Broadcom makes money by selling high-value semiconductor products and enterprise infrastructure software to hyperscalers, OEMs, enterprises, telecom equipment makers, storage vendors and other global technology customers. Its current business model is built around long-cycle chip and software development, deep customer engagement, outsourced and partner-based manufacturing, and large strategic acquisitions, with VMware now central to the software side of the company.
In Q2 fiscal 2026, Broadcom reported revenue of $22.187 billion, up 48% year over year. Semiconductor Solutions generated $15.009 billion, or 68% of revenue, while Infrastructure Software generated $7.178 billion, or 32% of revenue. Adjusted EBITDA was $15.244 billion, equal to 69% of revenue, showing the scale and margin profile of the combined semiconductor and software portfolio.
- Semiconductor Solutions: This segment includes custom AI accelerators and XPUs, Ethernet switching and routing silicon, Ethernet NICs, PHYs, optical components, broadband chips, wireless chips, storage products and industrial semiconductors. AI infrastructure is the main growth engine, with Q2 fiscal 2026 AI semiconductor revenue of $10.8 billion, up 143% year over year.
- Infrastructure Software: This segment includes VMware and other enterprise software assets focused on private cloud, virtualization, mainframe, cybersecurity and infrastructure management. VMware expanded Broadcom’s position in private cloud software, with growth tied to VMware Cloud Foundation demand, non-terminable license revenue and the transition toward subscription licensing.
- Sales channels: Broadcom sells through direct sales, distributors, channel partners, OEMs, contract manufacturers and software customers. The customer base is global, but revenue concentration is high. In fiscal 2025, one semiconductor customer represented 32% of total net revenue, and Broadcom estimated that its top five end customers represented about 40% of net revenue.
Broadcom’s competitive advantage comes from its position in custom silicon, Ethernet-based AI networking, large-scale customer relationships, broad infrastructure product coverage and high free-cash-flow generation. The company is not a pure merchant chip supplier. It works closely with large customers on custom accelerators and networking components, which gives it a differentiated role in hyperscale AI infrastructure.
Its direct competitors vary by product area. In AI compute and data-center acceleration, NVIDIA is the most important global peer and cross-category competitor. NVIDIA dominates general-purpose GPU systems, while Broadcom is positioned as an alternative and complement through custom AI accelerators and Ethernet networking silicon. In networking, Broadcom competes with companies such as Marvell and other semiconductor vendors. In infrastructure software, it competes with enterprise software and cloud infrastructure providers across virtualization, private cloud, cybersecurity and systems management.
Broadcom holds one of the strongest market positions among diversified semiconductor and infrastructure software companies. Its role in AI data centers is expanding quickly through custom accelerators, Ethernet switching, routing, NICs, PHYs, optics and full rack or system-level infrastructure. Management guided Q3 fiscal 2026 AI semiconductor revenue to $16.0 billion, implying more than 200% year-over-year growth, and guided total company revenue to about $29.4 billion.
China remains relevant as a shipment and delivery location, though Broadcom states that delivery geography does not necessarily reflect end-customer location. In fiscal 2025, 17% of net revenue came from shipments or deliveries to China, including Hong Kong, down from 20% in fiscal 2024. The exposure still matters because of U.S.-China trade restrictions, export controls, China-Taiwan risk and possible supply-chain disruption.
Overall, Broadcom’s market position is defined by a rare mix of AI semiconductor growth, Ethernet networking depth, enterprise software scale and disciplined cash generation. The main strategic question for investors is how durable the AI custom silicon cycle proves to be, especially given customer concentration and the speed at which hyperscaler demand shapes quarterly results.