Semiconductor Sell-Off: 1 AI Chip Stock to Buy, 1 to Hold, and 1 to Sell After Broadcom's Beat

The article rates Nvidia a buy, AMD a hold and Qualcomm a sell, centered on divergent AI-chip growth and valuation profiles. Nvidia generated $96.2B of quarterly revenue, up 106% year over year, and trades at a P/E of 28 despite a projected deceleration to 65% growth. AMD's first-half 2026 revenue rose 44% to nearly $22B, but its 132 trailing P/E and 68 forward P/E leave little room for an AI infrastructure spending slowdown. Qualcomm's handset revenue fell 9% in the first nine months of fiscal 2026 and Apple’s modem transition is expected to halve Qualcomm sales to Apple sequentially in the December quarter, despite 39% automotive revenue growth.

Analysis

The relevant distinction is not GPU demand growth but who captures incremental system-level economics. NVDA's moat remains the software, networking and rack-scale integration layer; a modest accelerator share loss need not materially impair earnings while customers remain locked into its deployment stack. AVGO is the less obvious beneficiary if hyperscalers increasingly favor custom silicon to constrain NVDA pricing, because it supplies the design/IP and connectivity ecosystem behind that substitution.

AMD is a high-beta share-gain vehicle, not a clean AI-spending hedge. Its upside requires both credible customer adoption of full-rack offerings and gross-margin resilience as it prices below NVDA; a product launch alone will not validate the multiple. Over the next one to three quarters, hyperscaler capex commentary, disclosed design wins, and the mix of data-center revenue versus inventory build will determine whether AMD can close the valuation gap or de-rate sharply.

QCOM's low headline multiple should be treated as a transition multiple: loss of a concentrated, high-margin modem revenue stream can create operating deleverage faster than automotive, IoT, or data-center initiatives can absorb fixed R&D. The underappreciated risk is that data-center buyers may view QCOM as another merchant supplier without a differentiated software ecosystem, forcing lower pricing and elevated customer-acquisition costs. Conversely, a named hyperscaler production deployment—not merely evaluation interest—would be the key falsifier of the bearish view.

Consensus likely overstates the read-through from any one semiconductor earnings guide. A digestion phase in AI infrastructure spending would first pressure merchant accelerator vendors and component inventories, while cloud platforms could benefit from improved returns on invested capital. That makes a relative-value posture preferable to a broad semiconductor directional bet over the next 1-3 months.

Key Decisions for Investors

• Maintain a core long NVDA versus AMD pair (long NVDA / short AMD, beta-adjusted) through the next two earnings cycles. The thesis is software-and-system monetization versus share-gain expectations; cover the AMD short if management discloses multiple production-scale rack deployments with sustained gross-margin expansion.
• Add a tactical long AVGO / short SOXX relative trade on evidence that hyperscaler custom-silicon programs are accelerating. Use the next cloud-provider capex updates as the entry gate; exit if AI networking and custom-chip backlog commentary weakens, since this trade depends on architecture diversification rather than aggregate AI demand.
• Underweight QCOM for the next 6-12 months; do not short solely on handset weakness given its valuation support. A more attractive catalyst-based short would be initiated only after confirmation that modem-related revenue losses are not offset by higher-margin automotive/IoT growth or a production hyperscaler data-center win.
• Watch META, AMZN, and AAPL capex and silicon disclosures as demand-quality indicators. Rising capex paired with slowing infrastructure depreciation or improving cloud/advertising margins supports NVDA/AVGO; capex restraint or explicit GPU-utilization normalization would favor reducing semiconductor beta broadly.