Thematic equity portfolio · long-only research note
Twenty positions built on one claim: that the demand curve for computation is growing faster than the physical systems that feed it — fuel, molecules, land and grid — can be rebuilt. Each bucket is a different link in that chain, and each position carries the condition that would prove it wrong.
The buckets are ordered because they depend on each other. Fuel, molecules and land are upstream of compute, and the minerals in its hardware sit beside them; the last bucket exists because the first three could turn out to be one trade wearing three hats.
The book indexed to 100 twelve months ago, in USD, rebalanced to target weight on the first session of each month. This is a backward-looking simulation of the current weights, not a track record — the positions were chosen with the benefit of knowing this window.
Bucket targets are set first; weights within each bucket are set line by line, and the change log records why. Equal capital is not equal risk: see the risk table. Diversifiers carry a large share of the book, which is a deliberate statement that the book's own thesis might be wrong.
Every position converted to the USD base at each day's rate. Bars are coloured by direction, not by bucket — this chart answers "what worked", and bucket identity is carried by the labels.
The gap between the two columns is pure FX — return that has nothing to do with whether the thesis was right.
A hedged return is not the local return. Rolling a short forward on the currency earns or pays the gap between the two interest rates, so a dollar investor hedging reais gives up roughly the Selic-minus-Fed spread every year. The toggle above the charts uses these rates.
Horizon.
Review and exit.
The Diversifiers sleeve exists in case the rest of the book is one trade. If it does its job, it should load lightly on the factors the thesis is long and explain little of its own movement through them.
Every line, with the case for it and the condition that kills it.
This page is rebuilt on a schedule from public market data. Each run re-fetches every position, validates the result against the previous one, and is discarded without publishing if it looks wrong — a stale page beats a corrupted one.
The parts of this that a careful reader should push on. They are here because leaving them out would make the numbers above look more certain than they are.
The KOSPI is an index, so a real book expresses it through a tracking fund or futures and eats the tracking error and FX. Gold is held as spot-backed shares rather than a dated future, so there is no roll and no expiry to manage.
It accrues the short-rate differential day by day on the full marked value, from dated daily series for USD (T-bills) and BRL (CDI) and a dated policy-rate table elsewhere. Real forward points also carry cross-currency basis, and TWD and BRL hedge offshore through non-deliverable forwards; the hedge note shows what one point a year of basis would cost.
A fund that listed part-way through the year has no history before it existed. The curve renormalises across whatever is trading on each date, so the headline return describes the positions that were actually present — not always the whole book.
Taipei and Seoul close before New York opens, so daily returns for those lines are measured against a different information set. Cross-market correlations here read lower than the true economic relationship; weekly sampling would reduce it.
The curve rebalances to target weight on a fixed schedule. A held book would drift with its winners, so this understates the concentration that would actually have built up.
Each refresh compares Yahoo's closes for five US-listed lines against Nasdaq's published closes, session by session.
Returns include reinvested dividends and are net of each fund's own fee, which is already in its price. The headline is gross of commission, spread, withholding and the FX spread; the net-of-cost series in the performance section charges stated assumptions for each.
Twelve months, ending on the day the note was written, for positions selected today. That is the most flattering possible framing, and the reason the drawdown column matters more than the return column.