DRAM's 166% Run Is Rewriting Semiconductor ETF Flows
Roundhill's DRAM ETF has surged 166% since its April launch and now holds $25B in assets, reshaping capital flows away from SOXX and SMH.
Key Points
- Roundhill's DRAM ETF has surged 166% since its April launch and crossed $25 billion in assets in June, making it the most successful U.S. ETF debut of the first half of 2026.
- SOXX and SMH cannot hold SK Hynix or Samsung Electronics, giving DRAM a structural monopoly on pure-play global HBM exposure that legacy semiconductor funds structurally cannot replicate.
- Traders should watch SOXX's $5.4 billion single-day inflow event as a benchmark — any repeat of that magnitude would confirm institutional money is still rotating into the semiconductor complex rather than locking in gains after a 113% YTD run.
The most explosive ETF trade of 2026 is not in artificial intelligence broadly — it is in the three companies that manufacture the memory chips AI cannot run without. Roundhill's Memory ETF (DRAM) has gained 166% since its April launch, grown to over $25 billion in assets, and pulled in nearly $10 billion in June alone, outpacing every fixed income fund, every international equity product, and every thematic rival on the monthly flow leaderboard. SOXX, for all its headlines, is up 113% year-to-date. DRAM has lapped it in a fraction of the time.
The Structural Edge No One Else Has
The core reason institutional money has flooded into DRAM at a pace that defies normal ETF ramp timelines is a structural gap in the two dominant semiconductor funds. As of July 9, neither the VanEck Semiconductor ETF (SMH) nor the iShares Semiconductor ETF (SOXX) holds SK Hynix or Samsung Electronics. Both names are Korean-listed companies, and both legacy funds are built on U.S.-listed or U.S.-focused index methodologies that exclude them. That is not a minor omission. SK Hynix, Samsung, and Micron together control more than 90% of the global high-bandwidth memory market — the HBM chips that Nvidia, AMD, and every hyperscaler building out AI inference infrastructure must buy before they can ship a single accelerator.
DRAM holds all three. That gives it something SOXX and SMH functionally cannot offer: direct, concentrated, single-ticker exposure to the companies that are actually supply-constrained in AI hardware. When a portfolio manager decides they want to own the HBM bottleneck rather than a broad semiconductor basket weighted toward equipment makers and designers, DRAM is the only liquid, regulated vehicle that delivers it. That is why the fund went from zero to $25 billion in roughly 75 days — a ramp that by any prior metric would be considered impossible for a thematic niche product.
Where the $10 Billion in June Actually Came From
June's near-$10 billion inflow into DRAM did not arrive in a smooth daily trickle. The semiconductor complex saw some of the most aggressive single-session positioning of the year, including a $5.4 billion single-day inflow into SOXX that ranked as one of the largest single-day ETF positioning events on record for a sector fund. That kind of flow does not come from retail investors clicking "buy" on a brokerage app — it comes from institutions executing large block creations, likely model portfolio rebalances, derivative hedging programs, or tactical allocations tied to quarterly earnings cycles.
SOXX itself is up 113% year-to-date, with the bulk of that gain compressed into the second quarter, meaning the managers adding $5.4 billion in a single session were not buying at the bottom — they were chasing momentum at elevated valuations. That context matters. ETF inflows topped $1 trillion year-to-date through June, and the semiconductor slice of that total is disproportionately large relative to the sector's weight in the broad market. The $4.1 billion that went into SOXX in June — separate from that single-day spike — combined with DRAM's $10 billion haul means roughly $14 billion in net new semiconductor ETF exposure was created in one calendar month. That is a crowding risk as much as it is a momentum signal.
What Traders Should Watch Next
The forward-looking question is whether DRAM's flow momentum survives its own success. At $25 billion in assets and a 166% return in less than a quarter, the fund has moved from discovery trade to consensus trade — the precise moment when the easy money typically slows. The three names that constitute over 90% of the portfolio are not immune to the broader macro setup: the 10-year Treasury yield sits at 4.54%, CPI is still running at 4.2% year-over-year, and the Fed funds effective rate is 3.62%, meaning the rate environment is not the tailwind for high-multiple tech hardware names that it was in 2020 and 2021.
SK Hynix and Samsung both report earnings this month. Any guidance cut on HBM pricing or volume — even a modest one — would land directly in DRAM's NAV with no diversification cushion. The fund's concentration is its structural advantage and its principal risk simultaneously. SOXX, by contrast, spreads across equipment, design, and fabrication, which dampens the upside but also the downside on any single segment shock. Traders holding DRAM through earnings should define that risk explicitly: a 10% drawdown in SK Hynix alone translates directly into roughly 10 points of NAV damage in a fund with that level of top-line concentration.
The specific level to watch on SOXX is the 113% YTD gain line — meaning any pullback that takes the fund below its Q1 high would signal that the institutional momentum buyers who showed up with $5.4 billion in a single session are reversing, not adding. For DRAM, the date is SK Hynix's earnings release. If HBM demand guidance holds or accelerates, the $25 billion fund has a credible path toward $35 billion before year-end. If it doesn't, the unwinding of a crowded, concentrated, 166%-gainer will not be orderly.
The Weekly Investor
Daily market analysis for active traders. Free.
