Want to invest in chips? Watch these 4 ETFs
Semiconductor ETFs seem like a simple way to buy the entire chip industry. But their structures differ significantly: some funds concentrate a large share of their capital in a few giants, while others spread it across dozens of smaller companies. The difference in returns over three years reached more than 20 percentage points annually, and the result is determined mainly by the weighting method, not the fee.
Key points
The semiconductor market is growing at an exceptional pace: global sales in the first eight months of 2026 topped USD 1 trillion for the first time, led by memory chips and chips for artificial intelligence.
Annual fees for the funds covered range from 0.33 to 0.55%, a narrow band, while differences in annual returns are measured in tens of percentage points.
Market-cap-weighted funds have benefited from the dominance of the largest companies over the past three years, while more evenly weighted funds offer different sensitivity to industry developments.
The October sell-off showed that a higher share of smaller companies in a portfolio does not cushion declines, but tends to deepen them.
Buying two market-cap-weighted funds at the same time mostly leads to buying the same stocks again, not to diversification.
Semiconductors in autumn 2026
Demand is driving memory and data centers
According to the Semiconductor Industry Association, global semiconductor sales reached USD 159.7 billion in August 2026 (three-month moving average), up 144.3% from a year earlier. Cumulative sales from January to August exceeded USD 1 trillion for the first time in history, and the SIA expects more than USD 1.6 trillion for the full year. In its August update, WSTS projects a 2026 market of USD 1,655 billion, roughly double the 2025 level.
Growth is unevenly distributed. According to WSTS, memory chips grew 305% year over year in the first half of 2026, while logic circuits grew 45%. The sharp rise in memory is driven by demand for high-bandwidth memory (HBM) for servers with AI accelerators, as well as capacity shortages that are pushing prices up. Part of the growth therefore reflects higher prices, not just larger volumes.
On the logic-chip side, Nvidia remains the main indicator. In the second quarter of fiscal 2027, it reported revenue of USD 96.2 billion, up 106% year over year, of which USD 89.0 billion came from data centers. Management's guidance for the third quarter is USD 108 billion and assumes no revenue from data-center computing chips in China. Moody's estimates capital expenditures by the largest cloud infrastructure operators at USD 785 billion in 2026 and nearly USD 1 trillion in 2027.
Cyclicality beyond AI
Outside AI, the picture is more varied. Sharp increases in memory prices are affecting consumer electronics: according to press reports, IDC expects global smartphone shipments to fall 13% in 2026, the largest decline in its records. Industrial and automotive electronics, by contrast, are recovering from the downturn of 2023 to 2025. Texas Instruments increased revenue 28% year over year to USD 6.06 billion in the third quarter of 2026, and management reported growth across industrial, automotive, and data-center markets. However, its fourth-quarter outlook (USD 5.80 to 6.30 billion) suggests stabilization rather than further acceleration.
Valuation: record profits, even higher expectations
In the week through 16. 10. 2026, the Philadelphia Semiconductor Index lost 11.4% and recorded eight consecutive losing sessions, its longest losing streak since 2008. The triggers were fears of a return of inflation (WTI oil above USD 100 per barrel), a 10-year U.S. Treasury yield of 5.10%, and growing bets on Fed rate hikes. Doubts about returns on AI investments also contributed. Nevertheless, as of that date, the index was still up 57.2% year to date.
As of October 2026, the valuation of large semiconductor portfolios is around 34 to 45 times trailing 12-month earnings. Forward earnings multiples are roughly one-third to one-half lower, at around 20 to 23. The difference between the two figures directly quantifies the earnings growth the market already expects. If the memory cycle turns, or data-center operators slow investment, expected earnings may not materialize, and the low “forward” multiple could prove illusory.
The long-term thesis remains strong: semiconductors are an input for computing power, electrification, and automation. But the industry combines high volatility, rapid technological change, geopolitical concentration of manufacturing in Taiwan, and pronounced inventory cycles. For investors, therefore, the question is not just whether to invest in chips, but also which fund structure to choose.
VanEck Semiconductor ETF
Methodology: concentrated in leaders
SMH tracks the MVIS US Listed Semiconductor 25 Index from MarketVector. It selects the 25 largest and most liquid semiconductor companies listed in the U.S. that generate at least 50% of their revenue from semiconductors or manufacturing equipment. Weights are based on free-float market capitalization, capped at 20% for any one company, and the index is reviewed semiannually. The selection also includes foreign companies listed in the U.S. as American depositary receipts (ADRs), at their full weight.
Portfolio composition
As of 30. 9. 2026, the fund held 26 positions and managed USD 74.9 billion.
Largest positions:
The five largest positions make up 44.29% and the ten largest 67.31% of the portfolio. Geographically, the U.S. accounts for 78.78%, Taiwan 9.24%, the Netherlands 5.48% (ASML), and South Korea 4.40%. The fund therefore combines chip designers (Nvidia, AMD, Broadcom), a contract manufacturer (TSMC), and suppliers of manufacturing equipment (Applied Materials $AMAT, KLA $KLAC, ASML $ASML, Lam Research $LRCX). Based on known weights, manufacturing equipment suppliers account for approximately 17 to 18%, by our estimate.
Performance and costs
According to VanEck's fact sheet as of 30. 9. 2026, the fund returned 86.89% over one year, an annualized 62.00% over three years, and 37.37% over five years, with distributions reinvested. The expense ratio is 0.35%, and the 30-day SEC yield is 0.15%. The portfolio's price-to-earnings ratio was 34.49, its price-to-book ratio was 11.98, and its weighted average market capitalization was USD 1.70 trillion.
Strengths and weaknesses
The main strength is clean exposure to companies that control the most profitable parts of the value chain, as well as exceptional liquidity. The weakness is that one-fifth of the fund is a bet on a single company. In the third quarter of 2026, SMH recorded the smallest decline of the four (-7.34%), so its concentration in the largest companies cushioned losses during this period. The opposite effect would occur if a slowdown in AI investment were first reflected in accelerators.