US AI & Semiconductor ETFs Complete Guide (2026) — SOXX vs SMH vs SOXQ vs AIQ vs BOTZ

Quick answer (June 2026): Of the five most-traded US AI & semiconductor ETFs, SOXQ leads 2026 YTD at roughly +96.7% thanks to higher memory and networking-chip weight, SMH (Top 10 ≈ 73%) is the most concentrated NVIDIA + TSMC bet, SOXX is the most diversified semi-only fund (Top 10 = 57.4%), AIQ spreads across 86 global AI names at 0.68% TER, and BOTZ is really an industrial-automation fund disguised as AI.

1. Five ETFs at a glance

There are over 100 US AI & semi ETFs, but only five matter for most retail investors. Here is the snapshot as of mid-2026.

TickerIssuerInceptionExpense ratioAUMHoldingsProfile
SOXXiShares (BlackRock)20010.35%~$33-40B34Broad semi, passive
SMHVanEck2011 (re-listed)0.35%$63.04B25Mega-cap concentrated
SOXQInvesco20210.19%~$1.2B30Low-cost PHLX index
AIQGlobal X (Mirae)20180.68%$9.76B86Broad AI + Big Data
BOTZGlobal X (Mirae)20160.68%$3.36B~45Robotics & industrial automation

Source: iShares, VanEck, Invesco, Global X official fact sheets (March 31 – May 29, 2026).

2. The expense-ratio trap — what you actually pay

US ETFs disclose a single “total expense ratio,” but real cost includes the bid-ask spread and tracking difference. Once you add those in:

TickerHeadline ERAvg spreadTracking diff (12M)All-in
SOXX0.35%0.02%±0.05%~0.37%
SMH0.35%0.01%-0.03% / +0.10%~0.36%
SOXQ0.19%0.04%±0.05%~0.23%
AIQ0.68%0.05%±0.10%~0.73%
BOTZ0.68%0.08%±0.15%~0.76%

On a $100,000 ten-year hold, SOXQ costs roughly $2,300 while AIQ/BOTZ cost ~$7,300 — a $5,000 gap that compounds.

3. Concentration — why “semiconductor ETF” is misleading

This is the most important table. SOXX, SMH and SOXQ all cover US-listed semis, but the weighting rules differ.

RankSMH (Top 5)SOXQ (Top 5)SOXX (Top 5, May 29)
1NVDA 19.35%NVDA 12.56%AVGO 6.90%
2TSM 11.70%AVGO 10.30%NVDA 5.88%
3AVGO 7.72%MU 6.97%AMD ~5.0%
4ASML 5.02%MRVL 5.16%AMAT ~4.8%
5AMD 4.73%AMD 4.32%MU ~4.5%
Top 10 total~73%~59%57.4%

SMH is essentially “buy NVIDIA at 19% plus TSMC at 11.7%.” SOXX spreads the load across NVIDIA, AVGO, AMD, AMAT, MU, LRCX, KLAC and TXN — equipment (AMAT/LRCX/KLAC) and memory (MU) carry roughly double the weight versus SMH. SOXQ sits between them.

AIQ is a different animal — 86 holdings, Top 10 only 35.51%, with NVIDIA, Microsoft, Alphabet, Meta, Apple and Amazon at 4-7% each, plus Samsung, TSMC, Toyota and Sony in the international sleeve (71% US allocation per Global X). BOTZ leads with ABB 8.8%, NVDA 8.5%, Keyence 7.9%, Fanuc 7.7% and Intuitive Surgical 7.3% — industrial 45%, IT 33%, healthcare 11%. It is closer to an industrial-automation fund than to a semi fund.

4. Returns — a 30-point spread on the same theme

Ticker2026 YTD (~Jun 6)1-year3-year CAGR5-year CAGR
SOXQ+96.7%+165%+59.4%n/a (2021 inception)
SMH+77.1%+138%+64.2%+33%
SOXX~+79.5%~+140%+50%+30%
AIQ~+35%+28%+30%+18%
BOTZ~+28%+22%+12%+5%

As of 2026-06-06. SMH/SOXQ YTD: 24/7 Wall St., 2026-06-04. 3- and 5-year figures from etf.com and Morningstar.

The reason SOXQ outran SMH in 2026 is simple: Micron and Marvell carry roughly double the weight versus SMH. As AI data-center spending rotated from GPUs into HBM memory and Ethernet switching silicon in early 2026, those names ran hardest and SOXQ absorbed it most directly.

5. Risks

  • Volatility: Semi ETFs have 35-45% annualized standard deviation — more than double the S&P 500. SOXX was -36% and BOTZ -42% in 2022.
  • Single-stock risk: An NVIDIA 30% drawdown alone takes SMH down 5.7%.
  • FX: All five are USD-denominated. No hedged share classes.
  • Sector clustering: AIQ and BOTZ look diversified by name count but co-move tightly.
  • Geopolitics: US export controls and Taiwan-Strait risk hit every semi ETF simultaneously.
  • Liquidity: SOXX, SMH, SOXQ and AIQ trade hundreds of millions per day. BOTZ is thinner ($50-80M/day).

6. How they compare to Korean semi ETFs

Domestic Korean AI/semi ETFs (KODEX AI반도체, TIGER AI반도체핵심소재) are effectively Samsung + SK Hynix bets — 30-40% in two names. US ETFs are NVIDIA-, AVGO-, TSM- centric, so AI exposure is more direct but FX risk is real for Korean investors. We will publish a side-by-side KR vs US comparison in the next post in this series.

The same pattern showed up in our quantum-ETF series: US quantum ETFs ran +120% while Korean ones did about +60% — US themes tend to outperform their Korean counterparts in the same theme by a wide margin during super-cycle phases.

7. Which one fits which investor?

  • Aggressive (want NVIDIA beta): SMH. 19% NVDA + 11.7% TSM captures the AI alpha most directly.
  • Balanced (diversify but keep fees low): SOXQ. 0.19% ER, NVDA + AVGO + MU + MRVL + AMD diversification, leading 2026 YTD.
  • Defensive (afraid of single-stock risk): SOXX. 34 names, Top 10 at 57.4%, longest track record (2001 inception).
  • Broad AI exposure (want big tech too): AIQ — but the 0.68% fee is the cost of admission.
  • Robotics & automation focus: BOTZ as a satellite, not a core position.

After 1-year returns of +130-165%, dollar-cost-averaging over 3-6 months is more prudent than a lump-sum buy at current levels.

8. FAQ

Q1. SMH or SOXX for a single pick? Over the last decade SMH beat SOXX by roughly 3.5%/yr, but that was NVIDIA + TSMC alpha. If AI leadership broadens, SOXX should narrow the gap.

Q2. Why did SOXQ outperform SMH in 2026? Memory and networking-chip weight is roughly double SMH’s, so the 2026 HBM/Ethernet rally hit SOXQ hardest.

Q3. Tax for Korean investors? Same as other US equities — KRW 2.5M annual exemption then 22% capital-gains tax; dividends 15% withholding. Separate from income tax.

9. Related

Source: https://www.ishares.com/us/products/239705/ishares-semiconductor-etf

Leave a Comment