AI supply-chain thesis — mapping bottlenecks, focus companies, and supply-chain exposure for investors.
**Bottleneck theme:** Custom Silicon **Focus:** $AVGO — Broadcom Inc. Broadcom is the custom-ASIC duopoly half (with $MRVL) — the silicon partner of choice for hyperscalers building their own AI accelerators (Google TPU, Meta MTIA, ByteDance, others) and also the leader in merchant AI fabric switching ASICs (Tomahawk 6 + Jericho 3-AI). Where NVIDIA sells the merchant GPU, Broadcom sells the custom silicon hyperscalers use to either supplement NVIDIA or build a credible second source. The Tomahawk 6 (102.4Tbps) and Jericho 3-AI roadmap also positions Broadcom as the dominant switch-ASIC supplier under Arista and Cisco AI fabrics — a second franchise on top of the custom-silicon business. The investment case is structural: every percentage point of hyperscaler accelerator spend that flows into custom silicon is incremental Broadcom revenue, and the network silicon under AI Ethernet fabrics is a second compounding driver. The bear case is concentration (a small number of hyperscaler programs drive most growth), TSMC CoWoS allocation (competing with NVIDIA), and an integrated software/service stack from VMware that the market has yet to fully reward. Pair with $MRVL (the smaller duopoly half) and $ANET (the system that uses Broadcom silicon).
Browse all AI supply-chain theses · Macro trends · Industries · Product categories