Scry Fund

AI Buildout Economics

2 analysis reports

The AI Buildout Constraint Scanner
Ranking AI Infrastructure and Mag 7 Stocks by Constraints and Fundamentals
Semiconductors9 July 2026Start with physical bottlenecks, then assess growth, quality, valuation, momentum, and risk for each stock

An AI buildout screen inspired by Gavin Baker's watts-and-wafers framework. It ranks 33 AI infrastructure and Mag 7 stocks by scarcity exposure, revenue growth and scale, momentum, quality, valuation, and risk. The 9 July 2026 run prioritizes TSM, NVDA, MU, ASML, KLAC, LRCX, AMAT, CRDO, AVGO, VRT, and GOOGL for research.

The Economics of a Gigawatt
Costs, Revenue, and Returns for a 1GW AI Datacenter
Semiconductors18 June 2026The base case works without scarcity pricing if throughput per watt rises faster than token prices fall

A model of one 1GW GB200-class datacenter: ~$38-41B upfront capex and ~$7.1B/year to run. Renting GPUs at $3.30/GPU-hour yields a 42.8% operating margin. The lab needs ~$31B/year of inference revenue per gigawatt at a 60% gross margin to justify compute costs. Throughput rises from H100 900k to B200 2.8M tok/sec/MW (3.1x), lifting revenue per gigawatt ~56% even if token prices halve. Figures match the author's sheet; the model flags ARK's $8.5B versus $7.14B annualized-cost difference.