with monopoly positions: Towa (compression molding monopoly), Auros (HBM4 IR metrology and thin-film measurement), and memory ETF exposure via EWY and SNDK.
Picks in this Sector (4 active)
Towa holds a monopoly over compression molding — the critical packaging step for HBM memory. Every major memory maker (MU, SK Hynix, Samsung) uses Towa. With all three printing from HBM3e and now ramping HBM4, their capex cycles feed directly into Towa equipment orders. Flat YTD despite record memory capex — markets haven't connected the dots yet. Earnings May 11 as catalyst.
Auros is an undiscovered Korean company at ~$270M MC building IR metrology and thin-film thickness measurement for HBM4 hybrid bonding — a segment that KLA previously monopolized. Now qualifying at Samsung factories with SK Hynix likely qualifying too. Two products simultaneously volume ramping H2 2026. A decade of R&D finally paying off.
SanDisk (spun off from WD) is the pure-play NAND memory supercycle vehicle. Q3 2026: revenue $5.95B (+252% Y/Y), EPS $23.41 (+62% beat), 78.4% gross margin. Q4 guidance: $7.75-8.25B revenue (23% above consensus). Memory companies reprice ahead of earnings via third-party NAND price hike reports — the earnings just confirm. Up 300% YTD.
EWY is the cleanest vehicle for exposure to Samsung and SK Hynix through a US-listed ETF. When everyone was doomposting 'KOSPI crash bubble' and 'Silver Crash', Serenity went long and held conviction. Now up ~3× on 2028 LEAPS. Korean memory makers are printing from HBM3e and ramping HBM4.