Is Broadcom Stock a Buy Now?

Broadcom reported fiscal Q3 AI semiconductor revenue of $16.7 billion, beating its $16 billion target and rising 221% year over year; CEO Hock Tan guided to $21.7 billion in Q4 AI revenue, up 236% year over year. Total Q3 revenue was $29.6 billion, up 86%, with $13.7 billion in free cash flow. The article says these results support the growth outlook despite competition from Marvell, while noting shares are 31% below their peak and trade at 28.3 times forward earnings.

Analysis

The key read-through is not simply stronger AI demand: it raises the cost of betting that custom accelerators are a near-term threat to Broadcom’s growth. But a larger market does not guarantee durable pricing power. If Alphabet can qualify Marvell as a credible second source, it may gain leverage in future design and supply negotiations even while Broadcom’s near-term volumes remain strong. Watch customer allocation and program wins, not just aggregate AI revenue.

For NVIDIA, custom silicon is a substitution risk in workloads where customers prioritize cost and efficiency; it is not necessarily a direct displacement of general-purpose GPU demand. Broadcom’s networking exposure may also benefit as accelerator clusters scale, though that does not remove customer-capex or execution risk.

Near term (days), semiconductor sentiment remains vulnerable to read-throughs from peers and macro risk appetite; the share-price discount alone is not a catalyst. Over 1–3 months, the next guide and evidence of customer/program concentration matter more than the current growth rate. Over 6–18 months, multi-sourcing, customer bargaining power, and whether AI infrastructure spending converts into sustained deployments will determine the multiple. The contrarian risk is treating one strong aggregate result as proof that competition is harmless. Falsify the constructive view if Broadcom cuts its AI outlook, misses subsequent delivery expectations, or customer disclosures indicate meaningful program migration; verify segment economics and customer concentration before underwriting margin durability.

Key Decisions for Investors

• Prefer a staged, risk-budgeted AVGO long over chasing a sharp post-results move; reassess after the next guidance update. The thesis is growth durability, not that the shares are cheap. Reduce or exit if AI guidance is lowered or reported execution no longer supports the outlook.
• Do not short MRVL solely on this evidence. Treat it as a watch: a confirmed material Broadcom program loss or customer allocation shift would strengthen a relative-value long AVGO/short MRVL trade; without that evidence, both can benefit from expanding custom-accelerator demand.
• Monitor GOOG’s accelerator sourcing and deployment disclosures for signs of multi-sourcing. A credible second source could pressure Broadcom’s negotiating leverage before it materially reduces reported volumes.
• Track NVIDIA for workload substitution rather than assume a broad GPU-demand collapse. Evidence that custom silicon is taking share in specific inference deployments would support relative underperformance risk; continued broad infrastructure expansion could leave both NVIDIA and Broadcom beneficiaries.