Marvell Has Raised Its Growth Outlook 3 Times Since December. The Stock Still Sits 28% Below Its High.

Marvell raised its fiscal-year revenue outlook to about $12 billion and its following-year outlook to roughly $18 billion, implying approximately 45% and 50% growth, respectively. Fiscal Q2 revenue reached a record $2.739 billion, while data-center revenue rose 46% year over year to $2.17 billion, or 79% of total sales; Q3 revenue guidance of $3.15 billion implies about 52% growth and adjusted EPS guidance is $1.10 versus $0.94 in Q2. The bullish AI-driven growth case is tempered by Q3 non-GAAP gross-margin guidance of 57.5%-58.5%, below 58.9% in Q2, increased custom-chip competition following Qualcomm's Amazon collaboration, and a valuation of roughly 57x current-year expected earnings.

Analysis

MRVL’s equity debate is shifting from AI demand validation to the quality of that revenue. Custom silicon and associated interconnect content can grow revenue rapidly while diluting mix; a sustained sub-58% gross-margin profile would make consensus EPS revisions materially less powerful than revenue revisions and cap multiple expansion. The key incremental variable for the next 1-3 months is whether operating leverage offsets mix dilution, not another top-line guide increase.

The more consequential competitive risk is cloud customers deliberately creating second sources. QCOM’s entrance validates a larger custom-AI-silicon TAM, but it also weakens the assumption that MRVL captures a stable share of each hyperscaler program. ASIC design wins are typically sticky once production ramps, yet next-generation socket allocation can change 12-24 months before reported revenue does; watch disclosed design-win count, customer concentration, and backlog conversion rather than management’s aggregate TAM framing.

The selloff may be directionally rational but potentially over-discounting near-term earnings durability if the current production ramps are already funded and qualification risk has passed. A re-rating requires evidence that gross margin bottoms while data-center growth remains elevated; absent that, MRVL can remain a high-growth but de-rating-prone semiconductor supplier. Thesis is falsified by another material margin-guide reduction, a hyperscaler capex slowdown, or QCOM/other ASIC vendors announcing production-scale wins at a current MRVL customer.

AMZN benefits strategically from dual sourcing even if it does not internalize economics immediately: supplier competition improves ASIC pricing, roadmap control, and supply assurance. NVDA is not a direct loser from isolated custom-chip programs, but broader ASIC adoption pressures the premium portion of its data-center multiple if hyperscaler workloads prove more substitutable than investors expect over the next 6-18 months.

Key Decisions for Investors

• Initiate a starter long MRVL only on confirmation that gross-margin guidance has stabilized or improved sequentially; use a 3-6 month horizon and size modestly given high earnings-duration risk. Target a recovery toward the prior valuation range if estimates continue rising; exit on a further 100bp-plus gross-margin-guide cut or a material reduction in next-year revenue outlook.
• Express the near-term quality concern through a relative-value trade: long MRVL / short SOXX in equal beta-adjusted dollars after the next earnings print if MRVL confirms margin stabilization. This isolates company-specific estimate upside from broad AI-semiconductor multiple risk; close if MRVL underperforms SOXX by 10% after a clean print or if sector capex guidance weakens.
• Do not short QCOM solely on its AI-custom-silicon announcement. Set an alert for disclosed production volumes, customer economics, and whether the program displaces rather than supplements incumbent suppliers; without those data, the financial impact is not sufficiently measurable.
• Monitor AMZN capex commentary and custom-silicon deployment metrics through the next two earnings cycles. A capex acceleration paired with broader ASIC deployment is constructive for MRVL’s addressable market but negative for its pricing power if supplier diversification becomes explicit.