Broadcom Has A Marvell Problem

Broadcom’s softer-than-expected fiscal Q3 AI revenue outlook has raised concern that its AI premium may be running ahead of near-term growth. The piece flags increasing competitive risk from Marvell’s Nvidia-backed momentum in custom silicon and AI networking, though Broadcom’s six core XPU wins still support multi-year AI revenue visibility through at least FY2028.

Analysis

The market is beginning to price AVGO more like a near-term AI spending proxy than a multi-year infrastructure compounder, and that is where the setup gets interesting. If the stock has already capitalized a large portion of FY27–FY28 XPU conversion, then any quarter-to-quarter moderation in AI revenue growth can compress the multiple even if the underlying design-win count remains intact. That makes this less about absolute demand and more about the timing gap between bookings and revenue recognition: the longer the ramp stretches, the more vulnerable the premium becomes to a de-rating event.

Competitive pressure is most acute in two places: custom accelerators and the networking stack that surrounds them. A credible second-order effect is that hyperscalers may split awards more deliberately to preserve negotiating leverage, which can slow AVGO share capture without reducing total AI capex. That would also pressure smaller vendors in the silicon ecosystem first — analog, interconnect, and niche networking names — because procurement teams typically cut excess overlap before they reduce core compute spend.

The contrarian read is that the “miss” in sentiment may be more about pace than destination. Six wins imply AVGO still has a strong line of sight to multi-year conversion, and the relevant risk is not losing AI relevance but losing the right to be valued as a near-perfect execution story. For NVDA, the signal is neutral-to-slightly positive: any forced diversification away from pure GPUs tends to reinforce Nvidia’s ecosystem moat, but it also introduces some cannibalization risk in scale-up interconnect over a 12–24 month horizon. Near term, the catalyst path is mostly guidance revisions and customer commentary over the next 1–3 quarters; the tail risk is a broader re-rating of AI infrastructure multiples if hyperscaler capex growth decelerates or if Marvell continues to win credibility as an alternative supplier.

Key Decisions for Investors

• Short AVGO vs long NVDA into the next earnings cycle (4–8 weeks): AVGO has the more fragile near-term multiple if AI revenue timing slips again, while NVDA retains cleaner ecosystem leadership; target a 10–15% relative move, stop if AVGO re-accelerates guide by >5%.
• Buy AVGO downside protection via 1–2 quarter puts or put spreads ahead of the next print: best risk/reward if the market has already priced FY27 visibility but remains sensitive to another softer guide; structure for a 2–3x payoff on a 1–2% portfolio debit.
• Pair long a basket of non-AVGO AI supply-chain laggards only if hyperscaler capex remains intact: look for any pullback in optical/interconnect or custom silicon names that have been dragged by sentiment, because the first-order pain is likely overdone relative to multi-year demand.
• Avoid chasing AVGO strength until the market proves it can absorb slower quarterly AI growth without multiple compression; if shares de-rate 8–12% on no-change fundamentals, that becomes the better entry for a 6–12 month long.
• For tactical longs, prefer NVDA on weakness over AVGO on strength: NVDA’s risk/reward is cleaner if the market starts rewarding platform control and punishing execution-timing uncertainty in adjacent infrastructure names.