Marvell: Rags To Riches Story Not Over

Marvell Technology has surged more than 220% since September, helped by stronger data center demand that now makes up 76% of total sales. Management lifted FY27 and FY28 outlooks, with interconnect revenue expected to grow 70% in FY27 and custom ASIC revenue projected to double in FY28. The improving growth profile reduces reliance on hyperscaler capex cycles and supports a more durable outlook for MRVL.

Analysis

MRVL is transitioning from a cyclical semi supplier to a more contract-like infrastructure compounder. The key second-order effect is not just higher revenue, but lower earnings beta: custom ASIC and interconnect exposure should make the next 12-24 months less sensitive to generic AI capex pauses because design wins are embedded earlier and revenue is tied to deployed sockets rather than broad ordering waves. That shift should compress the market’s perceived “AI hardware volatility discount,” which can justify a premium rerating even if headline growth decelerates.

The bigger winner set is likely the AI infrastructure ecosystem around MRVL’s customers. Hyperscalers get a lower-cost path to scale vs. off-the-shelf accelerators, while optical/packaging/power supply chains benefit from sustained high-density networking and custom silicon content per cluster. Competitively, this pressures merchant silicon and weaker networking vendors whose products are easier to substitute, because custom ASIC programs tend to lock in multi-year engineering dependencies and raise switching costs after tape-out.

The main risk is not near-term demand, but expectation saturation over the next 2-4 quarters. After a move of this magnitude, any sign of FY27/FY28 underdelivery, program slippage, or a slower-than-expected ramp in one hyperscaler can trigger multiple compression even if fundamentals remain strong. The most important watch item is whether customers keep diversifying away from one-off ASICs back toward more standard platforms if AI spend moderates; that would reintroduce cyclicality and weaken the de-risking narrative.

Consensus may be underpricing how much of the upside is already in the stock, but may still be underappreciating the duration of the earnings stream. In other words, the easy money may be gone on multiple expansion, yet the setup can still work on time arbitrage: a company with improving visibility can sustain high relative performance even if absolute upside slows. The contrarian risk is that investors mistake better visibility for perpetual outperformance; once the market fully discounts FY28, returns likely become more dependent on execution than narrative.

Key Decisions for Investors

• Maintain a tactical long MRVL, but size it below full conviction after the 220% run; use 1-3 month pullbacks to add, with a stop on any guidance reset or tape-out delay that reopens cyclical risk.
• Use call spreads rather than outright longs for the next 6-12 months to express upside while reducing multiple-compression risk; this is a cleaner way to stay involved after the rerating.
• Pair long MRVL vs. a basket of more merchant-silicon/networking-exposed semi names over 3-6 months to isolate the secular custom-ASIC de-risking theme.
• If you want to fade the move, do it with a small-sized put spread 2-4 quarters out rather than a naked short; the stock can stay expensive if hyperscaler order visibility remains intact.
• Monitor hyperscaler capex commentary and any evidence of program concentration; if one customer begins to dominate incremental growth, trim exposure because the market will reprice concentration risk quickly.