Broadcom

Agentic Economy
View Market Map
Layer 7: Physical Infrastructureas Broadcom

Broadcom is a global semiconductor and infrastructure software conglomerate that designs networking chips, custom AI accelerators (ASICs), and enterprise software. After acquiring VMware in 2023, Broadcom became one of the largest technology companies in the world by market capitalization.

In the AI infrastructure space, Broadcom is a major player in custom silicon: the company designs custom AI training and inference chips (XPUs) for hyperscale cloud customers including Google (which uses Broadcom-designed TPUs). Broadcom's networking semiconductor division also produces the switching and routing chips used in AI datacenter fabrics.

The Google-Broadcom-Anthropic Supply Chain

In April 2026, Broadcom announced a major supply assurance agreement to provide Google with custom Tensor Processing Units, along with networking and other components for Google's next-generation AI data center racks through up to 2031. As part of this expanded collaboration, Anthropic will access approximately 3.5 gigawatts of TPU-based computing capacity beginning in 2027 — a commitment spanning multiple gigawatts as Anthropic races to keep pace with enterprise demand that has driven its revenue run rate past $30 billion. This deal positions Broadcom at the center of the most consequential AI infrastructure alliance: Google designs the TPU architecture, Broadcom implements the silicon design and supplies networking components, TSMC fabricates the chips, and the resulting compute capacity powers both Google's own AI services and Anthropic's Claude — the leading enterprise AI platform.

Broadcom's dual position in both custom AI accelerators and datacenter networking makes it a critical supplier at the physical infrastructure layer of the agentic economy. As AI labs face acute compute crunches and the industry shifts from NVIDIA GPU monoculture toward custom silicon, Broadcom's role as the primary ASIC implementation partner for the world's largest pool of AI compute is increasingly strategic.

Further Reading