Fund

Fund Philosophy & Process

How PCA SOF thinks and decides. Read after Fund Overview and PCA SOF Investment Thesis.

Investment Philosophy (the beliefs)#

  1. Own the shovel and the gold. In a gold rush with uncertain miners, own the picks (NVIDIA, TSMC, Micron, Disco), the land (Constellation, Vistra), and a few promising miners (the software & consumer layer). → The AI Value Chain
  2. Chokepoints compound. Durable excess returns come from controlling a scarce, hard-to-replicate input, leading-edge fab capacity, CUDA, HBM, advanced-packaging tools, nuclear baseload.
  3. Concentration with conviction. 27 names, not 100. Diversification past the point of understanding is di-worse-ification. The fund would rather deeply understand 27 businesses.
  4. Secular over cyclical, but respect the cycle. Memory, foundry, and power are cyclical inside a secular uptrend. Size them for amplitude; add into capitulation, trim into euphoria.
  5. Moats are the only durable alpha. Every position must survive the question: what stops the customer's in-house team, a hyperscaler, or a low-cost rival from replicating this?
  6. Power is the new oil. Compute is gated by electricity; the fund took direct energy exposure before consensus.
  7. Geography is edge. Non-US winners (TSMC, Tencent, BYD, MercadoLibre, Adyen, BioNTech, Disco, Li Ning) are systematically under-owned by US benchmarks.
  8. Read the chain, not the name. Earnings are analysed for cross-holding read-throughs (a hyperscaler capex beat is a Micron/Marvell/TSMC signal).

Investment Process (idea → position → review)#

1. Sourcing, thematic top-down#

Start from a theme (Artificial IntelligenceAI Power Demand). Ask: where in this theme is value scarce and defensible? Map the value chain; identify the chokepoint at each layer.

2. Underwriting, bottom-up moat & financials#

For each candidate: moat type & durability, unit economics, capital allocation, balance-sheet resilience, management quality, and a 3-statement view of the next 3-5 years. The fund underwrites a base / bull / bear and demands asymmetry (bull-bear skew favourable).

3. Role assignment, portfolio fit#

Before sizing, define the holding's role: Return Driver, Compounder, Hedge/Diversifier, or Optionality (§8 of claude.md). A name with no distinct role is rejected even if cheap.

4. Sizing, conviction × asymmetry ÷ correlation#

Position size rises with conviction, liquidity, and bull/bear asymmetry; it falls with correlation to the existing book. Highly correlated AI names share a risk budget so the portfolio isn't secretly one trade. → Position Sizing Framework

5. Monitoring, the living thesis#

Each holding has explicit thesis-validation signals and sell triggers. Quarterly earnings update 09 Earnings Analysis and feed Cross-Holding Read-Throughs.

6. Sell discipline#

Exit on: moat impairment, theme maturity fully priced, value-destructive capital allocation, or a superior expression of the same theme. The fund does not sell on multiple compression alone if the thesis is intact, but it trims into euphoria.


Risk Management Philosophy#

  • Internal hedges over derivatives. The non-AI ring (payments, healthcare, EM consumer) is the structural hedge against AI-capex beta. → Return Drivers vs Hedges
  • Concentration is managed, not avoided. The fund accepts single-name risk where moats are deepest; it manages it via theme/correlation budgeting, not by over-diversifying.
  • Pre-mortem every position. Every company note carries a Major Risks pre-mortem. The portfolio-level pre-mortem is Risk Master Note.
  • Respect tail risk it cannot diversify. Taiwan Strait Risk is the one exposure the fund cannot hedge internally, it is sized and monitored explicitly.

How The Manager Thinks (mental models)#

  • "Follow the constraint." Whatever is most supply-constrained in the AI build-out has the most pricing power, that's where to be.
  • "Two ways to win." Prefer positions that pay off on the base case and have free optionality (Meta: ads base case + AI/Reality Labs optionality; Alphabet: search base + cloud/Waymo optionality).
  • "The second-order trade." Everyone owns NVIDIA; fewer connect NVIDIA's GPUs to a power shortage and own Constellation Energy. The fund hunts second-order beneficiaries.
  • "Cyclical inside secular." Don't confuse a memory down-cycle with a broken thesis.

Linked Notes#

Back to PCA SOF