Marvell stock sinks 8%: is its $120B Google deal taking too long to pay off?

Marvell shares fell nearly 8% premarket despite a solid fiscal Q2: revenue rose 37% to $2.74B and beat expectations. The company also raised guidance to about $12B revenue for fiscal 2027 and about $18B for fiscal 2028. The selloff reflects investors focusing on the timing of the Google AI-chip opportunity rather than near-term execution.

Analysis

This looks less like a fundamental miss and more like a duration problem: the market is refusing to pay today for revenue that only becomes visible in later fiscal years. That usually happens when investors think the path from design win to meaningful shipment is long, customer-concentrated, and subject to qualification delays; in that setup, out-year guidance supports the model but does not support the multiple. If the ramp slips even one quarter, the stock can de-rate sharply because the near-term numbers still won’t justify AI premium valuation.

The cleaner beneficiary is GOOGL, which gains lower inference cost and better internal economics from custom silicon even if the supplier’s revenue recognition is back-end loaded. The second-order loser is the broader merchant AI semis basket: if hyperscalers keep leaning into bespoke silicon, it pressures the narrative for NVDA and, to a lesser extent, AVGO on share capture at the margin, even while overall AI capex remains healthy. That said, this is not a demand destruction signal; it is a monetization-timing signal.

Contrarian risk: the selloff may be overdone if investors are extrapolating timing noise into structural doubt. Custom chip programs often look disappointing right until ramp inflects, because revenue lags tape-out and deployment by multiple quarters. The key falsifier is not the stock price but the next 1-2 quarters of AI-related sequential growth and any downgrade to the long-range revenue bridge; if that bridge holds, the gap-down becomes a buying opportunity rather than a warning.

Key Decisions for Investors

• Do not chase MRVL on the headline gap; wait for the next two quarters of evidence that Google-related revenue is converting on schedule. If management cannot show sequential AI acceleration, expect further multiple compression rather than earnings-led support.
• Pair trade: long GOOGL / short MRVL for 1-3 months to express platform-margin capture versus supplier timing risk. Thesis breaks if MRVL starts showing clear quarterly step-ups in AI revenue or if Google commentary confirms an earlier-than-expected ramp.
• Watch list: NVDA and AVGO as indirect reads on hyperscaler custom-silicon substitution. If the market starts rewarding bespoke chips over merchant accelerators, expect relative weakness in GPU-heavy names over the next 3-6 months.
• Optionality only for patient capital: consider a small, deferred MRVL call spread only after channel checks or the next earnings call confirm deployment milestones. The risk/reward is poor until the market sees evidence, not guidance, of shipments.