Broadcom — AI supply-chain exposure

Broadcom is the AI chain's second silicon story and much of its nervous system. It co-designs the custom accelerators (ASICs) that hyperscalers build to reduce their dependence on merchant GPUs, and it dominates the high-speed networking — switching and optical interconnect — that moves data between tens of thousands of chips in an AI cluster.

That gives Broadcom two distinct hooks into the build-out: it benefits when the largest buyers want their own silicon, and it benefits from the sheer scale of interconnect every cluster needs regardless of whose accelerator wins. The model reads it as a producer across compute and networking — a name levered less to any one chip than to the wiring of AI at scale.