If there is one exchange-traded fund, or ETF, that has become the poster child of the AI-driven semiconductor boom in 2026, it is the Roundhill Memory ETF (ticker: DRAM).
Since launching at the start of April 2026, DRAM has seen explosive asset growth and, according to Reuters, has become the fastest-growing ETF launch in history, surpassing even the pace previously set by the iShares Bitcoin Trust ETF (IBIT). Currently, DRAM has just under $9.7 billion in assets.
[Sign up for stock news with our Invested newsletter.]
Investors searching for bottlenecks within the AI infrastructure buildout have increasingly turned toward semiconductors, with earlier attention focused on chip fabrication plants, hyperscaler spending, data centers and even electricity demand from utilities.
One area that remained relatively overlooked until recently was memory chips. These include DRAM, or dynamic random-access memory, which temporarily stores active data for rapid processing, and NAND flash memory, which provides longer-term storage used in servers and smartphones. AI workloads require enormous amounts of high-bandwidth memory, making these chips critical.
However, exposure to this niche is concentrated among only a handful of companies. Some of these firms, such as Micron Technology Inc. (MU), are U.S. listed. Others, however, are foreign listed, such as South Korean company SK Hynix Inc. (000660.KS).
For many U.S. investors, directly accessing international shares can introduce added brokerage limitations, currency conversion costs or liquidity concerns. DRAM circumvented these drawbacks by packaging all three major memory players into a single U.S.-listed ETF, where they now collectively account for nearly 75% of portfolio assets.
Part of DRAM’s success stems from its active structure. Rather than being tied to a rigid semiconductor index, the fund relies on portfolio managers and research analysts to determine holdings and portfolio weights, which in this case aligned very well with the AI memory trade at precisely the right time.
Moreover, DRAM itself remains highly liquid. The ETF currently maintains a very tight 0.03% 30-day median bid-ask spread, allowing investors to trade it much like a normal U.S.-listed stock.
Still, investors should understand that DRAM may not necessarily be the ideal long-term semiconductor holding. The ETF charges a relatively high 0.65% expense ratio, which is common for active thematic ETFs but materially more expensive than many index-based semiconductor funds.
DRAM is also extremely concentrated, with only 15 holdings and roughly three-quarters of assets allocated to a handful of companies. That concentration can amplify returns during bull markets, but semiconductors remain cyclical, meaning it can also magnify downside risk during industry slowdowns.
Here are seven of the best semiconductor ETFs to buy in 2026:
| ETF | Expense Ratio |
| iShares Semiconductor ETF (SOXX) | 0.34% |
| VanEck Semiconductor ETF (SMH) | 0.35% |
| VanEck Fabless Semiconductor ETF (SMHX) | 0.35% |
| Global X AI Semiconductor & Quantum ETF (CHPX) | 0.50% |
| Invesco PHLX Semiconductor ETF (SOXQ) | 0.19% |
| Invesco Semiconductors ETF (PSI) | 0.56% |
| SPDR S&P Semiconductor ETF (XSD) | 0.35% |
iShares Semiconductor ETF (SOXX)
“The potential benefits of investing in semiconductor ETFs include exposure to a high-growth industry with strong fundamentals, diversification across multiple companies in the industry and the potential for long-term capital appreciation,” says Sean August, CEO of the August Wealth Management Group. SOXX is a popular option with over $33 billion in assets at a reasonable 0.34% expense ratio.
This ETF tracks 30 U.S. listed semiconductor firms represented by the NYSE Semiconductor Index. Top holdings for SOXX currently include Micron, Broadcom Inc. (AVGO), Nvidia Corp. (NVDA), Intel Corp. (INTC) and Advanced Micro Devices Inc. (AMD). SOXX maintains excellent liquidity with a low 0.02% 30-day median bid-ask spread and traded an average of 7.4 million shares over the same period.
VanEck Semiconductor ETF (SMH)
“Near-term, the backdrop for semiconductors remains supportive as hyperscalers continue to spend heavily on AI infrastructure, and demand stays firm across compute, memory and networking enablers,” explains Nick Frasse, product manager at VanEck. “Long term, we think the opportunity expands further as AI adoption spreads from data centers into enterprise use cases like edge computing and robotics.”
SMH is currently the largest pure-play semiconductor ETF with $63 billion in assets. It passively tracks the MVIS U.S. Listed Semiconductor 25 Index, which favors the largest and most liquid companies. As a result of Nvidia’s historical run-up, the stock is now SMH’s largest holding at a 17.5% weight, followed by Taiwan Semiconductor Manufacturing Co. Ltd. (TSM) at 9.7%. SMH charges 0.35% in fees.
VanEck Fabless Semiconductor ETF (SMHX)
“We believe SMHX is well positioned for an environment where design, architecture and intellectual property matter more and more,” Frasse says. “As advanced compute becomes harder to scale efficiently, the ability to deliver better performance per watt and per dollar becomes increasingly valuable, which we see as a key advantage for fabless semiconductor companies moving forward.”
One reason investors may prefer a fabless semiconductor strategy is to reduce exposure to manufacturing and geopolitical risks tied to fabrication plants. For example, much of the global semiconductor supply chain still depends heavily on TSM. Instead of owning the manufacturers themselves, SMHX focuses on innovative chip designers for a 0.35% expense ratio.
[Read: 5 Cloud Computing ETFs to Buy Now.]
Global X AI Semiconductor & Quantum ETF (CHPX)
“The semiconductor shift is most visible on three fronts: the transition from general-purpose processors to AI-optimized chips, the use of high-bandwidth memory to handle AI’s data intensity and the rise of ultra-fast interconnect solutions that bind AI servers together,” says Tejas Dessai, director of thematic research at Global X ETFs. Alongside its AI thematic ETFs, Global X also offers CHPX for semiconductors.
“CHPX provides exposure to companies across the entire global compute stack, from AI semiconductors to data center equipment, power infrastructure and quantum technologies,” Dessai says. Unlike SOXX and SMH, the top two holdings in CHPX are foreign companies, with an 8.7% allocation to ASML Holding NV (ASML) and 8.6% to TSM. However, CHPX is less affordable with a 0.5% expense ratio.
Invesco PHLX Semiconductor ETF (SOXQ)
“AI remains the dominant structural driver across the semiconductor value chain,” says Rene Reyna, head of thematic and specialty product strategy at Invesco. “Data center buildouts are driving demand for numerous chip categories such as graphics processing units (GPUs), central processing units (CPUs), networking, memory, power management and manufacturing equipment.”
SOXQ is one of Invesco’s two semiconductor-themed ETFs. It passively tracks the PHLX Semiconductor Index, a benchmark of the 30 largest U.S.-listed semiconductor companies. While SOXQ has a significant overlap with SMH and SOXX, it undercuts both in terms of fees, at a 0.19% expense ratio. For a $10,000 investment in SOXQ, that implies just $19 a year in fee drag versus 0.35% and 0.34% for SMH and SOXX, respectively.
Invesco Semiconductors ETF (PSI)
“Memory has become a major upside driver for semiconductors because AI servers require high-bandwidth memory and advanced DRAM, creating a pricing and supply squeeze,” Reyna explains. “This can support pricing power and capital spending across the semiconductor ecosystem, benefiting chip designers, memory suppliers, foundries and equipment companies in particular.”
PSI is Invesco’s other semiconductor ETF. It tracks a more sophisticated benchmark, the Dynamic Semiconductor Intellidex Index. The 30 companies in PSI have been screened for price momentum, earnings momentum, quality, management action and value, making it more curated than a market-cap weighted semiconductor ETF. However, PSI is also pricier with a 0.56% expense ratio.
SPDR S&P Semiconductor ETF (XSD)
“When looking for semiconductor ETFs, investors should consider factors such as the expense ratio, the underlying index or benchmark, the fund’s holdings and diversification strategy, and the ETF’s historical performance,” August says. “It is also important to assess the fund’s liquidity to ensure that it is easy to buy and sell.” This is important because not all semiconductor ETFs select and weight stocks the same.
XSD is a good example. This ETF tracks the S&P Semiconductor Select Industry Index. Every quarter, the 40-plus stocks in XSD are systematically reset back to equal weights, producing a natural “buy low, sell high” effect along with a stronger tilt toward small- and mid-cap stocks. This can decrease concentration risk, but the drawback is capped upside if large-cap semiconductor stocks continue to outperform.
More from U.S. News
7 Best Space Stocks and ETFs to Buy in 2026
Can AI Pick Stocks? 5 AI Investing Apps to Try
7 Best Semiconductor ETFs to Buy for 2026 originally appeared on usnews.com
Update 05/18/26: This story was published at an earlier date and has been updated with new information.