AI & Semiconductors
A concentrated bet on the AI infrastructure buildout. Semis are the picks-and-shovels of the AI revolution. High conviction, high volatility — this is not for the faint of heart.
Holdings
| Symbol | Name | Weight | Price | 1D | 3M | YTD | Yield | AUM |
|---|---|---|---|---|---|---|---|---|
| SMH | VanEck Semiconductor ETF | 35% | $436.88 | ... | ... | ... | 0.3% | $47.6B |
| QQQ | Invesco QQQ Trust, Series 1 | 30% | $610.96 | ... | ... | ... | 0.5% | $396.3B |
| SOXX | iShares Semiconductor ETF | 20% | $386.21 | ... | ... | ... | 0.4% | $24.3B |
| IGV | iShares Expanded Tech-Software Sector ETF | 15% | $74.61 | ... | ... | ... | — | $9.8B |
Investment Thesis
Every AI model needs chips to train and run inference. NVIDIA alone has seen revenues explode from $27B to over $100B in two years, and the broader semiconductor industry is entering a supercycle driven by data center buildouts, edge AI, and the automotive transition. SMH captures the leaders — NVIDIA, TSMC, Broadcom, ASML — while SOXX provides broader diversification across the semiconductor value chain. QQQ adds the tech companies actually deploying AI (Microsoft, Google, Amazon, Meta), and IGV captures the software layer that will monetize AI through SaaS products. The risk is significant: semiconductor stocks are cyclical, and if AI spending disappoints or the cycle peaks, these stocks could fall 40-50%. But the structural demand thesis — that computing power needs will grow exponentially for decades — is compelling.
Portfolio Construction
Key Considerations
- Extremely concentrated in one theme — if AI investment slows, losses will be severe
- Semiconductor stocks are cyclical and can drop 40%+ in downturns
- Significant overlap between SMH, SOXX, and QQQ in top holdings like NVIDIA
- Valuations are stretched after the 2023-2025 AI rally