The Roundhill Memory ETF (CBOE:DRAM) has become the retail trade of the year. Launched on April 2, 2026, DRAM was pitched as “the first-ever memory stock ETF”, and buyers have piled in fast enough that the fund is being described as the fastest asset gatherer in ETF history, reportedly crossing $25 billion in assets in roughly four months. The appeal is obvious: memory pricing has ripped higher alongside AI infrastructure spend, and DRAM lets a US brokerage account own Samsung and SK hynix without wrestling with foreign listings. The fund is up 107.78% since inception. At face value, it seems like a fine idea; however, on closer inspection, it appears to be a fund most memory bulls do not actually need.
What DRAM Actually Is Under the Hood #
Strip the branding away and DRAM is a three-stock portfolio with a tail. Samsung Electronics sits at 24.99%, SK hynix at 24.22%, and Micron Technology at 23.83%. Those three names are 73% of the fund. The remaining sleeve, spread across Kioxia, Sandisk, Western Digital, Seagate, Nanya and Winbond, adds storage and Taiwanese memory exposure but does not change the concentration story. Geographically, 49.25% of the fund sits in South Korea, 37.65% in the United States, 6.31% in Taiwan, and 4.87% in Japan.
For that exposure, holders pay 0.65% annually. That is not egregious for a thematic ETF, but it is meaningful when the same trade can be built directly. It also introduces a wrinkle US taxable holders should understand: Samsung and SK hynix pay foreign dividends, and the fund’s Korean exposure carries withholding tax and won complications that a domestic name does not.
A Simpler Trade Has Beaten the Fund #
If your thesis is “AI needs more memory,” the cleanest expression is the one US pure-play in the basket: Micron Technology (NASDAQ:MU | MU Price Prediction). Since DRAM’s launch on April 1, 2026, Micron is up 162.86%, versus 107.78% for the ETF that holds it as a top position. Year to date, Micron is up 238.94%, and over the trailing year it has returned 736.3%. That gap is the mechanism. DRAM’s Samsung weight dilutes the memory thesis with a conglomerate that also sells phones, appliances, and displays. SK hynix is a purer bet, but Micron has been the highest-beta pure-play in the group during this cycle, and the ETF’s rules-based weighting caps how much of that upside flows through. You are paying 0.65% a year to have the fund partially hedge the very trade you were trying to make.
If You Want Diversification, Own Semis Broadly #
The counter to single-stock risk is a broader semiconductor fund. The iShares Semiconductor ETF (NASDAQ:SOXX) is up 72.87% year to date, and the VanEck Semiconductor ETF (NASDAQ:SMH) is up 55.62%. Both trail Micron, but both include Micron alongside logic, foundry and equipment names that benefit from the same AI capex wave without depending on a single pricing cycle. For an investor whose real goal is AI-linked chip exposure rather than a memory-specific bet, SOXX at a lower expense ratio is the sturdier vehicle.
Tradeoffs Worth Naming #
Owning Micron alone means accepting the full swing of memory pricing. If DRAM contract prices roll over, MU falls harder than DRAM will, because the fund’s Samsung and Korean allocations blunt some of the move. That is the real cost of the purer expression. It is also worth noting DRAM is the only clean US-listed way to overweight SK hynix without buying ADRs, so investors who specifically want Korean memory exposure have a real reason to keep at least part of the position.
How to Think About the Swap #
In a taxable account, do not blow up a 100%-plus gain to save 65 basis points. The tax bill will dwarf the fee savings for years. In a retirement account, the math flips: rotating some or all of the DRAM position into Micron for higher beta, or into SOXX for broader semis, is a cleaner expression of most investors’ actual thesis. A middle path is trimming DRAM back to a position sized for Samsung and SK hynix access only, and holding Micron directly for the US memory book.
DRAM is a concentrated fund wearing a diversified label, at a price most of its owners can beat by naming the two or three things they actually wanted to own.
Contact [email protected] for any questions or corrections.