Broadcom: The Custom ASIC Opportunity

Custom AI accelerators — application-specific integrated circuits (ASICs) designed for specific AI workloads — represent a growing alternative to NVIDIA's general-purpose GPU architecture. Google's TPU, Meta's MTIA and Apple's Neural Engine are all custom ASICs designed and manufactured with Broadcom's assistance.

The economics of custom ASICs are compelling for hyperscalers running AI workloads at scale. A custom chip optimised for a specific inference workload can deliver 3–5x better performance per watt than a general-purpose GPU, translating directly into lower data centre operating costs at the scale of billions of daily AI queries.

Broadcom's networking silicon — the Tomahawk and Jericho product families — is equally important. AI training clusters require ultra-low-latency, high-bandwidth interconnects between thousands of GPUs. Broadcom's networking chips are the dominant solution for this application, with market share exceeding 70%.

The VMware acquisition, completed in 2023, adds a recurring software revenue stream that is transforming Broadcom's financial profile. VMware Cloud Foundation subscriptions are growing at 30%+ annually, and the software business now contributes approximately 40% of total revenue at significantly higher margins than hardware.

Research Disclaimer: This article is intended solely for informational, educational and research purposes. Nothing in this article constitutes investment advice or a recommendation to buy or sell any security. BGAM™ scores, scenarios and classifications are model-based analytical assessments for illustrative purposes only.