Joe
Full-Time Trader/Investor, Semiconductor & AI Infrastructure Inflection Points
Anthropic's reported move to Samsung's 2nm process for its ASIC adds another reason to own the DRAM memory-sector ETF, which has 25% exposure to Samsung. Per @pequityresearch, Anthropic plans to adopt Samsung's 2nm process for its ASIC project to cut development cost and avoid competing with TSMC for capacity. x.comÂ
The memory-sector ETF is preferable to Micron because SK hynix and Samsung are cheaper and Samsung remains the dominant DRAM player, with its foundry business coming along for free. SK hynix's new US listing could also outperform Micron from its starting price, and a Samsung ADR likely follows in the medium term if it goes well. The three have similar market caps but not similar capacity and scale. x.comÂ
Samsung Electronics has no way out of the DRAM shortage for years, which means further memory price increases. Its foundry business adds long-term optionality given a shortage of logic capacity. Nikkei reports a surge in inquiries for Samsung's advanced chip manufacturing from existing and potential customers including BYD, Google, AMD and Tesla. x.comÂ
Samsung has the most upside among the three memory makers because it holds memory, foundry and packaging under one roof, letting it shorten development cycles and give customers more options. Its HBM4 collaboration on AMD's MI455X with 432GB of HBM4, plus foundry talks, shows a major AI customer taking it seriously again. Expect strong DRAM, NAND and HBM4 contributions. @JonahLupton notes it trades at 4.4x NTM EPS, and HBM prices rising at least 100% in 2027 may not be in estimates. x.comÂ
SK hynix's new direct US listing is a small disadvantage for the Roundhill Memory ETF, which investors had bought partly to own SK hynix indirectly. Owning memory companies directly is preferable to the ETF. x.comÂ
More memory CapEx may yield proportionally fewer bits as HBM consumes more wafer capacity, supporting SK hynix and Samsung's pricing. Both reported double-digit quarterly DRAM and NAND price gains; Samsung targets 60-70% of mid- to long-term capacity under LTAs. SK hynix has the stronger execution record, but Samsung could benefit enormously from conventional memory without winning HBM4. If trends persist, normalized margins could rise, though memory remains cyclical. x.comÂ
Prepaid contracts and demand-paced fabs weaken a broad 2028 memory-glut case. SK hynix says over half its 2027-28 capacity is signed with deposits; 'sold out through 2029' includes negotiations. Samsung targets 60-70% contracted capacity. Micron's three projects begin output in 2H28; tooling follows demand. Reuters says CXMT's 20-30%-yield HBM3E may ship this year to export-restricted Chinese buyers, but performs like older HBM3. Commodity DRAM risks 2028-29 oversupply; NAND could ease in 2H27. … Show more
The memory ETF is not worth buying yet, because it is essentially just the KOSPI plus Micron and more forced selling is likely in Korea. x.comÂ
Higher HBM stack requirements favor Samsung Electronics shares, even as the broader semiconductor industry suffers and DRAM is in dire straits. Daishin Securities says plans for 4-high and 6-high HBM in low-end inference appear shelved, while Nvidia's 8-high Rubin option appears dropped amid base-die supply constraints and customers' preference for bandwidth. x.comÂ
Memory should outperform over the next 6-12 months because demand is not slowing and supply is not catching up. The DRAM ETF is the easiest way to gain exposure, with Micron, Samsung and SK hynix as its top three positions. x.comÂ
The Roundhill Memory ETF is hard to justify because 75% of its holdings sit in just three stocks, so investors pay a fee for little diversification. x.comÂ
The Roundhill Memory ETF is a genuinely great way to get memory exposure, especially without access to foreign stocks. It concentrates on the three leaders, Micron at 24.6%, Samsung at 24.1% and SK Hynix at 23.1%, plus SanDisk, Kioxia, Western Digital, Nanya and Winbond. x.comÂ
