This is analysis, not personalized investment advice. Do your own homework before making decisions.
SMH is a conviction play on the chip buildout. Semiconductors power the future economy: AI data centers, autonomous vehicles, cloud infrastructure, edge devices, and high-performance computing. SMH gives you that exposure in one fund, but it does so with real concentration: roughly 27 holdings and Nvidia alone at around 18% to 20%. If you do not have a view on AI infrastructure, you do not need SMH.
- AI infrastructure believers who want chip exposure without picking individual winners
- Investors comfortable with concentration in Nvidia, TSMC, Broadcom, AMD, and equipment makers
- Dare-slot users who want a real equity-based conviction bet instead of crypto or leverage
- Core portfolio buyers who need broad diversification
- Investors who cannot tolerate chip cycles, including 40%+ drawdowns
- Anyone buying only because AI is hot without a holding-period plan
You get focused upside, not safety. SMH can crush broad indexes in a semiconductor bull market, but it can also fall much harder when chip demand slows, inventories build, or Nvidia's valuation resets.
Key metrics
Fund snapshot
| Full name | VanEck Semiconductor ETF |
| Ticker | SMH |
| Exchange | Listed on Nasdaq; charting prefix used here: AMEX:SMH |
| Underlying index | MVIS US Listed Semiconductor 25 Index |
| Expense ratio | 0.35% |
| Total net assets | ~$63.0B (VanEck, May 15, 2026) |
| Inception | December 20, 2011 on VanEck's current fund page |
| Distribution frequency | Annually |
| Sponsor | VanEck |
| Holdings | 27 holdings (VanEck, May 14, 2026) |
| Nvidia weight | 17.61% reported; think roughly 20% |
| 30-day SEC yield | Low income profile; roughly 0.2%–0.3% when available |
| Average daily volume | Roughly 8M–9M shares |
Performance snapshot
| Period | SMH return | Category avg | vs S&P 500 |
|---|---|---|---|
| 1 year | ~126% | Semiconductor category varies widely | Massive outperformance in chip bull cycle |
| 3 year (ann.) | ~65% | Above broad technology category | Outperformed |
| 5 year (ann.) | ~38% | Above broad technology category | Outperformed |
| 10 year (ann.) | ~37% | Above broad technology category | Outperformed |
Performance figures are rounded, time-sensitive, and based on adjusted market-price history through May 15, 2026. SMH's recent numbers are inflated by a powerful AI/chip bull cycle and should not be treated as normal expected returns.
What it is and why it matters
What SMH actually owns
SMH owns a concentrated basket of U.S.-listed semiconductor companies. The top names are the center of the modern compute stack: Nvidia for AI accelerators, Taiwan Semiconductor Manufacturing for foundry capacity, Broadcom and AMD for chip design, Micron for memory, plus equipment and analog leaders such as Lam Research, Texas Instruments, Qualcomm, and Analog Devices.
This is not a broad technology ETF. It does not own every software platform, internet marketplace, or cloud company. It owns the companies that make, design, or equip the chips those platforms need.
Why the AI infrastructure angle matters
AI is not magic floating in the cloud. It is servers, GPUs, memory, networking, cooling, fabs, lithography, packaging, power, and capital spending. When the world builds larger models and more data centers, semiconductor demand becomes the physical bottleneck. SMH is a way to bet that the bottleneck remains valuable.
The fund's Nvidia position is the key reader number: VanEck reported Nvidia at 17.61% on May 14, 2026, so a practical mental model is roughly 20% of SMH is Nvidia. That is why SMH moves like a chip-conviction fund, not like a normal diversified stock ETF.
The semiconductor cycle
Chips are deeply cyclical. Demand surges, companies over-order, capacity gets built, inventories rise, pricing weakens, and the cycle resets. In 2022, SMH fell about 45% from its January high to its October low. That was not a fund failure; that was the semiconductor cycle showing up in the price.
The lesson is simple: SMH can be a great expression of AI infrastructure conviction, but it is not a smooth ride. If a 40% drawdown would make you sell, the position is too large or the thesis is not yours.
Cost analysis
Expense ratio in context
SMH charges 0.35% per year. On a $10,000 position, that is about $35 annually. On $50,000, it is about $175. That is not ultra-cheap like a plain S&P 500 ETF, but it is reasonable for a focused thematic ETF with high liquidity and a long trading history.
The bigger cost is not the 0.35% fee. The bigger cost is concentration risk: if semiconductors go out of favor, the whole fund can fall together. You are paying for a sharp tool. Use it only where a sharp tool belongs.
Tool: Dare Risk Calculator
Use the Dare Risk Calculator to see what a 5% or 10% SMH position would do to your total portfolio if the semiconductor cycle cuts the fund by 30%, 45%, or more.
SMH vs. the competition
SMH vs. SOXX
SMH and SOXX are the two most obvious semiconductor ETF comparisons. SMH tracks the MVIS US Listed Semiconductor 25 Index and holds roughly 27 stocks. SOXX, the iShares Semiconductor ETF, holds roughly 35 stocks. Both are top-heavy, both are direct chip bets, and both can be dominated by the same handful of winners.
The difference is sharpness. SMH is the more concentrated tool and has often had a larger Nvidia exposure. SOXX is still concentrated, but it spreads the bet across a slightly broader semiconductor basket. If you want the cleaner conviction bet, SMH fits. If you want a little more breadth inside the same theme, compare SOXX.
SMH vs. SOXQ
SOXQ is Invesco's PHLX Semiconductor ETF. Its main appeal is cost: SOXQ charges 0.19% versus SMH's 0.35%. That fee gap matters over time, especially if returns cool down from the AI boom years.
SMH's advantage is scale, liquidity, and mindshare. It is much larger and more heavily traded. SOXQ is the cheaper expression of the same broad semiconductor idea, but SMH is the more established Dare-slot vehicle.
SMH vs. QQQ
QQQ is broad Nasdaq-100 exposure. It owns megacap technology, communication, consumer, and growth companies. SMH is a pure semiconductor bet. That difference matters: Nvidia is roughly 20% of SMH, but closer to roughly 10% of QQQ depending on current market weights.
If you want growth with more diversification, QQQ is the broader tool. If your thesis is specifically that chips remain the toll road for AI infrastructure, SMH is the cleaner expression. Do not confuse the two jobs.
| Feature | SMH | SOXX | SOXQ | QQQ |
|---|---|---|---|---|
| Expense ratio | 0.35% | ~0.35% | 0.19% | ~0.20% |
| Holdings | ~27 | ~35 | ~30 | 100+ |
| Nvidia exposure | ~18%–20% | High, but usually different weight | High, but smaller fund | Roughly half of SMH's weight |
| Best for | Focused chip conviction | Slightly broader chip exposure | Lower-cost chip exposure | Broad Nasdaq growth |
Verify current fund data with each sponsor. Holdings and weights move daily.