Risk
Thematic Exposure
Sector and bucket labels can hide a risk that cuts across both: several holdings can carry completely different sector tags and bucket assignments while still depending on the same underlying bet. This page is that single-theme lens, computed directly from the live Holdings weights.
The theme: AI infrastructure, compute & power
Semiconductors and memory, enterprise software tied to AI capex, AI data-center and compute infrastructure, and the power generation that feeds it are, functionally, one bet: that AI-driven compute and power demand keeps growing. Sixteen of my 32 positions and one leveraged ETF touch this theme in some form. The IPS requires flagging single-theme clustering when it appears — it doesn't cap it, because a written cap here would be arbitrary rather than principled, but disclosure is the discipline in its place.
Every position in the theme
| Ticker | Sub-theme | % of book |
|---|---|---|
| SNDK | Semiconductors / memory tied to AI capex | 10.43% |
| META | Hyperscaler AI capex spender | 8.97% |
| AMD | Semiconductors / memory tied to AI capex | 8.15% |
| MU | Semiconductors / memory tied to AI capex | 5.91% |
| PANW | Enterprise software tied to AI capex | 5.29% |
| INTC | Semiconductors / memory tied to AI capex | 4.43% |
| TLN | AI data-center / power infrastructure | 2.55% |
| NBIL | AI data-center / compute (leveraged) | 2.10% |
| DDOG | Enterprise software tied to AI capex | 2.01% |
| NOW | Enterprise software tied to AI capex | 2.37% |
| AVGO | Semiconductors / memory tied to AI capex | 1.39% |
| WYFI | AI data-center / compute infrastructure | 0.77% |
| FTNT | Enterprise software tied to AI capex | 1.48% |
| CRWV | AI data-center / compute infrastructure | 1.18% |
| NBIS | AI data-center / compute infrastructure | 1.30% |
| CRWD | Enterprise software tied to AI capex | 1.38% |
| VST | AI data-center / power infrastructure | 0.82% |
Why this is disclosed, not capped
The IPS's diversification clause calls for flagging single-theme clustering "when it appears," deliberately short of a hard numeric cap. A hard cap on a theme this broadly defined would either be meaningless (set so high it never binds) or would force selling good businesses for a bad reason (correlation, not fundamentals). The honest response is transparency: this page, kept current, plus the standing risk this creates. The NBIS and CRWV selloffs in July 2026 were a live example of exactly this correlation showing up on a single bad week — multiple names in this table moved together on the same news, regardless of their individual fundamentals.
What would change this
Nothing here is a plan to trim the theme — it's a deliberate view that the underlying demand is real and durable. What would change my mind: a sustained deceleration in hyperscaler capex guidance across multiple names in the same earnings season, not a single stock's bad quarter. Until then, this table exists so the concentration is checkable, not just described in prose once and forgotten.