What Is the Roundhill DRAM ETF? Buying the AI Memory Supply Chain with One ETF

The Roundhill DRAM ETF is a U.S.-listed thematic ETF focused on global memory chip and storage hardware companies. With one ticker, you can gain exposure to companies such as Micron, Samsung Electronics, SK hynix, SanDisk, Kioxia, Western Digital, and Seagate. Its advantage is high thematic purity and access to Asian memory leaders. Its limitation is also clear: concentrated holdings and strong industry cyclicality mean it cannot replace an S&P 500 ETF or a global broad-market ETF.

Key Takeaways

• DRAM is an actively managed ETF focused on global memory chip companies.
• Its top three holdings account for nearly 75%, making it more concentrated than broad-market funds.
• HBM is an important driver, but the fund also covers NAND, SSD, and HDD.
• AI supports long-term demand, while memory pricing cycles drive short-term earnings leverage.
• The 0.65% expense ratio is higher than semiconductor ETFs such as SOXX and SMH.
• DRAM is more suitable as a satellite position than as a standalone core allocation.

What Is the Roundhill DRAM ETF and How Does It Work?

The Roundhill Memory ETF trades under the ticker DRAM and was listed on Cboe BZX on April 2, 2026. Its objective is to seek capital appreciation through a portfolio of global memory companies. It is actively managed, so it does not mechanically replicate a fixed index. In practice, however, the portfolio is usually adjusted around quarterly rebalancing rather than through frequent short-term trading.

Based on the fund’s 0.65% annual expense ratio, if you invest $10,000, assume a 5% annual return, and the expense ratio remains unchanged, the prospectus example estimates first-year fund operating costs at about $68. This does not include brokerage commissions, bid-ask spreads, foreign exchange conversion costs, or taxes.

Which Companies Can Be Included in the DRAM ETF?

The fund does not include every company loosely associated with “storage.” To qualify, a company generally needs to derive at least 50% of its revenue or profit from developing or manufacturing the following products:

• High-bandwidth memory, or HBM;
• Dynamic random-access memory, or DRAM;
• NAND flash or SSDs using NAND;
• NOR flash;
• Hard disk drives, or HDDs;
• Specialty memory and embedded memory.

In addition, the fund usually requires eligible companies to have a market capitalization of at least $10 billion and average daily trading volume of no less than $5 million. Under normal circumstances, the fund invests at least 80% of its net assets in memory company stocks, depositary receipts, or financial instruments that provide related exposure.

Why Does the Fund Use Total Return Swaps?

DRAM holds companies listed in the United States, South Korea, Japan, mainland China, and Taiwan. Exposure to certain overseas stocks may be obtained through depositary receipts, forwards, or total return swaps. Roundhill states that one purpose of using swaps is to meet diversification and tax requirements for U.S. regulated investment companies.

Swap instruments can reduce some of the operational hurdles involved in accessing overseas markets, but they also introduce additional risks:

• Counterparties may fail to perform;
• Contract valuation may deviate from the underlying stocks;
• Positions may be difficult to unwind during market stress;
• Derivative structures make the fund harder to understand.

Active management does not mean the fund manager can avoid every industry downturn. If weighting decisions, stock selection, or swap arrangements fail to meet expectations, DRAM may still underperform the broader group of memory stocks.

Summary: DRAM is an actively managed global memory-themed ETF. Its core investment scope includes HBM, DRAM, NAND, SSD, and HDD. The fund uses stocks, depositary receipts, and certain derivatives to gain exposure to memory companies, while applying business revenue, market capitalization, and liquidity criteria. What you buy is not the entire semiconductor industry, but a concentrated thematic portfolio built around memory manufacturers.

What Companies Do You Actually Buy Through the DRAM ETF?

When you buy the DRAM ETF, your main exposure comes from Micron, Samsung Electronics, and SK hynix. As of June 30, 2026, these three companies together represented 74.84% of the fund, meaning their earnings, valuations, and stock price movements largely determine the ETF’s direction. The remaining holdings mainly add exposure to NAND, SSD, HDD, NOR flash, and niche memory products.

The data comes from the fund’s DRAM holdings and asset allocation as of June 30, 2026. During the same period, about 49.04% of assets were exposed to South Korean companies and 39.51% to U.S. companies, so “global allocation” does not mean evenly diversified by region.

What Do HBM, DRAM, NAND, and HDD Each Contribute?

HBM stacks multiple layers of DRAM vertically and uses wider data channels to feed data to GPUs or AI accelerators. It directly benefits from AI training, high-performance computing, and large-scale inference demand. Its unit value and technical barriers are generally higher than those of standard server memory.

Commodity DRAM is used in servers, PCs, smartphones, and other computing devices. Even when AI demand is strong, commodity DRAM is still affected by consumer electronics shipments, inventory changes, and capacity adjustments. As a result, the DRAM ETF is not only a bet on AI GPUs; it also carries traditional electronics-cycle exposure.

NAND flash is used for long-term data storage and forms the basis of enterprise SSDs, personal computers, smartphones, and data center storage systems. HDDs are slower than SSDs, but their lower cost per unit of capacity makes them suitable for training data, backups, logs, and cold data storage.

Which Segments Does the DRAM ETF Not Fully Cover?

Although DRAM is tied to the memory industry, it does not fully cover the following segments:

• Semiconductor equipment companies such as ASML;
• Foundries such as TSMC;
• Advanced packaging such as CoWoS;
• Memory controllers and high-speed interface chips;
• AI server system manufacturers;
• Cloud computing operators;
• Database and enterprise storage software.

In addition, SK hynix is now listed on Nasdaq through SKHY, allowing international investors to gain direct ADR exposure. DRAM’s convenience is therefore no longer only about accessing Korean stocks that may be difficult to trade directly. Its stronger value is that it combines multiple memory leaders into one portfolio and adjusts their weights automatically.

Summary: DRAM is not evenly diversified across a large number of memory companies. It is centered on Micron, Samsung, and SK hynix, with additional exposure to SanDisk, Kioxia, Western Digital, and Seagate. It covers the key products from high-bandwidth memory to flash memory and hard drives, but it does not include the full AI infrastructure chain, such as equipment, foundry, advanced packaging, servers, or cloud services.

Why AI Memory Demand Can Drive the DRAM ETF

AI can raise the long-term demand ceiling for memory companies, but it does not eliminate industry cycles. Training and inference require higher bandwidth, larger capacity, and faster data access, which supports HBM, server DRAM, and enterprise SSD demand. At the same time, supply expansion, inventory, yield rates, and product pricing still determine the actual profits of memory manufacturers.

AI data centers require a full memory and storage hierarchy. GPUs and AI accelerators need HBM to deliver data quickly. CPUs require server DRAM to run workloads. SSDs support frequent access to training datasets and model files. HDDs store large amounts of data at lower cost.

How AI Training and Inference Increase Memory Demand

• Larger model parameters and longer context windows increase memory capacity requirements;
• Multi-GPU clusters need higher bandwidth to reduce processor idle time;
• Higher inference concurrency increases KV cache and server memory demand;
• Training data, model checkpoints, and logs require enterprise SSDs;
• Historical data, backups, and archives continue to support nearline HDD capacity growth.

Taking Samsung HBM products as an example, high-bandwidth memory uses a stacked structure to increase capacity and data throughput within a single package. Target applications include AI training, inference, GPUs, and high-performance computing. However, HBM requires advanced process technology, stacking, packaging, and yield improvement, so capacity expansion is often slower than for standard memory.

Why the Memory Industry Remains Highly Cyclical

Memory chips are relatively standardized products, so prices are highly sensitive to supply-demand gaps. A typical cycle often moves through the following stages:

1. Server, AI, or consumer electronics demand improves;
2. Channel and customer inventories decline;
3. DRAM and NAND contract prices rise;
4. Manufacturer margins and cash flow improve rapidly;
5. Companies increase capital expenditure and expand capacity;
6. New capacity comes online and supply-demand conditions ease;
7. Product prices and earnings expectations decline.

Indicators for Judging Whether the AI Memory Trade Can Continue

You should not only watch GPU shipments. The following indicators also matter:

• HBM customer qualification, product generation, and yield;
• DRAM and NAND contract prices;
• Memory company inventory days;
• Wafer capacity utilization;
• Capital expenditure and expansion pace;
• Enterprise SSD and nearline HDD shipments;
• Cloud provider AI capital expenditure;
• Memory company gross margins and next-quarter guidance.

DRAM attracted more than $6 billion in assets within five weeks of launch, making it one of the fastest-growing thematic ETFs at the time and showing how quickly investor attention shifted toward AI memory. Fast inflows can improve fund scale and liquidity, but they may also create crowded positioning, making prices more sensitive to earnings misses, capacity expansion, or changes in market sentiment.

Summary: AI supports long-term demand for HBM, server DRAM, enterprise SSDs, and high-capacity HDDs. However, the medium- and short-term returns of the DRAM ETF are still jointly driven by pricing, inventory, capacity, earnings, and valuation. The AI memory trade is more likely to continue when end demand grows faster than new supply, and when company earnings can keep up with valuation expansion.

How Is the DRAM ETF Different from SOXX, SMH, and Memory Stocks?

DRAM’s advantage is concentrated memory exposure with access to both U.S. and Asian leaders. The trade-off is a higher expense ratio, fewer holdings, and stronger sensitivity to the memory cycle. SOXX and SMH cover a broader set of semiconductor companies and are usually more suitable as industry-level allocations. Buying individual stocks further concentrates returns into one company.

SOXX and SMH hold chip designers, manufacturers, equipment companies, and other semiconductor businesses. When GPUs, foundries, or equipment stocks rise while memory prices weaken, they may significantly outperform DRAM. Conversely, when HBM, DRAM, and NAND all enter an upcycle, DRAM’s earnings leverage is usually more concentrated.

DRAM vs. Buying Memory Stocks Directly

Buying DRAM can reduce single-company risks such as product delays, customer losses, yield issues, or management execution problems. However, it does not eliminate memory industry risk. Because the top three holdings account for almost three-quarters of the portfolio, the remaining holdings may not fully offset losses if Samsung, SK hynix, and Micron fall at the same time.

Buying Micron or SKHY directly does not involve a fund management fee, and you can decide your own entry timing and position size. However, single-stock investing requires continuous analysis of product cycles, customer qualifications, capital expenditure, competition, and valuation, making the research burden higher.

Beyond Expense Ratios, Actual Trading Costs Matter

The 0.65% fund expense ratio is continuously deducted from fund assets, but it is not your full cost of trading DRAM. Actual costs may also include:

• Brokerage commissions and platform fees;
• ETF bid-ask spreads;
• Execution slippage;
• Currency conversion costs;
• External agency fees;
• Taxes related to dividends or capital gains.

You can first use U.S. stock market data to check DRAM, SOXX, SMH, and major memory holdings, then compare fund fees with order-level costs. Biya U.S. stock trading fees show that U.S. stock trading commission is $0, with a platform fee of $0.005 per share, subject to a minimum of $0.99 per order and a maximum of 1% of trade value; external agency and trading activity fees total $0.00396 per share. Specific fractional-share fees, applicable conditions, and other costs should be based on the fee disclosure and order page.

DRAM and RAM Are Not the Same Product

The RAM 2X Long DRAM Daily Target ETF seeks two times the daily performance of DRAM. It began trading on June 24, 2026, and has a net expense ratio of 1.25%. It rebalances daily, so over periods longer than one trading day, cumulative returns may deviate significantly from two times DRAM’s period return.

If DRAM repeatedly rises and falls in a volatile market, RAM may be affected by compounding path dependence and volatility decay. Even if DRAM rises slightly over a given period, RAM could still lose money. It is therefore mainly designed for professional traders who can monitor positions daily.

Summary: If you want broad semiconductor industry exposure, SOXX or SMH is usually more balanced than DRAM. If you have a strong view on HBM and the memory pricing cycle, DRAM offers higher thematic purity. If you can research individual companies deeply, memory stocks may be an alternative. RAM is a daily leveraged trading tool and should not be treated as a long-term enhanced version of DRAM.

What Are the Risks of the Roundhill DRAM ETF and Who Is It Suitable For?

DRAM is more suitable for investors who already have a broad-market core position, can tolerate large drawdowns, and are willing to track memory pricing and capacity cycles. It is not suitable for investors seeking low volatility, stable dividends, or a single fund to handle long-term asset allocation. When deciding whether to buy, position sizing is usually more important than short-term price forecasting.

The fund documents highlight risks related to memory companies, the semiconductor industry, the South Korean market, non-diversification, active management, and swap agreements. During market stress, if authorized participants or market makers reduce activity, bid-ask spreads may widen and ETF prices may temporarily deviate from NAV.

Which Investors May Be Suitable for DRAM?

• Investors already holding the S&P 500, global equities, or other broad-market ETFs;
• Investors treating DRAM as a satellite thematic allocation;
• Investors who understand the differences among HBM, DRAM, NAND, and HDD;
• Investors willing to track product prices, inventory, and capital expenditure;
• Investors able to tolerate large interim drawdowns;
• Investors with clear position limits and rebalancing rules.

When Should Investors Avoid Chasing the Theme?

• Your only reason for buying is recent outperformance;
• Your portfolio already has heavy exposure to SOXX, SMH, Micron, or SK hynix;
• The market discusses AI demand but ignores capacity expansion and valuation;
• Memory company capital expenditure accelerates while demand growth begins to slow;
• You cannot tolerate a 20% or larger drawdown in a thematic fund;
• You plan to use living expenses or short-term funds to invest.

Pre-Investment Checklist

1. Does your current portfolio already hold a large amount of semiconductor and technology exposure?
2. How much overlap does DRAM have with SOXX, SMH, or QQQ?
3. Can you tolerate a memory industry downcycle?
4. Have you checked the fund expense ratio, spread, platform fees, and FX costs?
5. Have you set thematic position limits, stop-loss rules, or rebalancing conditions?
6. Are you reviewing the latest dated holdings rather than relying on launch-period data?

Summary: DRAM is a highly concentrated, high-beta, and highly cyclical thematic ETF. It is more suitable as a supplement to a broad-market portfolio than as a replacement for core assets. You need to assess long-term AI memory demand, the current industry cycle, top-three holding valuations, and your own risk tolerance, while using position sizing to reduce the impact of being wrong.

Before deciding whether to allocate, you can check DRAM’s latest holdings, NAV, bid-ask spread, and overlap with your existing technology positions. Users who meet local availability, identity verification, and platform rules can use Biya to view U.S. stocks, Hong Kong stocks, and other asset classes, and compare DRAM, SOXX, SMH, Micron, and other memory-related names. You can also download the Biya App to manage watchlists and trading records. Service availability depends on the user’s location, identity verification result, platform rules, and applicable laws and regulations. All fees are subject to the actual order and fee disclosures. Public market information, product structure, and fee comparisons do not constitute investment advice.

FAQ

Is the Roundhill DRAM ETF a Pure HBM ETF?

No. Although DRAM is heavily influenced by HBM demand, its investment scope also includes commodity DRAM, NAND, SSD, HDD, NOR flash, and embedded memory. Samsung, Micron, and SK hynix each have multiple memory businesses, so the fund’s performance is not determined by HBM alone.

Does the Roundhill DRAM ETF Pay Regular Dividends?

DRAM may distribute income based on dividends from its holdings and the fund’s annual distribution policy, but the amount and frequency are not guaranteed. The fund’s primary objective is capital appreciation rather than stable cash flow. Actual distributions, tax treatment, and payment amounts should be verified through fund announcements and brokerage account statements.

Do International Investors Face Currency Risk When Buying the DRAM ETF?

Yes. DRAM trades in U.S. dollars, but its underlying companies are based in the United States, South Korea, Japan, mainland China, and Taiwan. The fund’s NAV may be indirectly affected by local currencies, securities markets, and depositary receipt prices. You may also incur FX conversion and trading costs when converting your local currency into U.S. dollars.

Is the DRAM ETF Less Risky Than Micron Stock?

DRAM diversifies some of Micron’s single-company risks, such as customer, product, and execution risk, but its top three holdings account for nearly 75%, so memory industry cyclicality remains significant. Micron stock is more concentrated. DRAM offers relatively more company-level diversification, but that does not mean its volatility will always be lower.

Is the DRAM ETF Better for Long-Term Dollar-Cost Averaging or Cycle Trading?

Both approaches require conditions. Long-term dollar-cost averaging requires confidence that AI memory demand can outlast multiple pricing cycles while keeping thematic exposure controlled. Cycle trading requires tracking prices, inventories, capital expenditure, and valuation. Recent outperformance alone should not determine the buying approach.

What Costs Exist Besides the DRAM ETF Management Fee?

In addition to the 0.65% annual expense ratio, investors may face platform fees, commissions, bid-ask spreads, slippage, FX conversion costs, external agency fees, and taxes. Costs vary by platform, account location, and order size, so investors should rely on order details and local regulatory requirements.