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 forecast roughly $18 billion for the following year, supported by accelerating AI data-center demand. Fiscal Q2 revenue reached a record $2.739 billion, up 37% year over year, while data-center revenue rose 46% to $2.17 billion, representing 79% of total sales; Q3 guidance implies approximately 52% total revenue growth and adjusted EPS of $1.10. Risks include gross-margin pressure, with Q3 non-GAAP margin guided to 57.5%-58.5% from 58.9%, increased custom-AI-chip competition after Qualcomm's Amazon collaboration, and a still-elevated valuation of about 57x current-year earnings.

Analysis

MRVL’s central valuation question is not demand durability but conversion quality: incremental custom-silicon revenue can grow rapidly while contributing less gross profit per dollar than its legacy networking mix. If gross margin settles below 58% rather than rebounds as volumes scale, the market is likely to treat MRVL as a lower-multiple ASIC integrator rather than a scarce AI infrastructure supplier; that distinction can cap multiple expansion despite continued estimate revisions.

The more consequential competitive signal is hyperscaler multi-sourcing. AMZN’s willingness to add QCOM increases its negotiating leverage across future chip generations, and similar supplier diversification could pressure MRVL’s pricing, NRE recovery, and attachment rates in optical/interconnect. The near-term test is the next earnings report: upside requires both another revenue guide raise and evidence that mix dilution has bottomed; revenue upside alone may no longer be enough to re-rate the shares.

Consensus appears too focused on whether the forward revenue target is achievable and insufficiently focused on whether custom programs create repeatable, high-return platforms or one-off capacity ramps. Over 6-18 months, successful in-house or alternative-vendor deployments at cloud customers would shift value toward AMZN and competing ASIC suppliers, while MRVL bears utilization and margin risk. Conversely, stable margins alongside continued program wins would demonstrate that the current drawdown has already discounted the competitive concern.

Key Decisions for Investors

• Initiate a measured long MRVL only ahead of the next earnings print if shares remain below the prior post-earnings gap area; target a 15-20% rebound over 1-3 months on a revenue-guide raise plus gross-margin stabilization. Exit if forward non-GAAP gross-margin guidance falls below 57% or management fails to lift the following-year outlook.
• Use a relative-value structure: long MRVL / short SOXX in equal dollar amounts for 3-6 months. This isolates potential company-specific estimate revisions from broad AI-semiconductor multiple compression; close if MRVL underperforms SOXX by 10% following a guide increase, which would indicate the market is repricing structural margins rather than discounting a temporary issue.
• Do not short QCOM solely on its cloud-custom-chip announcement. Treat it as a watch item until disclosed design-win economics, production timing, or material data-center revenue contribution establish that it can displace MRVL rather than merely become a second source.
• Monitor AMZN capex commentary and any disclosed supplier concentration changes over the next two quarters. A capex deceleration or evidence of meaningful multi-vendor allocation is the key falsifier for MRVL’s 6-18 month revenue and margin thesis.