Broadcom's Future $100 Billion Custom Chip Business Is Your Sign to Buy Now

Broadcom's custom AI chip business is scaling rapidly, with the relevant division growing 106% year over year to $8.4 billion in Q1. Management believes annual sales for custom AI chips could exceed $100 billion by the end of next year, versus likely current annual sales below $30 billion. The article argues this could help Broadcom's total revenue more than double from $68 billion over the last 12 months.

Analysis

The market is likely underestimating how quickly custom silicon changes the AI profit pool: the real transfer is not from Nvidia to Broadcom, but from hyperscalers’ operating expense lines into capex-heavy, multi-year strategic commitments. That typically benefits the design house first, then the foundry/advanced packaging stack later, while pressuring GPU suppliers only at the margin because custom ASICs mostly absorb stable, high-volume inference and narrow training workloads rather than replace general-purpose compute outright.

The second-order winner set is broader than AVGO: any acceleration in custom chip deployment tightens demand for leading-edge wafer starts, CoWoS-like packaging capacity, HBM, and EDA/IP tools, but with lumpy timing because the revenue bridge only converts after tape-out, qualification, and rack-level deployment. The risk is that consensus extrapolates headline design wins into near-linear revenue, when the bigger variable is whether hyperscalers can actually absorb the full silicon output into deployed clusters without power, networking, or software bottlenecks creating slip into 2027.

The main contrarian point is that this is becoming a crowded “pick-and-shovel” trade, and the stock may already discount a high-teens to low-20s growth regime. If custom chip revenue merely lands at the low end of expectations, multiple expansion becomes the vulnerable leg; the cleaner trade is not a chase-long, but owning AVGO versus other mega-cap AI beneficiaries with less visible monetization paths. Near term, any delay in ramp or weaker AI capex commentary from hyperscalers would hit sentiment hard even if the long-term thesis remains intact.

Key Decisions for Investors

• Maintain a tactical long AVGO, but only on pullbacks after hyperscaler earnings or AI capex-related weakness; upside is strongest over 6-12 months, while the main risk is multiple compression if the market has already priced the 2027 ramp.
• Express the view as a pair: long AVGO / short NVDA for 3-6 months if you want a relative-value trade on AI workload migration; this benefits if custom ASIC penetration rises faster than general-purpose GPU spend without requiring a broad AI selloff.
• Add a secondary basket long in the AI supply chain that benefits from rising custom chip volumes but lower headline multiple risk, with preference for foundry/advanced packaging exposure over pure AI software names; this is a 6-18 month trade with better second-order leverage.
• Avoid chasing INTC as an AI proxy here; its participation is too indirect and execution-sensitive. If held, use it only as a small optionality position, not a core expression.
• If AVGO rallies sharply into a guidance event, consider buying 3-6 month put spreads to hedge a disappointment risk; the setup is asymmetric because the stock likely prices in perfection before the revenue inflection is fully visible.