Marvell Q2: I Agree With The Market

Marvell (MRVL) reported Q2 revenue up 37% YoY and slightly beat both revenue and EPS expectations. Data center revenue grew 46% YoY, while management guided to >60% growth in fiscal 2028 supported by strong AI and custom ASIC demand, and non-GAAP operating margin improved to 36.6%. Despite GAAP dilution concerns from substantial stock-based compensation, the outlook and margin trajectory underpin a bullish setup for the stock.

Analysis

The real takeaway is not just that AI demand is intact, but that hyperscalers still appear willing to fund a second silicon stack alongside merchant GPUs. That is structurally positive for MRVL and, by extension, other custom-silicon and networking beneficiaries such as AVGO, because the spend migrates from generic compute toward application-specific silicon, interconnect, and packaging-heavy content. The loser set is subtler: if custom ASIC adoption keeps accelerating, it gradually pressures the assumption that NVDA owns all incremental AI capex growth, even if the transition is measured rather than abrupt.

Near term, the market will likely focus on whether the growth trajectory is backed by contracted design wins or just a strong demand read-through. The key risk is that AI capex remains lumpy and customer concentration can turn a great quarter into a difficult comp cycle within 1-2 quarters; if data center growth decelerates materially or margin expansion stalls, the multiple can compress quickly. The stock-based comp issue matters because GAAP earnings and per-share value creation are meaningfully weaker than the headline non-GAAP optics; that caps how much of the beat should be capitalized into a higher long-duration multiple.

Contrarian view: consensus may be over-extrapolating fiscal 2028 commentary into a clean straight line, while underweighting dilution, customer concentration, and the possibility that some AI spending shifts from growth to optimization. I would treat this as a quality bullish setup, but not a blanket buy-the-dip story unless the next two quarters confirm sequential data-center acceleration and convert commentary into backlog, not just TAM rhetoric. The upside is real, but the burden of proof is now on sustained order flow, not another one-quarter beat.

Key Decisions for Investors

• Stay long MRVL on pullbacks, but only if the next earnings cycle confirms continued data-center acceleration; thesis weakens if data-center growth drops sharply or management softens the fiscal 2028 trajectory.
• Pair trade: long MRVL / short NVDA for 1-3 months if you want to express the custom-ASIC substitution theme; this works best if hyperscaler capex shifts toward bespoke accelerators and networking rather than merchant GPUs.
• Use AVGO as the cleaner quality hedge against MRVL execution risk: long MRVL vs. short a small basket of higher-quality semi names if you believe MRVL has more operating leverage to AI content expansion, but smaller margin for error.
• Do not chase on the initial gap-up; wait for the first post-earnings consolidation or the next sell-side note cycle before adding risk, since the current move may already discount the near-term beat.
• Watch SBC as the falsifier: if dilution remains elevated while GAAP profitability lags, treat the rally as a multiple event rather than a durable compounding story.