Why Semiconductor Investors Are Rotating From SMH’s Nvidia Concentration to PSI’s Equal-Weight Approach
One rival semiconductor ETF has quietly lapped the most popular chip fund on the market over the past year, and the reason comes down to a single position that most investors never think to question.
This post may contain links from our sponsors and affiliates, and Flywheel Publishing may receive compensation for actions taken through them.
If you own the VanEck Semiconductor ETF (NASDAQ:SMH), you own the default way to bet on chips. The fund is the largest and most liquid semiconductor ETF on the market, sitting at roughly $71.1 billion in assets with a slim 0.35% expense ratio, and it delivers exactly what most investors want: concentrated exposure to the names driving the AI buildout. That is also SMH’s problem. NVIDIA (NASDAQ:NVDA | NVDA Price Prediction) now accounts for 21.7% of the portfolio, and over the past year an equal-weight-leaning competitor has quietly outrun it by more than 20 percentage points.
Why SMH’s Design Became a Headwind
SMH holds only about 25 stocks and weights them by modified market cap, which pushes megacap winners to the top and keeps them there. Beyond Nvidia, the fund’s next tier includes Taiwan Semiconductor at 9.29%, Applied Materials at 5.74%, AMD at 5.63%, and Micron at 5.67%. The top ten holdings account for the majority of the fund’s assets.
That structure worked brilliantly when Nvidia was compounding at triple-digit rates, but it has worked less well over the past twelve months. Nvidia returned 26.55% in the year ending September 23, 2026, a fine result, but still trailed the rest of the sector. Memory names, semiconductor equipment stocks, and analog chipmakers largely performed better. That said, because SMH’s largest single position was also its slowest large mover, the fund returned approximately 87.26% over the same window.
PSI: Same Sector, Less Nvidia, More Breadth
The Invesco Semiconductors ETF (NYSEARCA:PSI) tracks a modified equal-dollar-weighted index of about 33 positions. In its most recent full NPORT filing, Nvidia sat at just 3.91% of net assets, well behind top holdings like MaxLinear at 7.98%, AMD at 6.26%, and Texas Instruments, Broadcom, Micron, KLA, Lam Research, and Applied Materials, each between 3.9% and 5%.
That means when the broader semi complex rallies rather than just the AI megacap, PSI captures more of the upside per dollar invested. Over the past year, PSI returned 111.7%, roughly 107.45% on a price-only basis. Against SMH’s 89.3%, that is a real spread, and it came without leverage, without a thematic tilt, and without straying outside the same industry.
The mechanism is straightforward. When your top position is ~22% of the fund and it advances 27%, you need everything else to fire hard just to match a broad-based rally. PSI does not have that anchor. Its largest position is under 8%, and the fund holds roughly 33 chip names spanning designers, equipment makers, memory, and analog. Breadth did the work.
Where SOXX Fits
If you want a middle ground, the iShares Semiconductor ETF (NASDAQ:SOXX) tracks the ICE Semiconductor Index with a cap at roughly 8% per name. Nvidia was 6.81% of SOXX in its June 30 filing, and the fund delivered a one-year total return of 110.52%. It carries the cheapest fee of the three at 0.33%, which matters if you plan on holding long term.
What the Swap Actually Costs
PSI’s expense ratio is 0.55%, meaningfully higher than SMH’s 0.35% or SOXX’s 0.33%. On a $10,000 position, that is roughly $20 more per year versus SMH—trivial next to a 20-point performance gap, but it will compound in flat markets. PSI is also smaller, at about $2.0 billion in net assets as of its April NPORT versus SMH’s $77.2 billion. Bid-ask spreads are wider, and PSI behaves less like a broad index proxy and more like a small- and mid-cap chip basket. When Nvidia leads, PSI will trail.
Selling SMH in a taxable account after a run this size will trigger capital gains. If your position lives in an IRA or 401(k), the switch is frictionless. In a brokerage account, a partial reallocation—trimming SMH and adding PSI or SOXX rather than swapping fully—preserves the AI-megacap exposure while diluting the single-name concentration.
How to Think About the Trade
SMH remains a concentrated bet on the largest chipmakers, and if Nvidia reaccelerates from its current 27% one-year pace, SMH will look right again. The question is whether you want your semiconductor allocation to rise and fall with one stock or with 30. If you already own Nvidia directly, PSI or SOXX gives you the rest of the industry without doubling down. If you do not, splitting your allocation between SMH and PSI captures both the megacap engine and the broader base at a blended fee under 0.45%. And if you want exposure to the AI buildout without piling further into chipmakers at all, we profiled seven suppliers powering the data-center side of the trade in a free report you can grab here. The past year rewarded breadth. The next year may not, but knowing which bet you are making is the point.
Contact [email protected] for any questions or corrections.
Ryne Mauck is an investment writer covering exchange-traded funds, retirement planning, and portfolio strategy. Through his work at 24/7 Wall St. and other investment platforms, including Seeking Alpha, he aims to provide clear, research-driven insights that help investors make more informed decisions while maintaining a long-term approach to investing.
Ryne holds a B.Sc. in Finance and an M.A. in Political Science. He is a formerly registered Municipal Advisor Representative and has passed the Series 50, Series 63, and Series 65 exams. His articles are not intended to be, nor should they be interpreted as, financial advice.
