Broadcom vs. AMD: Which AI Chip Stock Is the Better Buy?

AMD reported 50% YoY revenue growth in Q2, while Broadcom posted 48% YoY sales growth and guided to 84% YoY revenue growth in fiscal Q3, with AI semiconductor revenue more than tripling YoY. The piece frames a potential CPU demand tailwind for AI/agentic deployments (e.g., shifting CPU-per-GPU ratios), favoring AMD, but argues Broadcom is the less risky valuation at 70x P/E vs AMD’s 120x and higher net margin (42% vs 19.9%). Overall, it positions Broadcom as the more compelling setup at current levels despite AMD’s CPU-growth angle.

Analysis

This is a mix-shift story, not a simple “AI up = both up” trade. The incremental dollars from agentic deployments likely flow to the chips that sit closest to rack architecture decisions, so any rise in CPU-per-GPU intensity is a direct earnings lever for AMD and only an indirect one for the broader AI silicon stack. The second-order risk is that higher CPU content raises total system cost, which can force hyperscalers to slow deployment cadence if ROI slips—good for near-term component ASPs, but a governor on unit growth later.

The market may be underestimating how much of AMD’s upside is already embedded after a strong run, while still discounting Broadcom’s ability to compound from a higher-margin base. AVGO’s valuation still gives it more room to absorb a guidance hiccup; AMD needs the CPU thesis to show up in reported data-center mix over the next 1-3 quarters or the multiple can compress fast. For NVDA, any CPU attach-rate increase is not a direct hit, but it can marginally dilute GPU economics by lifting platform cost and elongating customer payback periods.

The contrarian view is that the “CPU boom” narrative is real but timing-sensitive: design wins and rack re-architectures usually take longer to translate than the sellside expects. If agentic AI remains a pilot-stage workload through year-end, AMD’s relative upside could be pushed into 2026, while AVGO keeps harvesting existing custom-silicon demand. Falsifier for the AVGO-over-AMD view: AMD prints sequential data-center acceleration and management cites meaningful CPU share gains in AI servers; falsifier for the AMD bull case: CPU attach does not move materially in the next two earnings calls.

Key Decisions for Investors

• Long AVGO / short AMD as a 1-3 month relative-value pair into the next two earnings cycles; thesis is lower multiple + stronger margin quality versus AMD needing an unproven CPU mix inflection. Cut if AMD reports clear data-center acceleration or AVGO’s AI growth guide slows materially.
• If you want CPU upside exposure, buy AMD on pullbacks only after confirmation from hyperscaler capex commentary or server-channel checks; otherwise the stock looks optically expensive relative to the timing of the catalyst.
• Avoid chasing NVDA on this headline set; the risk/reward is less direct than the AVGO/AMD spread unless you see evidence that higher CPU content is accelerating overall AI system deployments rather than just redistributing spend.
• Set an alert for AMD next-quarter data-center revenue and segment commentary: if growth re-accelerates and management references rack-level CPU attachment, the relative-value short can be closed quickly.
• Watch AVGO’s next AI revenue guide as the key falsifier; if guide-up continues, the stock can keep working even if the broader AI tape cools.