Why Semiconductor ETFs Miss the Real Story
The most popular semiconductor ETFs — VanEck SMH and iShares SOXX — are marketed as broad semiconductor exposure. In reality, they are heavily concentrated in 4-5 mega-caps, with NVIDIA alone representing roughly 25% of SMH. This cap-weighting structure means they track NVIDIA's stock price, not the health of the semiconductor supply chain. Critical bottleneck companies in packaging, testing, and thermal management have negligible weight despite being irreplaceable chokepoints. For investors seeking genuine AI infrastructure exposure, cap-weighted ETFs are the wrong instrument.
The Concentration Problem
SMH holds 25 semiconductor companies, but the top 5 (NVIDIA, TSMC, Broadcom, ASML, Qualcomm) represent over 50% of the fund. In practice, buying SMH is a leveraged bet on NVIDIA with some diversification noise around it. When NVIDIA drops 15% on an earnings miss, SMH drops 4% even if every other semiconductor company reported strong results.
SOXX uses a modified equal-weight approach that's slightly better, but still overweights mega-caps and completely excludes many critical supply chain companies that are too small for the index.
What They Miss
The AI build-out depends on dozens of companies that ETFs either exclude or underweight:
- BESI (€5B) — the leading hybrid bonding equipment maker, critical for advanced packaging. Not in SMH.
- Advantest (¥4T) — dominates chip testing equipment. Tiny weight in SOXX.
- Modine ($4B) — thermal management for AI data centers. Not in any semiconductor ETF.
- DISCO (¥2T) — precision wafer dicing, irreplaceable process step. Not in SMH.
- Kulicke & Soffa ($3B) — wire bonding and advanced packaging. Not in SMH.
These companies sit at constraint points in the supply chain where demand exceeds supply — precisely where the best risk-adjusted returns exist.
The Alternative: Functional Indexing
Closelook's Functional Index addresses this by weighting constituents by their supply chain function rather than market capitalization. In equal-weight mode, BESI has the same weight as NVIDIA. This means the index reflects the structural health of the entire AI supply chain, not just the performance of the biggest companies.
The practical implication: investors can use SMH/SOXX for core semiconductor exposure, but should overlay the Functional Index to detect divergences, identify bottleneck opportunities, and monitor supply chain stress that cap-weighted indices miss.