AMD or Broadcom: If I Could Only Own 1 AI Chip Stock Through 2030, Here's My Pick

AMD reported Q2 revenue up 50% year over year to $11.5B, data-center sales up 107% to $6.7B, and non-GAAP EPS up 246% to $1.66; it estimates a $220B CPU market and $1.4T AI-accelerator opportunity by 2030. Broadcom’s fiscal Q3 2026 revenue rose 86% to $29.6B, while AI semiconductor revenue surged 221% to $16.7B; it projects $115B in AI semiconductor revenue for fiscal 2027 and $230B in AI revenue for fiscal 2028. The article gives Broadcom a slight edge based on its 19.4x forward earnings multiple versus AMD’s 39.5x, while noting potential margin pressure at Broadcom and uncertainty over how long the AI tailwind will last.

Analysis

The key distinction is not “AI exposure” but who controls workload economics. Custom ASIC demand can expand hyperscaler compute capacity while taking selected workloads—and bargaining leverage—away from NVIDIA (NVDA); it does not imply broad GPU substitution. Broadcom (AVGO) benefits if multi-year customer programs convert into durable deployments, but customer concentration and the shift toward lower-margin hardware can make revenue growth a poor proxy for earnings quality. Watch gross-margin trajectory and customer-program execution, not AI revenue alone.

AMD’s (AMD) agentic-AI thesis is more indirect: additional inference and autonomous workloads could raise CPU demand, but that only accrues if CPU content and pricing improve rather than being absorbed by existing server platforms or competing architectures. Intel (INTC) is exposed to share loss, though AMD’s market opportunity estimates are not evidence that share gains will persist. META and Alphabet (GOOG/GOOGL) may gain from lower-cost, workload-specific silicon, but could face higher switching and supply risks if custom-chip programs underdeliver.

Near term, earnings revisions and hyperscaler capex commentary matter more than long-dated TAM claims. Over 6–18 months, the central test is whether AVGO converts commitments into profitable shipments and whether AMD demonstrates measurable CPU demand beyond the current AI cycle. The cited forward multiples are source-provided and should be independently checked on comparable earnings definitions; they are not sufficient grounds by themselves for a valuation trade. A reversal in customer schedules, AI spending, or margins would undermine the bullish case.

Key Decisions for Investors

• Watchlist: consider a beta-adjusted long AVGO / short AMD pair only after independently validating the forward-earnings valuation gap and confirming AVGO’s AI program outlook is not being offset by gross-margin compression. Size modestly; the relative-value thesis fails if AVGO guidance or margins weaken faster than AMD’s.
• Ahead of the next earnings updates, track AVGO AI-segment growth alongside consolidated gross margin and customer concentration commentary. Treat growth without margin or cash-conversion support as lower-quality; reassess on program delays or guidance cuts.
• For AMD, require evidence that CPU demand is translating into incremental server sales, pricing, or market-share gains before paying for the agentic-AI upside. A slowdown in data-center growth or renewed competitive pressure from INTC would falsify the near-term thesis.
• Do not treat ASIC adoption as an automatic short of NVDA: diversified hyperscaler spending can support both custom silicon and GPUs. Revisit only if customer disclosures show sustained workload migration and weaker GPU-related spending.