We are not investing in artificial intelligence.
We are investing in the substrate of an economy
whose participants are no longer constrained
by the rate of human birth.
For readers who will not, in this session, read all five chapters: this is the essay in roughly seven hundred words.
From the agricultural revolution forward, every economic model on Earth has assumed that the productive capacity of an economy expands at roughly the rate at which humans can be born and educated. That assumption is being broken in real time by four converging mechanisms — artificial intelligence, semiconductors, robotics, and commercial space — that produce the substrate of an economy whose participants are no longer constrained by the rate of human birth.
Hyperscaler capital expenditure on this rotation has been guided at $660–$690 billion for 2026 alone, against $360 billion in 2025. Cumulative committed capital through 2030 is approaching $7 trillion. AI capex as a share of U.S. GDP has surpassed every major historical mobilization other than the world wars. The rotation is not theoretical. It is in the financial signature of the public equity market right now.
The seven trillion dollars do not disperse evenly. They funnel through five physical chokepoints whose owners cannot be replicated within any short horizon: lithography (ASML's monopoly on EUV scanners — sixty-six total deliveries through 2027), high-bandwidth memory (the SK Hynix–Micron–Samsung oligopoly producing 72% operating margins), advanced packaging (TSMC CoWoS, fully booked through 2026), power and grid (transformer lead times of 128 weeks, GSU lead times of 144 weeks, grid-interconnect queues at a 2,100-day median), and optical interconnect (Lumentum's 50–60% EML share, Credo's 73% AEC share, Broadcom's 80%+ switch silicon).
These positions show up on the income statements of the past four quarters. Their durability is not a forecast; it is observable in already-audited filings.
Beginning January 13, 2023 — six weeks after ChatGPT's public release and ten months before the first major sell-side AI thematic research — I began compounding personal capital against the thesis articulated in this essay. Forty months later, the account had grown the equivalent of one dollar to roughly $131. The maximum drawdown was −36.5% over three months in spring 2024. I held the position; the entire drawdown was recovered in the single subsequent month.
The record was generated by concentrated leveraged-thematic-ETF positions — an instrument set with daily-reset compounding mechanics that magnify both upside and downside, and that nobody should mistake for a recommendation. The point of Chapter III is not the number. The point is what I held through, and the vintage of the conviction.
Of fifteen legendary investors examined — Buffett, Soros, Druckenmiller, Tudor Jones, Griffin, Tepper, Cohen, Ackman, Einhorn, Loeb, Coleman, Dalio, Simons, Singer, Marks — the great majority compounded a documented personal record before institutionalizing. The narrative templates that have attracted institutional capital across these investors cluster into five archetypes: the Systematizer, the Value Detective, the Convexity Hunter, the Activist, and the Synthesizer. The thesis-driven concentrated approach belongs to the Convexity Hunter tradition — Druckenmiller, Tudor Jones, Soros — and the structural reasons it keeps producing the same shape of investor are the subject of Chapter IV.
I am investing my own capital against this thesis. I am writing about it. The essay you are reading is what comes out of three years of trading and reading and asking the same question. The closing is short, and it does not ask anything of you.