$Broadcom(AVGO.US)’s upcoming Q3 earnings test whether the AI semiconductor boom is diversifying beyond Nvidia's dominance. Wall Street expects $29.4B revenue and $3.24 adjusted EPS, with Broadcom guiding for ~$16B in AI revenue (up >200% YoY).

❓ This raises a key question: Does Nvidia need to lose for Broadcom to win in AI infrastructure?

Previous Q2 results showed revenue of $22.2B (+48% YoY) and AI revenue of $10.8B (+143% YoY), driven by high EBITDA margins (69%) and free cash flow (46%). Instead of competing head-on with Nvidia's GPUs and CUDA ecosystem, Broadcom partners with hyperscalers (such as Meta on its 2nm MTIA silicon and OpenAI on inference chips) to build custom AI accelerators (ASICs/XPUs) and high-speed Ethernet networking.

🐂 The Bull Case: The AI infrastructure market is expanding rather than acting as a zero-sum game. Nvidia maintains dominance in frontier computing, while hyperscalers deploy custom silicon for predictable, cost- and power-efficient workloads. Broadcom benefits from custom chips and cluster networking, enabling rapid growth alongside Nvidia.

🐻 The Bear Case: Broadcom faces extreme expectations, valuation pressure, customer concentration risk among a few giant hyperscalers, and Nvidia's entrenched ecosystem advantages. Meeting high guidance may not be enough for Wall Street.

👀 Looking Ahead: Broadcom sees potential for its AI business to surpass $100B annually by 2027. Success depends on custom silicon shifting from experiments to permanent infrastructure, sustained capex, expanding inference, and execution across tech roadmaps.

⭐ 🗝️ Key takeaway: Ultimately, the key metrics I am watching are whether Broadcom hits its $16B target, broadens customer adoption, and proves custom silicon and GPUs can co-exist as dual pillars of AI compute.

My Trade: In the meantime, I will just sit on my current position.