SMH

SMH holds the compute thesis without requiring a choice of a single winner. Design, manufacturing and equipment are different businesses with different economics, and the view here is about compute demand rather than about which layer of the stack captures it. There is a tension worth naming. This portfolio moved away from index products on purpose, so holding a fund needs a reason beyond convenience. The reason is that semiconductor outcomes turn on process nodes, customer concentration and capital cycles I do not think I can handicap reliably at the single-name level. Where I cannot rank the companies, I would rather own the demand than guess the winner.

The fund is concentrated rather than diversified. A handful of the largest semiconductor companies dominate its weighting, so it behaves more like a leveraged position on a few names than a broad sector holding. Buying the basket avoids single-name selection risk. It does not avoid concentration.

Semiconductors are also cyclical. The industry has always alternated between shortage and glut, and an AI-driven cycle is not exempt from that pattern.