Risk Model

The AI Infrastructure Portfolio uses a balanced, fundamentals-based risk model focused on diversification, valuation discipline, and exposure management across multiple layers of the AI value chain. Since this is a thematic thesis portfolio, risk is not eliminated but intentionally concentrated around a defined technological trend: AI infrastructure. The goal of the risk model is to quantify, contain, and contextualize that concentration rather than avoid it.

1. Concentration and Sector Risk

The portfolio’s largest source of risk comes from its high weighting in semiconductors and data center infrastructure. Roughly two-thirds of the allocation is tied to the semiconductor industry, which is inherently cyclical and sensitive to capital expenditure trends. To offset this, the remaining holdings in power, cooling, and electrical systems (Vertiv, Eaton) provide countercyclical exposure and steadier cash flows.

2. Valuation Sensitivity

Valuation risk is monitored through forward P/E and price-to-sales multiples relative to historical sector averages. A sustained deviation above the 90th percentile of the past 5 years triggers a review of that holding’s weight. This helps avoid concentration in overextended names while allowing flexibility to maintain exposure to long-term winners like NVIDIA and ASML.

3. Earnings and Cyclical Volatility

Holdings are analyzed for earnings stability using standard deviation of quarterly EPS over the last 12 quarters. High-volatility names (e.g., Micron, Super Micro) are capped at ≤12% allocation to limit drawdown potential, while more stable, cash-generative holdings like Eaton and Broadcom help smooth performance.

4. Geopolitical and Supply Chain Risk

Exposure to Asia-based manufacturing, particularly through TSMC and ASML’s China-related revenue, introduces geopolitical sensitivity. The model accounts for this by pairing fabrication-heavy holdings with U.S.-based infrastructure and networking firms that have diversified manufacturing footprints.

5. Correlation and Cross-Sector Balancing

Historical correlation analysis is used to maintain diversification across sectors. Semiconductor names often exhibit >0.75 correlation with each other, while infrastructure names (Vertiv, Eaton) typically move at <0.45 correlation. Weighting between these groups is calibrated to maintain a portfolio-level beta near 1.2 relative to the S&P 500, offering growth potential without excessive volatility.

6. Risk Monitoring Metrics

• Maximum Position Size: 17% (NVIDIA cap)
• Sector Cap: 65% semiconductors
• Expected Annualized Volatility: ~22%
• Sharpe Ratio (est.): ~1.15
• Portfolio Beta: ~1.2

Summary:
The portfolio’s risk model emphasizes thematic conviction with structural balance. It accepts volatility in exchange for exposure to a high-growth secular trend but offsets it through cross-sector diversification, fundamental screening, and disciplined weighting. This approach maintains the integrity of the AI infrastructure theme while managing downside through liquidity, quality, and diversification buffers.