Artificial intelligence chips need to process massive amounts of data quickly -- and to access it in an optimal way, they need that data stored nearby on high bandwidth memory (HBM) chips.
For example, Nvidia's (NVDA -1.01%) new Rubin GPU is designed to use as much as 288 gigabytes of HBM4, the fourth generation of high bandwidth memory. That's nearly three times the memory used in Blackwell, its predecessor architecture. And rival Advanced Micro Devices' (AMD -3.54%) MI400 series accelerators are designed to use up to 432 gigabytes of HBM4.
Image source: Getty Images.
Because of these growing needs, there are shortages across the entire memory market, and rectifying that will require investment across the semiconductor supply chain. The iShares Semiconductor ETF (SOXX -4.40%) gives investors exposure to Micron Technology (MU -7.24%), the chip designers driving HBM demand, and the equipment suppliers helping manufacturers expand production.
Why the memory supply remains tight
Increasing memory supply will require the few major players in that niche to build large and complex new chip foundries -- and they are doing so. However, the pace of their expansion is constrained by long construction times, shortages of skilled workers, permitting requirements, and the need for additional energy infrastructure. Micron management says it expects supplies of DRAM and NAND to remain tight beyond 2027.
For now, growth in production of HBM can tighten the broader memory market. Each new HBM generation requires more wafer supply compared to conventional memory, putting additional pressure on the supply available for non-HBM. Expanding output requires memory producers to invest in more advanced manufacturing and packaging equipment.
NASDAQ: SOXX
Key Data Points
These trends can benefit the components of the iShares Semiconductor ETF, which has nearly 21.4% of its assets in semiconductor manufacturing equipment makers such as Lam Research, Applied Materials, and KLA. Lam Research supplies etching and deposition tools used in advanced memory production. Applied Materials provides equipment for DRAM manufacturing and HBM packaging. KLA's inspection tools help chipmakers detect defects and improve production yields.
Capturing the broader memory opportunity
The iShares Semiconductor ETF also provides exposure to the companies producing memory and driving its demand. The fund owns 30 stocks and charges a 0.34% expense ratio.
Micron accounted for 8.33% of the ETF's value as of July 21, giving investors direct exposure to the HBM, DRAM, and NAND memory markets. Together, Nvidia and Advanced Micro Devices accounted for another 17% of the portfolio, and they are increasing the amount of memory they embed in their AI accelerators. The ETF's other major holdings, including Broadcom and Taiwan Semiconductor Manufacturing, provide wider exposure to the AI chip ecosystem.
However, iShares Semiconductor ETF is far from a pure-play memory fund. It does not hold stakes in the other two major HBM suppliers, South Korea-based SK Hynix and Samsung Electronics. The ETF's top 10 holdings accounted for about 60.8% of assets as of July 21, making it considerably more concentrated than a broad market ETF.
Is SOXX a buy right now?
The iShares Semiconductor ETF closed at $552.69 on July 21. Investors with access to fractional ETF trading could buy roughly 0.18 shares with $100.
The recent semiconductor sell-off has improved the entry point, as the ETF now trades about 16% below its June 22 peak. However, the ETF remains up sharply for 2026, and traded at more than 66 times earnings as of July 21. That premium valuation increases investors' downside risk if AI spending slows, memory prices weaken, or new memory supply reaches the market faster than expected.
Despite those risks, iShares Semiconductor ETF offers a diversified way to invest across the AI memory supply chain rather than betting on a single memory producer. Investors who can tolerate semiconductor volatility and plan to hold for several years can consider putting $100 (or more) into this ETF.
