Macro views
A dated record of how the thinking developed. Each entry is the view as it stood at that point; read top to bottom and the worldview assembles itself, one constraint at a time.
The largest capital reallocation of the decade is the physical build-out behind compute and electrification — and the market keeps paying for the first-order winner while underpricing the chain behind it. The binding constraint is not generation; it is the grid and the materials that feed it. Quantitative tightening and a slow drift away from dollar reserves set which version of that trade works, and when. Expressed through a concentrated structural-growth book — held with open eyes about the correlation it carries in a risk-off rotation.
The electrification story was being told as a generation story. The real chokepoint is the grid itself — transformers, turbines, transmission — where order backlogs run for years and supply can't be willed into existence. GEV sits on the scarce side of that equation. A contrarian committee view at the time; the backlog data has since caught up to it.
The market kept treating the AI power-demand step-change as something to be faded. It looks closer to a regime: independent power producers with nuclear and gas baseload gain pricing power as grid constraints bind. The mispricing was in the durability, not the direction.
On the surface a shipping story; underneath, a chain that ends at the Fed. The Strait carries close to a fifth of the world's oil, so a disruption pushes crude — and the transport costs riding on it across trucks, rail, and air — straight into broad inflation. Set that against a US labor market weaker than the headline unemployment rate admits, and the conditions for stagflation are in place. The bind is the Fed's own: it has historically answered inflation with hikes, but tightening into a softening labor market is its own risk — and it would also choke the borrowing financing the AI build-out (data centers, photonics, commodities infrastructure), one of the few engines still carrying the economy. I held the view through gold and worked the exit on the transitory-versus-Fed-timing question. A headline that looks purely geopolitical, traced the whole way to monetary policy and the future of AI capital — that interconnectedness is the part of markets I find most compelling.
Stepping back from any single name: electrification isn't a theme to trade around the edges, it's where an enormous share of capital is being structurally redirected. Top-down and bottom-up arrived at the same place — the macro current and the GEV fundamentals are the same trade.
Follow the infrastructure behind the headline. AI demand routes through data centers, data centers through copper, and new copper supply takes fifteen to twenty years to bring online. A structural supply crunch is the predictable result — and it's already surfacing as M&A in the miners, buying reserves rather than building them.
Gold is doing more than tracking real yields; it's reading a slow diversification away from dollar reserve assets against a noisier geopolitical backdrop. Carried as an active position and monitored as attribution against Fed policy — not as a static hedge left in a drawer.
The working hypothesis behind my MSc thesis: balance-sheet reduction doesn't deflate asset classes evenly — liquidity-sensitive assets respond first and hardest. The open questions are the interesting ones — whether the asymmetry depends on the pace of QT or its size, whether it reverses on cessation, and whether it's stable across the 2017–19 and 2022–24 episodes. Tested on FRED and Bloomberg credit data.
Where the worldview points geographically: Southeast Asia, where Western growth capital is still under-allocated against digital leapfrogging and supply-chain rewiring — and the Singapore–Taiwan–Korea axis as the software-and-silicon layer of the AI supercycle. Two distinct theses, a capital-gap story and a supercycle story, not one.
Most capital trades the first-order effect — AI, therefore the chip. The edge sits further down the causal chain: the power, the cooling, the grid, the materials, eventually the labor and insurance exposure. Fewer eyes are on where the energy actually flows. The method is to map the chain explicitly and position where the map thins out.