$DRAM: Roundhill Memory ETF

AI infrastructure breaks without HBM, DRAM, NAND, HDD capacity, and enterprise storage.

DRAM targets the memory and storage layer behind AI data centers, with concentrated exposure to global leaders in HBM, NAND, DRAM, and data center storage.

DRAM gathered $6.5B in its first 36 trading days and now sits near $10B in AUM, making it one of the fastest-growing ETF launches ever.

$000660.KS — SK hynix | 28.2%

SK hynix is the HBM kingpin. The company controls roughly 70% of the HBM market and supplies Nvidia’s Blackwell and Vera Rubin systems. Capacity is effectively booked for the next 18 months.

$MU — Micron | 24.9%

Micron is the U.S. HBM and DRAM challenger. Management has described current demand as the largest memory supply gap in semiconductor history, with HBM ramps aimed at enterprise AI data centers.

$005930.KS — Samsung | 20.9%

Samsung is the scale player. The company is pushing into sixth-generation HBM4 mass production for future AI accelerators while using its huge memory footprint to challenge SK hynix and Micron.

$285A.T — Kioxia | 6.5%

Kioxia gives $DRAM NAND exposure. The company focuses on enterprise flash storage for hyperscalers, where AI workloads create new pressure on server-grade SSD capacity and storage density.

$SNDK — SanDisk | 5.1%

SanDisk adds advanced NAND and storage capacity exposure. The company joined a $2.5B private placement tied to factory upgrades and signed multi-year supply agreements to secure memory availability.

$STX — Seagate | 4.4%

Seagate is the AI data lake play. Its heat-assisted magnetic recording roadmap targets high-capacity cloud storage, including 100TB HDDs built for dense, power-aware data center workloads.

$WDC — Western Digital | 4.0%

Western Digital supplies enterprise HDD infrastructure. Hyperscaler demand has reportedly absorbed HDD capacity through 2026, with long-term contracts extending visibility into future AI storage needs.

$2408.tw — Nanya | 3.3%

Nanya adds edge memory exposure. The company is tied to SRAM and specialty memory for localized computing, where smaller AI models increasingly run closer to devices instead of central clouds.

DRAM concentrates exposure in the companies supplying the memory and storage layer under the AI stack.