AI Momentum Continues— Time to Position Semiconductor ETFs?
Yesterday, U.S. tech stocks broadly advanced, with the $纳斯达克(.IXIC)$ rising 1.2%. $英伟达(NVDA)$ gained 4% intraday, while $美光科技(MU)$ surged as much as 7%.
From an ETF perspective, large-scale semiconductor ETFs delivered relatively steady gains, with $半导体指数ETF-HOLDRs(SMH)$ up 1.70% and $iShares费城交易所半导体ETF(SOXX)$ up 1.96%, reflecting capital flowing back into core names. ETFs with more diversified and mid-to-small-cap exposure showed stronger momentum, with $INVESCO SEMICONDUCTORS ETF(PSI)$ up 3.05%, $First Trust Nasdaq Semiconductor ETF(FTXL)$ up 2.01%, and $SPDR S&P Semiconductor ETF(XSD)$ up 1.91%. Leveraged products posted the strongest performance, as $三倍做多半导体ETF-Direxion Daily(SOXL)$ rose 5.86% and $ProShares Ultra Semiconductors(USD)$ gained 3.00%.
On March 16, Nvidia CEO Jensen Huang stated at the GTC conference in San Jose that the company expects AI chips to generate around $1 trillion in revenue by 2027, while also unveiling the next-generation Vera Rubin architecture.
As model sizes continue to grow, bottlenecks in AI systems are shifting. In practice, a significant portion of GPU time is spent waiting on data movement, making memory bandwidth an increasingly critical constraint—often referred to as the “memory wall.” Under this backdrop, high-bandwidth memory (HBM) is becoming more important.
On March 17, Micron announced that its HBM4, along with SOCAMM2 and PCIe 6.0 SSD solutions designed for Nvidia’s Vera Rubin platform, has entered mass production and will begin shipping in the first quarter of 2026. This indicates Micron has moved into actual supply for Nvidia’s next-generation platform.
From a technical standpoint, Micron’s HBM4 delivers bandwidth of 2.8TB/s—more than double that of HBM3E—while supporting higher-density stacking. These improvements directly enhance data access efficiency during AI inference, and demand for HBM is expected to increase as models scale further.
In terms of valuation, $台积电(TSM)$ trades at a price-to-book ratio of about 8.95, while $美光科技(MU)$ stands at 8.46, bringing the two close. Micron’s re-rating reflects its growing role in AI-related memory demand, particularly through its participation in HBM.
Korean enterprises are also part of the HBM supply chain. SK Hynix trades at a PB of 5.66, while Samsung Electronics is at 2.67. Both operate on the supply side of memory, and their recent gains largely reflect the broader industry trend of tight chip supply.
Over the long term, AI demand remains strong, while supply of advanced memory like HBM is difficult to ramp quickly, keeping memory prices elevated. As Nvidia continues to push new platforms forward, increasing compute power will also drive higher requirements for memory capacity and bandwidth.
ETF Highlights:
$半导体指数ETF-HOLDRs(SMH)$ is the largest semiconductor ETF, with about $44 billion in assets and a 0.35% expense ratio. Its top holdings—Nvidia 18.92%, TSMC 10.72%, Broadcom 7.24%, and Micron 7.15%—account for nearly 44% combined, showing a strong concentration in AI leaders.
$iShares费城交易所半导体ETF(SOXX)$ manages around $20 billion with a 0.34% expense ratio. Its top two holdings are Micron at 9.64% and Nvidia at 7.25%, with a more balanced allocation compared to SMH.
$INVESCO SEMICONDUCTORS ETF(PSI)$ has approximately $1.2 billion in assets and a 0.56% expense ratio. It uses a quantitative approach and tilts toward mid- and small-cap stocks, resulting in higher volatility and stronger upside during rallies.
$First Trust Nasdaq Semiconductor ETF(FTXL)$ manages about $1.5 billion with a 0.60% expense ratio. It employs a multi-factor strategy, balancing profitability and growth characteristics.
$SPDR S&P Semiconductor ETF(XSD)$ has around $1.4 billion in assets and a 0.35% expense ratio. Its equal-weight methodology avoids overconcentration in large-cap names and offers broader exposure across the sector.
$三倍做多半导体ETF-Direxion Daily(SOXL)$ and $ProShares Ultra Semiconductors(USD)$ , with assets of roughly $12.5 billion and $1.6 billion respectively, both carry a 0.75% expense ratio. These leveraged ETFs provide 3x and 2x daily exposure and are more suitable for short-term trading.
Disclaimer: Investing carries risk. This is not financial advice. The above content should not be regarded as an offer, recommendation, or solicitation on acquiring or disposing of any financial products, any associated discussions, comments, or posts by author or other users should not be considered as such either. It is solely for general information purpose only, which does not consider your own investment objectives, financial situations or needs. TTM assumes no responsibility or warranty for the accuracy and completeness of the information, investors should do their own research and may seek professional advice before investing.
