DRAM ETF Hits $1B in 14 Days on AI Memory Chip Demand

The Roundhill Memory ETF (DRAM) crossed $1 billion in assets under management on 18 April 2026, just 14 trading days after its 2 April debut. The milestone ranks among the fastest ETF launches in history, driven by investor demand for direct exposure to memory semiconductors at the centre of AI infrastructure buildouts. Daily net creations exceeded $100 million in sessions following the threshold, signalling sustained institutional appetite rather than speculative momentum.

The fund’s rapid accumulation reflects a confluence of factors: AI-driven memory demand that has pushed DRAM prices up 148% year over year, supply constraints that have turned high-bandwidth memory (HBM) into what analysts describe as the “true bottleneck” in data centre expansions, and the fact that U.S. investors previously had no direct way to access Samsung Electronics and SK Hynix, the two South Korean manufacturers that control over 70% of the HBM market. DRAM now offers a pure-play vehicle where none existed. This article explains the fund’s trading characteristics, the supply-demand imbalance driving memory prices sharply higher, how DRAM compares to the broader iShares MSCI South Korea ETF (EWY), and what the early momentum signals about the sector’s transition from cyclical commodity to secular AI growth story.

Record inflows position DRAM among fastest ETF launches ever

The $1 billion mark arrived on 18 April 2026, 14 trading days after launch. By 21 April, assets under management had reached approximately $1.2 billion. Daily net creations continued at a pace exceeding $100 million in multiple sessions, indicating the initial surge was not a one-day headline event but sustained investor positioning.

DRAM’s rapid accumulation coincides with broader international stock rotation dynamics, where U.S. investors have shifted billions into overseas markets after domestic tech leadership stalled; Korean semiconductor exposure sits at the intersection of this geographic reallocation and AI infrastructure thematic demand.

The fund’s trading metrics show institutional rather than retail-driven adoption. Average daily volume sits at $213 million, with options trading exceeding 11,000 contracts per day. Bid-ask spreads average 0.05%, tight enough to allow efficient entry and exit for large positions. The expense ratio of 0.39% positions DRAM as a premium thematic fund, reflecting the specialised nature of the strategy.

The Roundhill Memory ETF reached $1 billion in assets under management within 14 trading days of its 2 April 2026 launch, placing it among the fastest ETF accumulations in history.

Liquidity matters because thematic launches often attract early attention but struggle to maintain trading depth. DRAM’s options market and consistent volume suggest the fund has moved beyond speculative interest into operational use by portfolios seeking targeted memory exposure.

Why memory semiconductors are attracting record investor attention

The memory chip sector has moved from cyclical oversupply to structural shortage in the span of 18 months, driven by AI workloads that demand high-bandwidth memory far faster than manufacturers can scale production. The supply constraint has been severe enough that traders coined the term “RAMageddon” to describe the bottleneck.

DRAM prices surged 148% year over year by April 2026. HBM prices escalated tenfold amid the supply constraints. AI-driven HBM demand is projected to increase 70% year over year, with broader AI-driven memory demand forecast to grow 40-50% annually. The global memory market is forecast to reach $400 billion by 2027.

The pricing surge reflects a structural shift rather than cyclical volatility. S&P Global Market Intelligence analysis citing Visible Alpha consensus estimates projects Samsung’s traditional DRAM revenue per bit will rise 116% in 2026, while HBM production constraints tighten supply across the memory product stack.

Three companies dominate the market. Samsung Electronics and SK Hynix, both South Korean firms, hold a combined HBM market share exceeding 70%. Micron Technology, the U.S.-based manufacturer, controls 20-25% of the overall DRAM market. The concentration reflects the technical and capital barriers to entry in advanced memory production, where fabrication facilities require multi-billion dollar investments and years to bring online.

The global memory market is forecast to reach $400 billion by 2027, driven by AI infrastructure demand that has transformed the sector from cyclical commodity to secular growth story.

The supply-demand imbalance is not a short-term pricing anomaly. AI data centre buildouts require HBM to feed processors running large language models and other compute-intensive workloads. The physical constraints on HBM production mean the shortage cannot be resolved quickly, supporting the case for sustained elevated pricing through 2026 and beyond.

What the DRAM ETF holds and how it differs from alternatives

Samsung Electronics and SK Hynix trade only on the Korea Exchange. They are excluded from major U.S. semiconductor ETFs, leaving investors who wanted direct exposure to memory leaders with limited options. The iShares MSCI South Korea ETF (EWY) offered indirect access, but its structure dilutes memory exposure with broader Korean market holdings.

DRAM positions itself as the first U.S.-listed ETF providing targeted global memory exposure. The fund’s portfolio concentrates on the three HBM leaders and related memory supply chain names. EWY, by contrast, holds Samsung and SK Hynix as components within a diversified Korean equity basket where the two memory manufacturers represent roughly 44% of holdings. The remaining 56% consists of non-semiconductor Korean companies, including industrial, defence, and consumer names.

EWY has attracted $6.2 billion in year-to-date inflows as of 20 April 2026, posting a year-to-date return of 54.38%. The fund added $235 million after DRAM’s debut, suggesting many investors view the two vehicles as complementary rather than substitutes. The choice depends on whether an investor seeks concentrated memory exposure or broader Korean market participation.

EWY maintains appeal for diversified Korea exposure

Korean defence and industrial stocks within EWY have performed well alongside the semiconductor surge, offering diversification benefits that DRAM does not provide. Investors seeking memory gains alongside wider Korean market participation may prefer EWY’s structure. DRAM’s launch appears to have absorbed some memory-focused demand that might otherwise have flowed entirely into EWY, but EWY’s continued inflows suggest demand for both thematic concentration and geographic diversification remains robust.

For investors weighing concentrated memory exposure against diversified Korean market participation, our detailed comparison of EWY versus DRAM positioning strategies examines the specific return profiles, volatility characteristics, and portfolio fit considerations for each vehicle across different market scenarios.

How high-bandwidth memory became AI’s critical bottleneck

High-bandwidth memory sits directly next to AI processors in data centre servers, feeding them data at speeds required to run large language models and other compute-intensive workloads. Traditional DRAM lacks the bandwidth to support these applications. HBM solves the problem by stacking multiple memory layers vertically and connecting them with through-silicon vias, multiplying data transfer rates.

The shift to AI workloads created demand for HBM that manufacturers could not anticipate or scale quickly enough to meet. Roundhill CEO Dave Mazza described memory as the “true bottleneck” in AI infrastructure, where demand far outpaces supply. The constraint is not abstract. Data centres ordering AI servers face wait times for HBM that delay deployments, and edge computing applications requiring on-device AI processing encounter similar supply limitations.

The HBM supply constraint has accelerated custom AI chip partnerships like Broadcom’s $100 billion OpenAI deal, where hyperscalers are designing memory architectures optimized for their specific workloads rather than waiting for commodity solutions to catch up with demand.

The technical barriers to scaling HBM production are not easily overcome through capital investment alone. IEEE peer-reviewed research on HBM architectural and manufacturing constraints identifies physical limits in vertical stacking and through-silicon via design that require years of process development to address, supporting the multi-year supply gap thesis.

Samsung Electronics is focusing on 3D DRAM advancements and AI-optimised chip development to address the shortage. SK Hynix plans to extend DDR4 lifecycles through 2026 to free up capacity for HBM production. Micron Technology announced expanded HBM production capacity on 4 April 2026, targeting AI workloads specifically.

• Samsung Electronics: Focusing on 3D DRAM advancements and AI-optimised chip development
• SK Hynix: Extending DDR4 lifecycles through 2026 to address HBM shortages
• Micron Technology: Announced expanded HBM production capacity on 4 April 2026, targeting AI workloads

Memory is the “true bottleneck” in AI infrastructure, where demand far outpaces supply, according to Roundhill CEO Dave Mazza.

The physical constraints on HBM production create a multi-year supply gap. Fabrication facilities require years to build and billions of dollars to equip. The shortage will not resolve in a single quarter, supporting the case for sustained demand and pricing power through 2026 and potentially beyond.

What early momentum signals about the memory investment thesis

The $1 billion milestone in 14 trading days signals a broader shift in how investors view memory semiconductors. The sector historically traded as a cyclical commodity, with boom-bust pricing cycles driven by oversupply and demand fluctuations. The AI-driven shortage has reframed memory as a secular growth story, with structural demand that outpaces supply for years rather than quarters.

Bloomberg Intelligence and Morningstar analysts cited potential for 30-40% annual returns if AI demand sustains at current trajectories. Institutional investor sentiment surveys indicate 75% plan increased allocations to AI-enabling technologies, a category that includes memory as a critical component. Samsung and SK Hynix stock prices rose 15-20% in April 2026, reflecting the market’s view that memory pricing power will persist.

• Bloomberg Intelligence and Morningstar: Cite potential for 30-40% annual returns if AI demand sustains
• Institutional sentiment surveys: 75% of institutional investors plan increased allocations to AI-enabling technologies
• Samsung and SK Hynix stock prices: Rose 15-20% in April 2026
• No competing launches: No new memory-focused ETFs announced by other issuers as of 21 April 2026

Analysts cite potential for 30-40% annual returns from memory semiconductors if AI demand sustains at current trajectories, though concentration risk remains.

The absence of direct competitors gives DRAM a first-mover advantage, but concentration in a narrow subsector means the fund will move sharply with memory pricing cycles. Investors should weigh the upside case against the volatility profile. If AI demand softens or if manufacturers scale production faster than expected, memory prices could compress quickly, reversing recent gains.

The stable geopolitical and regulatory backdrop as of 21 April 2026 has supported flows. No significant Korea-related geopolitical disruptions have emerged, and no major AI regulations have impacted the memory semiconductor sector. This stability may not persist, and any shift in U.S.-China trade policy or regional tensions could affect investor sentiment toward Korean manufacturers.

For readers evaluating whether memory’s current pricing power represents secular growth or cyclical peak, our full examination of AI infrastructure investment risks explores the demand sustainability thesis, capital intensity return profiles, and scenarios where $6 trillion in AI spending may not produce the returns infrastructure suppliers currently price in.

Conclusion

The Roundhill Memory ETF’s record launch reflects three converging forces: unmet investor demand for pure-play memory exposure, the AI-driven supply shortage pushing DRAM and HBM prices sharply higher, and the first-mover status that made DRAM the only U.S.-listed vehicle offering targeted access to Samsung Electronics and SK Hynix. The fund’s trading metrics and sustained inflows indicate institutional adoption rather than speculative momentum.

The distinction between DRAM’s concentrated memory thesis and EWY’s diversified Korean equity approach matters. DRAM offers direct exposure to the sector analysts describe as AI infrastructure’s critical bottleneck. EWY provides broader geographic participation that includes memory exposure alongside other Korean industries. The continued asset growth for both funds suggests investors see value in each structure.

The fund’s momentum will depend on whether AI demand sustains its current trajectory and whether memory pricing remains elevated through 2026. Investors should assess their existing semiconductor or Korean equity holdings to determine whether DRAM fills a gap in their portfolio or duplicates exposure they already have. Concentration in a narrow subsector delivers amplified returns when the thesis plays out, but it also carries volatility that may not suit all portfolios.

This article is for informational purposes only and should not be considered financial advice. Investors should conduct their own research and consult with financial professionals before making investment decisions.