The Roundhill Memory ETF, trading under the ticker DRAM, has crossed $20 billion in assets under management, marking a historic milestone for the fund that launched just 11 weeks earlier on April 2, 2026. The surge reflects investor appetite for dram stock exposure amid an unprecedented global memory shortage driven by artificial intelligence infrastructure buildout.
DRAM became the fastest-growing exchange-traded fund by asset accumulation pace in history, reaching $9.8 billion in just 43 days according to CNBC reporting in May. The ETF’s trajectory has accelerated since then, doubling its holdings in the span of a month as memory stocks rallied on supply constraints and surging demand from AI data centers.
The fund’s concentrated portfolio amplifies its exposure to the memory bottleneck. Samsung Electronics, SK Hynix, and Micron Technology account for roughly 72% to 73% of the ETF’s holdings, each representing about 24% to 25% of assets. These three companies dominate global production of DRAM and high-bandwidth memory (HBM) chips essential to AI systems. SK Hynix leads in HBM supply to AI accelerators, while Micron and Samsung compete in both traditional DRAM and premium memory segments.
The memory shortage driving the rally reflects a structural imbalance in semiconductor supply. Analysts estimate that AI data centers will consume approximately 70% of high-end memory chip production in 2026, a dramatic shift from historical cycles where consumer and enterprise computing shared demand more evenly. Industry sources including Everstream Analytics and Avnet reported in early 2026 that global semiconductor revenue was projected to increase 62.7% year-over-year in 2026, with DRAM and HBM experiencing the strongest growth due to supply constraints.
Memory has emerged as the critical constraint in AI infrastructure expansion. CNBC described memory as the “biggest bottleneck in the AI buildup,” highlighting that while compute power from GPUs captures investor attention, the data throughput required to feed those processors depends entirely on memory bandwidth and capacity. As AI models scale from chatbots to autonomous agents, each iteration demands more memory per system, intensifying competition for limited wafer capacity from the three dominant manufacturers.
The ETF’s rapid growth mirrors broader investor recognition of this supply-demand imbalance. Retail and institutional investors have piled into memory stocks seeking exposure to what some market participants view as one of the last undervalued corners of the AI trade. Micron, Samsung, and SK Hynix trade at significantly lower multiples than compute-focused AI stocks despite benefiting from the same infrastructure buildout, according to analyst commentary tracked in May and June 2026.
Roundhill Investments designed the actively managed fund with a 0.65% expense ratio to give U.S. investors direct access to global memory companies. Two of the fund’s three largest holdings, SK Hynix and Samsung, trade primarily on foreign exchanges, making the ETF structure a practical vehicle for retail exposure to the oligopoly controlling memory supply.
Sources
- Crypto Briefing — Reported DRAM ETF crossing $20 billion in AUM as of June 18, 2026
- CNBC — Described memory as the “biggest bottleneck in the AI buildup” and reported DRAM reaching $9.8 billion in 43 days (May 15, 2026)
- Roundhill Investments — Confirmed AUM of $21.85 billion, launch date April 2, 2026, and 0.65% expense ratio via official fund details
- 24/7 Wall St. — Detailed portfolio composition showing Samsung 25%, SK Hynix 24%, and Micron 24% weightings (May 15, 2026)
- Everstream Analytics — Reported 70% of memory chip production destined for AI data centers in 2026 (January 16, 2026)
- Avnet — Estimated AI data centers could consume 70% of high-end DRAM in 2026
- Stocktwits — Reported global semiconductor revenue projected to increase 62.7% year-over-year in 2026, with strongest growth in DRAM and NAND
- Investopedia — Documented memory shortage as critical bottleneck in AI data center buildout (May 27, 2026)
