$TSM

i asked Tessara a simple question: If you could own one “fat pitch” in AI infra, what is the cleanest asymmetric bet? it looked across its constraint stack + database: memory, Nvidia, photonics, power, energy, advanced packaging. the answer it gave me was $TSM. Reason is simple: Nvidia fabs at TSMC. AMD fabs at TSMC. AWS Trainium fabs at TSMC. Google TPU fabs at TSMC. Meta MTIA fabs at TSMC. the custom silicon risk that caps Nvidia’s dollar share is the bull case for TSMC. every successful ASIC potentially fragments accelerator value away from Nvidia. but it still has to be manufactured somewhere. and at the leading edge, that somewhere is usually TSMC. Tessara’s constraint database shows TSMC as the broadest “convergence winner” across the AI chokepoint chain: advanced nodes, lea