This chip stock has more than tripled in the past year. Analysts say buy more after latest earnings

Marvell Technology delivered a fiscal Q1 earnings and revenue beat, along with better-than-expected fiscal Q2 data center revenue guidance and a profit forecast above consensus. Several analysts raised price targets sharply, including Barclays to $275 from $150 and Deutsche Bank to $240 from $120, though Morgan Stanley and Goldman stayed cautious. The stock had already surged 207.6% over the past 12 months and was down 3% in premarket trading despite the upbeat report.

Analysis

The key market read is not simply that MRVL printed well, but that the market is still underappreciating the duration of AI networking capex. When a supplier in the data-center interconnect stack can lift guide again after already re-rating hard, it suggests the spend is broadening from a few hyperscaler “proof points” into a more sustained buildout across optics, switching, and custom connectivity. That is bullish for the entire high-speed networking supply chain, but especially for names with genuine design wins and less exposure to generic accelerator ASP pressure.

The second-order implication is margin mix: if custom silicon ramps are still ambiguous while networking content rises, MRVL’s upside path likely comes from attach-rate expansion rather than a single blockbuster program. That favors suppliers with multi-socket exposure and penalizes vendors tied to one or two large custom compute ramps, because investors will pay up for visibility and punish concentration risk. It also means the next leg of the trade may come from follow-through estimate revisions across component and test/equipment vendors, not just further multiple expansion in MRVL.

The contrarian risk is that the stock is now trading against very high expectations, so the next disappointment could be about slope, not absolute demand. If hyperscaler budgets normalize or lead times compress, the market may rotate from “beats” to “proof of sustained share gains,” which is exactly the bar the skeptics are setting. In that setup, the trade becomes vulnerable over the next 1-3 quarters even if secular AI demand remains intact.

The cleaner risk/reward may be in expressing the theme through a relative-value pair rather than outright chasing MRVL after a >200% one-year move. The best upside is in names with similar AI exposure but less crowded positioning and more obvious estimate revision potential; the downside is that if networking demand slows, those same names will de-rate fastest. Short-dated upside calls can work only if you expect another guide-up within the next earnings cycle; otherwise the premium burn is likely to exceed the incremental fundamental surprise.

Key Decisions for Investors

• Prefer a relative-value long MRVL / short a weaker custom-silicon or AI-infrastructure peer over a cash long; 1-2 quarter horizon, aiming to capture continued networking-share upside while reducing multiple-risk from MRVL’s crowded ownership.
• Buy call spreads on MRVL into any post-earnings pullback rather than chase strength; 1-3 month expiry, targeting a move driven by the next estimate revision cycle, with defined downside if the market shifts to ‘show me’ mode.
• Go long a basket of high-speed interconnect beneficiaries versus semis with lower AI content; 2-4 quarter horizon, because the second-order beneficiary is broadening content in optics, switches, and test/validation rather than pure accelerator demand.
• Fade the most expensive AI-infrastructure names with the weakest customer visibility; 3-6 month horizon, since the article’s real message is that markets are rewarding visibility more than narrative, and names without repeatable guide-ups may underperform on the next data point.
• Set a tactical trim rule on MRVL upside into further multiple expansion: take partial profits if the stock rerates without a commensurate raise in forward revenue visibility, because the easiest money here may already be behind us.