Broadcom's AI Chip Revenue Grew 221% Last Quarter. A $1,000 Investment a Year Ago Would Be Worth This Much Today.

Broadcom reported fiscal Q3 AI semiconductor revenue of $16.7 billion, up 221% year over year, and forecast fiscal Q4 AI chip revenue of $21.7 billion, up 236%. Management projects AI semiconductor revenue to reach $230 billion by fiscal 2028, supporting an approximately $30 EPS target and an implied $135 billion in free cash flow. Shares reacted only modestly as total Q4 revenue guidance of $34.8 billion trailed some expectations above $35 billion and higher memory costs pressured gross margins, but the stock trades at roughly 12x management's fiscal 2028 earnings target. Key risks are customer concentration, memory/component supply constraints, and potential regulatory or public opposition that could slow data-center construction.

Analysis

The key re-rating question is not whether AI revenue grows, but whether custom ASICs convert from a hyperscaler-specific supplement into a durable share-taker versus merchant GPUs. AVGO captures more than accelerator silicon: each scale-out deployment pulls through high-speed switching, optics and connectivity content, creating a higher-value system attach rate than the market’s focus on chip revenue implies. That is incrementally adverse to NVDA’s inference mix and favorable to AVGO’s foundry/packaging ecosystem, principally TSM, while constrained memory availability shifts bargaining power toward MU and other HBM suppliers.

The company’s long-dated target should be discounted until customer commitments, unit economics and supply reservations can be independently corroborated. A small number of customers creates asymmetric downside: a single delayed data-center campus, model-efficiency breakthrough, or internal silicon redesign can create a material quarterly air pocket because custom programs have limited spot-market fungibility. Near-term gross-margin pressure from memory is less important than whether AVGO can preserve system-level pricing as component costs rise; margin resilience in the next two quarters is the cleaner validation signal than revenue alone.

Consensus appears to frame the muted reaction as a valuation opportunity, but the more likely explanation is that investors are assigning a low probability to full realization of the outer-year plan. That skepticism can unwind over 1-3 months if hyperscaler capex commentary confirms incremental ASIC deployments, but a broad AI-capex reset would compress AVGO’s multiple before the revenue shortfall is visible. Structurally, custom silicon is a credible 6-18 month share-loss vector for NVDA, though NVDA can offset it if total inference demand expands faster than ASIC substitution.

Key Decisions for Investors

• Initiate a 6-12 month long AVGO / short NVDA pair in equal dollar beta-adjusted size. The trade isolates custom-ASIC and networking share gains from broad AI-demand risk; target 15-20% relative outperformance, with a stop if AVGO reports AI revenue or backlog conversion below guidance in either of the next two earnings prints.
• Add AVGO only on confirmation that gross margin stabilizes despite memory inflation; use the next earnings release as the catalyst. A failure to demonstrate pricing pass-through, or a meaningful reduction in hyperscaler-specific shipment expectations, invalidates the long thesis.
• Maintain a tactical long MU basket exposure over the next 1-2 quarters as supply tightness raises memory content costs across AI systems, but take profits if memory pricing begins to normalize or AVGO and peer procurement commentary signals demand deferrals.
• Watch GOOG and META capex guidance and disclosed TPU/MTIA deployment cadence as leading indicators rather than treating AVGO management’s outer-year target as base-case earnings. A synchronized capex slowdown is the trigger to reduce AVGO exposure and cover the NVDA short.