Return Drivers vs Hedges
Which holdings are the offense, which are the defense, and how the fund hedges internally (without derivatives). Companion to Portfolio Construction Overview.
The Core Idea#
PCA SOF hedges its dominant factor, AI capex beta: not with options or shorts, but by owning uncorrelated secular growth. The non-AI ring is the structural hedge. This is the "internal hedging" philosophy from Fund Philosophy & Process.
Return Drivers (the offense)#
| Holding | Driver | Beta to | |, -|, -|, -| | NVIDIA | AI compute demand | AI capex (high) | | Micron Technology | HBM scarcity | AI capex + memory cycle (very high amplitude) | | TSMC | Leading-edge demand | AI capex + semi cycle | | Marvell Technology | AI networking + ASIC | AI capex | | Disco Corporation | Packaging/HBM tools | AI capex + semicap cycle | | Microsoft/Amazon/Alphabet/Meta Platforms | Cloud + AI monetisation | AI capex + ad/IT cycle | | Constellation Energy/Vistra | Datacentre power demand | AI capex via power markets (different transmission) | | ServiceNow/Salesforce/Snowflake/Datadog/CrowdStrike/Zscaler | Enterprise AI software spend | IT budgets + AI adoption |
These ~18 names rise + fall substantially together with the AI capital cycle. That correlation is the portfolio's biggest single risk, see Overlap & Diversification Map.
The Internal Hedges (the defense)#
| Holding | Independent driver | Why it's a hedge | |, -|, -|, -| | Adyen | Global commerce volume | Payments grow with GDP/commerce, not AI capex. | | PayPal | Commerce + self-help/buybacks | Value/turnaround; FCF-yield support. | | Tencent | China internet + policy | China-macro + value; low US-AI correlation. | | MercadoLibre | LatAm digitisation | EM consumer + fintech; geographic offset. | | Li Ning | China discretionary spend | Pure China-consumer; non-tech. | | BYD | EV adoption S-curve | Electrification; physical-economy. | | Netflix | Subscription media | Consumer-defensive; ad-tier optional. | | BioNTech | Clinical data (binary) | Truly uncorrelated; cash floor. | | IQVIA | Biopharma R&D spend | Healthcare cycle; wide-moat data. |
These 9 names are the shock absorber. If AI capex disappoints, this ring is what the fund leans on.
Power Names, A Special Case#
Constellation Energy + Vistra are return drivers tied to AI demand but with a different transmission mechanism (power prices + capacity markets + regulation, not chip cycles). They partially hedge silicon-cycle timing risk: a memory down-cycle doesn't necessarily hit power-market tightness. So they sit between "driver" + "diversifier."
The Diversification Honest-Truth#
- Effective hedge weight = Ring 3 (9 names) + the power transmission difference. If these are ~20-30% of the book, the fund retains meaningful AI-capex beta.
- The hedges are fundamental, not market-neutral: in a broad risk-off (rates spike, recession), correlations rise + even the hedges fall, just less. → Interest Rate Sensitivity, Risk Master Note.
- The cleanest diversifiers are BioNTech (binary/clinical) + the China cluster (Tencent/Li Ning/BYD, China-macro, not US-AI).
Linked Notes#
- Portfolio Construction Overview · Core vs Satellite Holdings · Overlap & Diversification Map · Position Sizing Framework
- Risk Master Note · AI Capex Cycle Risk · Macro Master Note · Knowledge Graph