Best Memory and Data Storage ETFs to Buy in Singapore

The next bottleneck in artificial intelligence may not be the processor. It may be the memory needed to keep that processor running at full speed.

Global memory-chip revenue is forecast to surge from US$216.3 billion in 2025 to US$633.3 billion in 2026, before reaching US$748.1 billion in 2027. Gartner calls the phenomenon “memflation”: average DRAM prices are projected to rise 125% in 2026, while NAND flash prices could climb 234%, with no meaningful pricing relief expected until late 2027.

Investors have moved just as quickly. The Roundhill Memory ETF (DRAM) launched on 2 April 2026 and accumulated nearly US$10 billion within its first 45 days, breaking records for early ETF asset growth. By mid-June, dedicated US memory ETFs collectively managed around US$20 billion. HBMX, KMEM and DISK followed within weeks, while Europe gained its first dedicated memory UCITS ETF through the Ireland-domiciled Defiance Memory UCITS ETF.

But the rush into memory ETFs also shows why investors need to look beyond the headline. Some funds place more than three-quarters of their portfolios in Micron, Samsung Electronics and SK Hynix—the three companies dominating advanced DRAM and high-bandwidth memory production. Others lean towards NAND flash, solid-state drives and persistent storage, while broader funds add semiconductor equipment, materials and advanced-packaging companies.

These are not interchangeable exposures. An HBM-focused ETF is primarily a bet on AI servers and memory bandwidth. A NAND-heavy fund depends more on flash-storage prices and demand for enterprise and consumer storage. A wider memory-supply-chain ETF may benefit from rising manufacturing investment but can behave more like a semiconductor-equipment fund.

The category is also extremely young, concentrated and volatile. Most dedicated memory ETFs have only been trading since the second quarter of 2026, leaving investors with little performance history and limited evidence of how the funds will behave through a full memory cycle.

This guide compares the main memory and data storage ETFs accessible from Singapore, covering:

1. Pure-play memory ETFs and broader semiconductor alternatives
2. HBM and DRAM exposure versus NAND, SSD and HDD exposure
3. Producer-focused funds and the wider memory supply chain
4. US-listed ETFs and the Ireland-domiciled UCITS option
5. Fees, concentration, liquidity and Singapore brokerage access
6. The risks of investing in a newly created and highly cyclical ETF category

Best memory and data storage ETFs at a glance

The dedicated memory ETF market is still small, but the funds are not interchangeable. The main difference is whether the portfolio concentrates on HBM and DRAM producers, extends into NAND and storage hardware, or owns the wider manufacturing stack.

Fund sizes are approximate and dated to mid-July 2026. Holdings and assets can change quickly, particularly for newly launched active funds.

Roundhill DRAM is the largest and most liquid dedicated fund, while KMEM makes the most concentrated bet on the three leading memory producers. DISK is the clearest option for investors looking beyond HBM towards NAND, SSD and storage companies. Defiance DRAM is the only dedicated Ireland-domiciled UCITS option.

The two funds using DRAM as a ticker are different products. Roundhill DRAM is a US-listed active ETF, while Defiance DRAM is an Ireland-domiciled UCITS ETF listed in Europe. Check the exchange and ISIN before placing an order.

What is a memory and data storage ETF?

A memory and data storage ETF invests in companies involved in storing, retrieving and moving digital information. Depending on its mandate, the fund may own memory-chip manufacturers, NAND and flash-storage producers, hard-disk-drive makers, memory-controller designers, or suppliers of the equipment and packaging used to manufacture advanced memory.

That distinction matters because “memory ETF” is a broad label. A portfolio concentrated in Micron, Samsung Electronics and SK Hynix behaves differently from one that also owns Kioxia, SanDisk, Seagate, Western Digital, semiconductor-equipment manufacturers and packaging suppliers.

Memory and storage technologies explained

HBM and DRAM are volatile memory: their contents disappear when power is removed. NAND, SSDs and HDDs provide persistent storage. Both are needed in an AI system, but they respond to different pricing cycles, customer demand and production constraints.

Why memory and data storage became an investable theme in 2026

AI systems need more memory bandwidth

AI accelerators can only process data as quickly as memory can supply it. As each generation of GPU becomes more powerful, the amount and speed of memory packaged alongside it must also increase.

HBM addresses this problem by stacking memory dies vertically and placing them close to the processor. That provides substantially more bandwidth than conventional memory while using less power per unit of data moved. Demand is therefore tied closely to shipments of AI accelerators and the capital expenditure of hyperscalers such as Microsoft, Amazon, Alphabet and Meta.

The investment case is not simply that more AI chips will be sold. It is that memory content per system is increasing, making HBM a larger part of the cost and performance of each AI server.

HBM production competes with conventional memory capacity

Memory manufacturers cannot add advanced capacity overnight. HBM requires additional wafer processing, stacking, testing and packaging, while production yields take time to improve.

As Micron, Samsung Electronics and SK Hynix direct more capacity towards higher-value HBM, less capacity may be available for conventional DRAM and NAND. That can tighten supply across the wider memory market rather than only the AI-specific segment.

The result is a theme with two linked drivers:

1.   Higher demand for advanced memory used directly in AI systems.

2.   Tighter supply and stronger pricing across conventional memory products.

Persistent storage demand is rising too

AI workloads do not end with HBM. Training datasets, model checkpoints, embeddings, logs and inference outputs still need to be stored.

The demand is already visible in company results. Seagate reported fiscal third-quarter 2026 revenue of US$3.11 billion, up 44% year on year, supported by nearline storage demand. SanDisk reported fiscal third-quarter 2026 revenue of US$5.95 billion, with data-centre revenue rising 233% year on year.

That broadens the investable theme beyond HBM. NAND, SSD and HDD manufacturers can benefit when the quantity of data generated by AI grows, even though their economics differ from those of advanced DRAM producers.

Memory remains a cyclical industry

Strong structural demand does not remove the memory cycle. DRAM and NAND prices have historically risen sharply during shortages and fallen just as quickly when new capacity, weaker device demand or excess inventories created oversupply.

Capacity decisions made today can influence supply several years later. High prices encourage manufacturers to invest, but fabrication plants and packaging facilities take time to build. This creates a lag between the investment decision and the additional supply reaching the market.

Investors therefore need to separate two ideas:

•    AI can increase the long-term amount of memory and storage required.

•    Memory-company earnings and share prices can still move through severe short-term cycles.

The memory and data storage ETF universe

Memory exposure is available through three broad fund types.

Data-centre REIT ETFs, cloud-computing ETFs and broad AI ETFs are adjacent exposures rather than direct substitutes. They may benefit from the same infrastructure buildout, but they do not have the same sensitivity to DRAM, NAND or storage-hardware prices.

Best US-listed memory and data storage ETFs

Four dedicated funds now trade on US exchanges. All launched between April and June 2026, and all are actively managed.

 Roundhill Memory ETF (DRAM): best for concentrated HBM and DRAM exposure

The Roundhill Memory ETF was the first dedicated US memory ETF and remains the category’s dominant fund by assets.

Its mandate covers companies involved in HBM, DRAM, NAND, SSDs, NOR flash, hard-disk drives and other forms of memory. In practice, the portfolio has been driven primarily by the largest memory manufacturers, including Micron, Samsung Electronics and SK Hynix, alongside storage names such as SanDisk and Kioxia.

DRAM is actively managed, so its holdings can change materially between disclosure dates. The fund also uses total-return swaps to obtain some of its exposure while remaining within US regulated-investment-company diversification rules. Investors should therefore assess both direct equity positions and swap exposure when calculating the true weight of each company.

Why it stands out: DRAM provides the most established single-fund route into the companies at the centre of the HBM and DRAM shortage. Its much larger asset base also makes it more practical to trade than the newer dedicated funds.

Main risk: The portfolio can be dominated by a few producers. The swap structure adds counterparty and valuation considerations, while the fund has not yet traded through a full memory cycle.

Best suited for: Investors seeking direct, high-conviction exposure to HBM and DRAM producers and who are comfortable with substantial concentration.

Tuttle Capital Concentrated Memory Stack ETF (HBMX): best for the wider production ecosystem

The Tuttle Capital Concentrated Memory Stack ETF takes a broader view of the memory industry.

HBMX targets approximately 20 to 35 companies across DRAM, NAND and HBM production, as well as advanced packaging, outsourced assembly and testing, substrates, materials, interconnects and semiconductor equipment. Companies generally need meaningful memory-related revenue or a substantial strategic focus on the memory stack to qualify.

This makes HBMX less dependent on the share prices of Micron, Samsung Electronics and SK Hynix. It can also capture companies that benefit when memory manufacturers raise capital expenditure, even when the price cycle for finished chips becomes less favourable.

Why it stands out: HBMX offers the broadest exposure to the equipment, materials, testing and packaging required to increase memory production.

Main risk: Its wider mandate dilutes direct exposure to memory pricing. At times, the fund may behave more like a semiconductor-equipment ETF than a pure memory ETF. The 0.95% expense ratio is also the highest among the dedicated US funds.

Best suited for: Investors who want to participate in memory capital expenditure without allowing three producers to determine most of the portfolio’s return.

Kurv Memory Select ETF (KMEM): best for concentrated memory producers

The Kurv Memory Select ETF is the most concentrated dedicated memory fund in this comparison.

Its look-through exposure as of 30 June 2026 placed 41.53% in SK Hynix, 19.85% in Micron and 18.81% in Samsung Electronics. Together, those three companies represented 80.19% of the portfolio. The remaining exposure included Kioxia, SanDisk, Western Digital and Seagate at much smaller weights.

Why it stands out: KMEM provides the strongest direct tilt towards SK Hynix and the three companies that dominate advanced memory production.

Main risk: A single-company weight above 40% is unusually high even for a thematic ETF. The fund is also new, with a wider 30-day median bid-ask spread than a large, established ETF.

Best suited for: Investors making a deliberate high-conviction bet on the leading memory producers, particularly SK Hynix. It is not a diversified semiconductor allocation.

Tema Memory ETF (DISK): best for NAND, flash and data storage exposure

The Tema Memory ETF is the clearest dedicated option for investors who want persistent storage alongside memory-chip exposure.

As of 17 July 2026, Kioxia and SanDisk were its two largest holdings at 16.33% and 16.00%. The portfolio also held SK Hynix, Samsung Electronics, Seagate, Western Digital, Micron, Nanya Technology and Innodisk.

That composition gives DISK a materially different earnings profile from DRAM and KMEM. NAND and storage companies are influenced by enterprise-storage demand, consumer-device shipments, flash prices and hyperscale data growth rather than HBM demand alone.

Why it stands out: DISK has the most differentiated persistent-storage exposure, with meaningful allocations to Kioxia, SanDisk, Seagate and Western Digital.

Main risk: NAND pricing has historically been highly cyclical. The thesis that AI-generated data will translate into stronger storage-company earnings can also be disrupted by price declines, efficiency gains or excess capacity.

Best suited for: Investors who believe the AI storage opportunity extends beyond HBM into NAND, SSDs, HDDs and enterprise storage.

Best UCITS memory and semiconductor ETFs

The UCITS market offers one dedicated memory ETF and several broader semiconductor funds with meaningful memory-company holdings.

Defiance Memory UCITS ETF (DRAM): best dedicated UCITS memory ETF

The Defiance Memory UCITS ETF is Europe’s first dedicated memory and storage UCITS ETF.

It is domiciled in Ireland and tracks the Indxx Defiance Global Memory Chip Select Index. Eligible companies must generate a substantial share of their revenue from memory or storage technologies, including HBM, DRAM, NAND, NOR flash, SSDs, memory controllers, emerging memory and mass-storage systems.

Unlike the actively managed US funds, Defiance DRAM follows a rules-based index. It uses physical replication, caps individual constituents at 10%, rebalances quarterly and reinvests income through an accumulating share class.

Why it stands out: It is the only dedicated memory and storage ETF currently available in an Ireland-domiciled UCITS structure. The accumulating share class also reinvests income automatically.

Main risk: The fund remains new and substantially smaller than Roundhill DRAM. Index caps reduce single-company concentration, but they can also limit exposure to the dominant producers during strong rallies.

Best suited for: Investors who prioritise an Ireland domicile, physical replication and an accumulating UCITS structure.

Ticker check: Roundhill’s US-listed fund and Defiance’s UCITS fund both use DRAM. Use the exchange and ISIN to identify the correct ETF.

Broader UCITS semiconductor ETFs with memory exposure

A broad semiconductor ETF can provide exposure to memory producers while spreading risk across chip designers, foundries and equipment companies.

AUM figures are approximate and use each issuer’s latest available reporting date. They are shown in the fund’s reporting currency and are not directly comparable without conversion.

The iShares fund provides the broadest global portfolio and can hold South Korean memory manufacturers directly. VanEck is larger and more concentrated but applies listing requirements that reduce direct exposure to companies traded only in Asia. Amundi and HSBC sit between those approaches.

A broad semiconductor ETF may suit investors who want memory exposure without making the entire allocation depend on a small number of manufacturers. The trade-off is that Nvidia, Broadcom, TSMC, ASML and other non-memory companies may drive most of the fund’s return.

Best way to invest in memory and data storage ETFs

Memory and data storage ETFs can be highly volatile, particularly because many are concentrated in a small group of companies and remain sensitive to DRAM, NAND and HBM pricing cycles. Regular investing can be a practical way to build exposure gradually without trying to time each market move.

With StashAway ETF Explorer, investors can choose an eligible ETF and automate recurring investments using cash or SRS. ETF Explorer normally charges US$1, excluding GST, per buy or sell order, with no additional management fee. Continue investing in the same ETF Explorer portfolio each month and buy orders can remain free in the following month.

This can be especially useful for smaller monthly investments, where recurring transaction fees would otherwise reduce the amount invested. Availability differs by ETF, so check whether the specific memory, storage or semiconductor fund is supported before setting up a recurring investment.

HBM, DRAM or NAND: which ETF owns the exposure you want?

Recent returns are not a useful starting point because most dedicated funds have only traded for a few weeks or months. The more important question is which part of the memory and storage stack the ETF owns.

An investor expecting HBM shortages to persist may prefer Roundhill DRAM or KMEM. Someone expecting AI data growth to support NAND and enterprise storage may find DISK more aligned. HBMX is the broader option when the objective is to own the businesses building memory-production capacity rather than only the chip producers.

Dedicated memory ETF vs semiconductor ETF vs data-centre ETF

These funds may all be described as AI infrastructure investments, but they sit at different points in the value chain.

A data-centre ETF owns the facilities that house servers. A cloud ETF owns platforms that sell computing and storage services. Neither provides the same earnings exposure as a fund directly holding Micron, SK Hynix, Samsung Electronics, Kioxia, SanDisk, Seagate or Western Digital.

How to compare memory and data storage ETFs

1. Measure direct memory and storage exposure

Start with the holdings rather than the fund name. Calculate how much of the portfolio is invested in:

•    HBM and DRAM producers

•    NAND and flash-storage companies

•    HDD and enterprise-storage companies

•    Semiconductor equipment, materials and packaging

•    Unrelated chip designers, foundries and software businesses

A fund with 70% in three memory producers is fundamentally different from one with 30% in equipment and packaging companies, even when both use “memory” in the name.

2. Check top-three concentration

Memory production is controlled by a small group of companies. This makes concentration partly unavoidable in a pure-play ETF, but the degree varies significantly.

KMEM had more than 80% of its look-through exposure in SK Hynix, Micron and Samsung Electronics as of 30 June 2026. Defiance DRAM caps individual constituents at 10%, while HBMX spreads its portfolio across a wider manufacturing ecosystem.

Higher concentration can produce stronger returns when the leading producers outperform, but it also increases company-specific, regulatory and execution risk.

3. Understand country and listing exposure

The memory supply chain spans South Korea, the US, Japan, Taiwan and China. Check whether the ETF owns local shares directly, uses depositary receipts, or obtains exposure through swaps.

This matters because different structures can affect:

•    Trading hours and price discovery

•    Currency exposure

•    Withholding taxes

•    Counterparty risk

•    Access to companies without US listings

An ETF limited to US-listed securities may exclude or underweight Samsung Electronics, SK Hynix and Kioxia even when those companies are central to the industry.

4. Compare active management with index rules

Neither structure is automatically better. The relevant question is whether the portfolio rules deliver the exposure you expect.

5. Treat AUM, volume and bid-ask spread separately

AUM measures the assets held by the fund. Trading volume measures how frequently ETF shares change hands. The bid-ask spread measures the cost of entering or leaving a position.

A large AUM does not guarantee a tight spread at every point in the trading day. Conversely, an ETF with low screen volume can still trade efficiently when its underlying securities are liquid and market makers are active.

For new funds:

•    Use limit orders rather than market orders.

•    Avoid trading immediately after the market opens.

•    Compare the market price with indicative NAV where available.

•    Check whether the main underlying Asian markets are open.

•    Review average daily trading value, not share volume alone.

6. Look beyond the expense ratio

The lowest TER does not necessarily mean the lowest total cost. A wide spread or expensive FX conversion can outweigh several years of fee differences. See how ETF expense ratios and other fees affect the actual cost of investing.

7. Do not annualise a few months of performance

Most dedicated memory ETFs launched in the second quarter of 2026. A one-month or three-month return should not be presented as an annualised indicator of expected performance.

The more useful early measures are:

•    Since-inception total return

•    Maximum drawdown since launch

•    Premium or discount to NAV

•    Bid-ask spread

•    Fund flows

•    Holdings concentration

•    Changes in portfolio composition

When comparing a new memory ETF with a long-established semiconductor ETF, use a common start date and label the short period clearly.

US-listed vs UCITS memory ETFs for investors in Singapore

The decision is not simply about which exchange has the highest recent return. Domicile changes how the fund distributes income, how it is accessed and how it may be treated for estate-tax purposes.

For non-US investors, US-situs assets can be subject to US estate-tax rules, with the general exemption potentially as low as US$60,000 depending on personal circumstances and treaty treatment. An Ireland-domiciled UCITS ETF avoids holding US ETF shares directly, although taxes can still arise within the fund on dividends from its underlying holdings.

The usual comparison between 15% and 30% US dividend withholding is less decisive here than it is for an income ETF. Memory and storage companies are global, dividend yields are generally low, and these funds are mainly held for capital growth.

Estate-tax exposure, accumulation, exchange access, spreads and portfolio construction may therefore matter more than headline dividend withholding. Tax treatment depends on individual circumstances and should be confirmed against current IRAS, IRS and fund documentation.

Risks of investing in memory and data storage ETFs

The same factor can benefit one ETF and hurt another. Higher HBM prices may support producer-heavy funds but squeeze customers. Higher capital expenditure can benefit equipment holdings in HBMX while raising the risk of future oversupply for DRAM and NAND manufacturers.

Where to buy memory and data storage ETFs in Singapore

After deciding which part of the memory and storage value chain fits your portfolio, the next consideration is whether your platform supports the ETF’s exchange listing and trading currency.

The dedicated US-domiciled funds, including DRAM, HBMX, KMEM and DISK, trade on US exchanges. The Ireland-domiciled Defiance Memory UCITS ETF is available through European exchanges, while broader UCITS semiconductor funds from iShares, Amundi, HSBC and VanEck commonly trade on the London Stock Exchange and Xetra.

Access therefore varies considerably between platforms. Some local brokerages cover Singapore, US and UK markets but charge relatively high minimum commissions. Global brokers generally offer lower trading costs and wider exchange access, while simplified investment platforms can be more practical for smaller or recurring purchases.

The cheapest platform depends partly on order size. A percentage-based fee may be competitive for smaller trades, while a flat commission can become more efficient as the investment amount increases. Minimum commissions, currency-conversion spreads, custody charges and exchange access should be considered alongside the advertised trading fee.

Availability should also be checked using the ETF’s full name, exchange and ISIN rather than its ticker alone. The Roundhill Memory ETF and Defiance Memory UCITS ETF both use the ticker DRAM on different exchanges, while UCITS ETFs can use different tickers across their USD, GBP and EUR trading lines.

Frequently asked questions

What is the best memory ETF?

There is no single best fund because each ETF targets a different part of the market.

Roundhill DRAM is the largest dedicated fund and offers concentrated HBM and DRAM exposure. KMEM has the highest concentration in SK Hynix, Micron and Samsung Electronics. DISK provides the strongest NAND, flash and storage tilt, while HBMX covers the wider manufacturing stack. Defiance DRAM is the dedicated option for investors who prefer an Ireland-domiciled UCITS structure.

Is there a data storage ETF?

There is no single large ETF devoted only to SSD and HDD manufacturers, but several memory ETFs include meaningful storage exposure.

DISK has the clearest tilt towards Kioxia, SanDisk, Seagate and Western Digital. Roundhill DRAM and Defiance DRAM can also hold NAND, SSD and mass-storage companies. Data-centre and cloud ETFs are separate categories because they own facilities and service providers rather than the companies manufacturing memory and storage hardware.

Is there a UCITS memory ETF?

Yes. The Defiance Memory UCITS ETF is an Ireland-domiciled, physically replicated, accumulating ETF with ISIN IE000CEUZ052.

It tracks a global memory-chip and storage index and is listed on the London Stock Exchange, Xetra, Borsa Italiana and Euronext Paris. It is the only dedicated memory UCITS ETF available as of July 2026.

Does a semiconductor ETF already include memory stocks?

Many broad semiconductor ETFs hold Micron, SK Hynix or other memory manufacturers, but the combined allocation varies widely.

The iShares MSCI Global Semiconductors UCITS ETF can hold South Korean producers directly within a broad global portfolio. Other funds may have larger weights in Nvidia, Broadcom, TSMC or semiconductor-equipment companies. Check the total weight of Micron, Samsung Electronics, SK Hynix, Kioxia, SanDisk, Seagate and Western Digital rather than assuming that every semiconductor ETF provides substantial memory exposure.

What is the difference between DRAM the ETF and DRAM the technology?

DRAM is dynamic random-access memory, the temporary working memory used in servers, PCs, smartphones and other electronic devices.

DRAM is also the ticker used by two separate ETFs. Roundhill DRAM is an active US-listed fund, while Defiance DRAM is an Ireland-domiciled UCITS ETF listed on European exchanges. Their structures, holdings and tax considerations differ, so investors should confirm the exchange and ISIN before trading.

Are memory ETFs suitable for long-term investing?

They can provide long-term exposure to rising AI and data-storage demand, but they remain concentrated thematic funds operating in a historically cyclical industry.

Most dedicated memory ETFs also have only a few months of operating history. They are generally more suitable as a targeted satellite allocation alongside a diversified portfolio than as a replacement for a broad global or semiconductor ETF.

How memory and data storage ETFs can fit into your portfolio

Memory and data storage ETFs can add direct exposure to one of the most supply-constrained parts of the AI infrastructure chain. They may complement a broad global ETF, S&P 500 ETF or Nasdaq-100 ETF when the goal is to own HBM, DRAM, NAND and storage companies that have small or no weights in a standard index.

The appropriate fund depends on the portfolio role. Roundhill DRAM and KMEM provide concentrated exposure to the leading memory producers. DISK adds a stronger NAND and persistent-storage tilt. HBMX extends into equipment, materials, testing and packaging. Defiance DRAM provides the only dedicated accumulating UCITS route, while broad funds such as iShares SEMI offer memory exposure with greater diversification across the semiconductor industry.

Before adding one, check the holdings already present in your portfolio. A global,  emerging-markets or Asia ETF may already own Samsung Electronics and SK Hynix, while a semiconductor ETF may carry meaningful Micron exposure. Adding a dedicated memory ETF can deepen the same country and company risks rather than diversify them.

AI can support a multi-year increase in memory and storage demand, but it does not eliminate pricing cycles, capacity expansions or valuation risk. The position size should reflect a category in which two or three companies can determine most of the return and where a supply shortage can eventually become oversupply.