Marvell: Surprisingly Cheaper Than In August, But There's A Catch

Marvell Technology maintains a Strong Buy rating after raising its fiscal 2028 revenue outlook to $18 billion, supported by broad-based connectivity growth. Operating margin is targeted at 38%-40% by Q4, while fiscal 2028 EPS is projected at $6.30-$6.70, improving the stock's forward valuation. Scale-up optics and near-packaged optics provide additional upside, with optics revenue expected to exceed $300 million in fiscal 2028 and ramp further in fiscal 2029.

Analysis

The investable question is not the long-dated revenue target but whether AI connectivity mix can sustain gross-margin expansion before the market has to underwrite fiscal-2028 assumptions. At a 38-40% operating-margin endpoint, modest execution misses have outsized EPS sensitivity: a 200bp margin shortfall on a multi-year revenue ramp would likely remove roughly 8-12% from normalized earnings power and compress the premium multiple assigned to AI infrastructure suppliers. The near-term catalyst path is quarterly evidence of data-center revenue mix, custom-silicon attach rates, and backlog conversion rather than another long-range target increase.

Near-packaged optics is strategically valuable but financially immaterial at its initially indicated scale relative to the broader revenue base; it should be treated as a fiscal-2029 optionality driver, not a reason to pay for current earnings. The second-order beneficiary set includes laser and optical-component suppliers such as COHR and LITE if optical interconnect intensity rises, while incumbent pluggable-optics vendors face architecture-transition risk if near-package designs gain hyperscaler qualification. The competitive risk is that AVGO and NVDA can bundle switching, silicon and systems more tightly, limiting Marvell's ability to retain margin as customers seek lower total cluster cost.

Consensus appears likely to extrapolate AI-networking demand in a straight line. The reversal scenario is a hyperscaler digestion cycle, delayed optical qualification, or a customer concentration disclosure that shows design-win revenue arriving later than expected; those risks can hit orders within 1-3 quarters while the long-term narrative remains intact. For the next 6-18 months, the critical falsifiers are data-center growth deceleration, operating-margin guidance failing to progress sequentially, or evidence that incremental AI revenue requires greater pricing concessions than management's model implies.

Key Decisions for Investors

• Accumulate MRVL only on post-results weakness rather than chase long-dated guidance; size as a 6-18 month AI-connectivity long and require the next two reports to show sequential data-center growth plus operating-margin progression. Exit or reduce if margin guidance stalls more than 200bp below the stated endpoint.
• Use a 12-18 month MRVL call spread rather than outright short-dated calls to express upside from design-win conversion; select strikes after confirming implied volatility and target a structure with at least 2:1 upside-to-premium-at-risk. This is preferable because qualification timing can slip by quarters without invalidating the structural thesis.
• Monitor COHR and LITE as higher-beta read-throughs to optical intensity, but do not treat them as direct substitutes for MRVL: initiate only after orders/backlog confirm datacenter optical demand rather than relying on architecture announcements.
• Set an earnings watch item for customer concentration, custom-silicon backlog, and optical revenue contribution. If hyperscaler capex guidance weakens or MRVL signals delayed production ramps, hedge the position with a short SOXX or reduce gross exposure; the likely drawdown would occur before fiscal-2028 estimates are formally revised.