Qualcomm Vs. Marvell: The Better AI Business Is The Worse Stock

Qualcomm is rated Buy with a FY2029 price target of $298, implying 69% upside, based on superior risk-adjusted return potential versus Marvell. While Marvell has a stronger current AI business, its valuation is viewed as pricing in near-flawless execution; its cited $120B Google opportunity is not committed backlog and carries uncertainty around revenue timing, margins, and warrant-related dilution.

Analysis

The relevant divergence is not AI exposure but the quality and duration of earnings embedded in valuation. MRVL is increasingly a levered expression of a small number of custom-silicon and optical design wins; any shift in hyperscaler deployment cadence, internal ASIC substitution, or product-mix-driven gross-margin pressure can produce a disproportionate FY27/FY28 estimate reset. AVGO is the more important read-through: sustained custom-ASIC share gains or improving AI networking attach rates at AVGO would challenge MRVL's addressable profit pool, not merely its revenue timing.

QCOM has a more balanced optionality profile: handset recovery, edge-AI content gains, automotive/IoT scaling, and potential licensing resilience provide multiple paths to earnings delivery. The market is likely underpricing the possibility that on-device AI expands premium Android silicon content without requiring a wholesale unit-volume recovery; this would support mix-led margin expansion over the next 6-18 months. Near-term, however, QCOM remains exposed to China handset demand and Apple modem transition risk, so upside requires evidence that non-handset growth can offset those headwinds.

The contrarian view is that neither name should be treated as a pure AI beta. MRVL's key risk is a de-rating from elevated expectations rather than an outright demand collapse, while QCOM's rerating catalyst is proof that diversification is reducing its handset multiple discount. The next two earnings cycles matter more than long-dated AI TAM estimates: watch AI revenue conversion, gross-margin guidance, customer concentration commentary, and diluted-share-count changes.

Key Decisions for Investors

• Initiate a 3-6 month beta-adjusted pair: long QCOM / short MRVL. Express equal dollar exposure initially and rebalance for realized beta; the thesis is relative estimate durability rather than directional semiconductor risk. Cover the MRVL short if its next two reports show accelerating AI revenue with stable or expanding gross margin and no adverse diluted-share-count revision.
• Add QCOM on post-earnings weakness only if handset commentary is soft but automotive, IoT, and licensing guidance remains intact. Target a 15-20% relative outperformance versus SOXX over 6-12 months; invalidate if management cuts full-year EPS guidance or signals a material acceleration in Apple-related revenue loss without offsetting design wins.
• Use AVGO results and hyperscaler capex commentary as an alert for the MRVL leg. A material upside revision to AVGO custom-silicon revenue, coupled with weaker MRVL AI-margin guidance, is a catalyst to increase the pair; conversely, evidence of incremental MRVL customer diversification warrants reducing the short.
• Avoid long-dated MRVL calls until warrant economics, revenue recognition timing, and gross-margin contribution from its major AI programs are independently quantified. The missing data are sufficiently material that upside participation currently offers inferior risk/reward to QCOM exposure.