Summary
The author argues that the most durable long-term AI infrastructure investments sit at physical bottlenecks that get tighter as AI scales, with customer-backed capacity and multi-year demand. The article is a teaser for a multi-part series that will name specific companies, but the current piece provides only the investment framework and no disclosed positions or explicit recommendations.
-
•The author prioritizes companies with multi-year infrastructure demand, customer depth, hard-to-replicate physical/technical capabilities, and ability to benefit from architecture changes rather than a narrow product cycle.
-
•The framework includes six criteria: being a real bottleneck, necessity to large customers for years, increasing content as architecture scales, manufacturing depth competitors struggle to copy, ability to invest through cycles while protecting cash flow, and multiple ways to win if markets evolve differently.
-
•The company discussed checks many boxes: physical layer of AI infrastructure, bottleneck that gets harder as AI scales, long customer relationships, manufacturing scale, cash flow, and customer-backed capacity agreements.
-
•The opportunity is described as 'much larger' after recent releases, with inference cited as the key driver that pulls optical components into scale-up.