Summary
Gaetano previews three deep-dive research projects—photonics durability, subsea/offshore data centers, and AI-RAN—arguing that the next AI infrastructure bottlenecks will shift to optical connectivity, ocean-based computing, and edge/telecom integration. For markets, the implication is that investors should prepare for thematic rotations into photonics suppliers, subsea cable and cooling companies, and telecom infrastructure plays before they are fully priced in.
-
•Microsoft's Project Natick ran subsea servers for two years with a lower failure rate than land-based servers, validating the subsea data center concept.
-
•The author plans to analyze which photonics companies can survive architecture changes (scale-out, scale-across, scale-up, CPO) vs. those dependent on a single product cycle.
-
•The AI-RAN project will cover Nokia, NVIDIA, and Blaize, examining how AI moves closer to factories, warehouses, defense, and telecom networks.
-
•The three projects are connected by the question: where are the next AI infrastructure bottlenecks, and which companies have real exposure before the market prices it in?
-
•The author explicitly withholds the name of the first photonics company until the full post, signaling a specific thesis tied to a recent event that moved the company from connectivity to a larger story.
-
•Subsea data centers offer cooling, land access, power proximity, and latency benefits, but face maintenance, corrosion, permitting, and insurance challenges.