Semiconductor Sell-Off: The 1 AI Chip Stock I'd Buy and 1 I'd Avoid

Broadcom’s AI semiconductor revenue rose 221% year over year in fiscal Q3 2026, driven by custom accelerators and AI networking; management expects further growth in the following quarter. The article favors Broadcom, citing a forward P/E of about 18, versus Intel’s roughly 51, while noting Broadcom’s exposure to a small group of large customers and Intel’s execution, manufacturing, and competitive risks. It frames the semiconductor sell-off as a potential buying opportunity but presents Intel as the higher-risk choice.

Analysis

The key distinction is cash-flow visibility, not simply which company is “more AI.” Broadcom’s custom silicon and networking can capture spend even when hyperscalers shift workloads away from general-purpose accelerators; that creates substitution risk for incumbent GPU platforms while making Broadcom’s results unusually sensitive to a small number of design wins and customer budgets. The 221% growth figure is not enough to underwrite durability without the AI revenue base, customer concentration, and conversion of announced programs into shipments. Watch order timing and customer capex, not just headline growth.

Intel is a different risk: its foundry ambition competes for capital and management attention before customer adoption and yields are established. A high forward P/E may overstate valuation risk if near-term earnings are unusually depressed, but it also leaves the stock exposed to multiple compression if process milestones slip or external customers do not commit. Validate the cited valuation on consistent earnings estimates before treating it as a clean relative-value signal.

Over days, a sector rebound can lift both irrespective of fundamentals. Over 1–3 months, hyperscaler capex commentary and Broadcom program updates matter more for AVGO; Intel process/yield milestones and credible foundry customer commitments matter more for INTC. Over 6–18 months, Broadcom’s customer concentration and Intel’s capital intensity are the defining risks. A contrary possibility: investors may be underpricing Intel’s option value if execution improves, while over-extrapolating Broadcom’s rapid growth from a potentially narrow customer/program base.

Key Decisions for Investors

• Prefer AVGO over INTC on a relative basis, but avoid treating the article’s forward P/E comparison as decisive until earnings periods, estimates, and the AVGO AI revenue base are verified.
• Potential trade: modest long AVGO / short INTC after confirming the semiconductor sell-off has stabilized; size as a relative-value position, not a broad AI bet. Reassess if Broadcom reports slowing AI orders or customer capex, or if Intel demonstrates repeatable process yields and meaningful external foundry commitments.
• For AVGO, track hyperscaler capex guidance, custom-accelerator program ramps, networking demand, and customer concentration. A slowdown in orders or a downward guidance revision would falsify the near-term growth thesis; VMware retention and recurring revenue quality should be checked separately rather than assumed to offset chip cyclicality.
• For INTC, treat process milestones, yields, foundry customer design wins, and capital-spending needs as gating evidence. Delays, weak customer adoption, or further deterioration in core earnings would argue against paying for turnaround optionality; sustained execution and customer commitments would invalidate the cautious view.