Is Marvell Stock a Buy on the Dip as AI Revenue Soars?

Marvell reported fiscal Q2 revenue of $2.74B (+37% YoY) and adjusted EPS of $0.94 (+40% YoY), slightly ahead of the $0.93 midpoint on $2.7B revenue. Data center revenue jumped 46% YoY to $2.17B, and management guided fiscal 2027 Q3 revenue of $3.15B (±5%) with adjusted EPS of $1.05–$1.15 and third-quarter data center revenue +75% YoY. The company also raised its fiscal 2027 revenue growth outlook to $12B (+45%) and lifted fiscal 2028 revenue to $18B (+50%), with data center +60% and custom chips projected to more than double—despite the shares dipping on valuation after the strong print.

Analysis

The market is paying for duration, not just the next quarter. MRVL can keep compounding if optical DSP and custom silicon stay on their current trajectory, but the valuation now assumes a clean multi-year ramp with no execution gaps; that is a fragile setup when the most meaningful incremental revenue is still back-half loaded. In the next 1-3 months, the stock likely trades more on hyperscaler capex sentiment than on fundamentals, so even a solid report can disappoint if investors start questioning how much of the AI spend is being monetized by vendors versus internalized by customers.

The clearest winners are the hyperscalers, not the supplier: AMZN, GOOGL, and MSFT can use custom silicon to lower inference cost and widen the gap versus smaller cloud peers, while also retaining bargaining leverage over MRVL. Second-order, that same leverage caps MRVL's long-run gross margin expansion and makes its revenue more customer-concentrated than the headline growth rate suggests. On the loser side, merchant networking and optical peers face a tougher selling environment if hyperscalers keep vertically integrating more of the stack; the risk is less about immediate demand and more about pricing power shifting up the chain.

Contrarian view: consensus is treating the growth revision as de-risking, but the real issue is timing mismatch. The Alphabet win is too far out to support the current multiple, so the upside case depends on no slowdown in interim AI spend and no pushout in custom-chip deployment. What would falsify the bearish overvaluation case is either a faster-than-expected conversion of the Alphabet program into revenue before FY2029, or another upward revision to FY2028 revenue/margin guidance; absent that, any disappointment in optical mix or hyperscaler order cadence could compress the stock 15-20% even without an earnings miss.

Key Decisions for Investors

• Do not chase MRVL after the recent run; wait for a 10-15% pullback or for another quarter of confirmation that data-center growth is translating into durable margin expansion before adding long exposure.
• Initiate a 3-6 month MRVL put spread on strength if the stock re-rates above ~34x FY2028 earnings; thesis is multiple compression, not a near-term revenue miss, so defined-risk downside is preferable to outright shorting.
• Pair trade: long GOOGL vs short MRVL for 1-3 months. Rationale: GOOGL captures the economics of custom silicon adoption and can absorb capex more efficiently; MRVL remains the supplier with longer monetization lag and greater pricing risk. Falsify if MRVL converts the Alphabet program into visible bookings sooner than expected.
• Overweight NVDA on pullbacks relative to MRVL for AI infrastructure exposure. NVDA still has superior pricing power and faster monetization of AI demand; MRVL's upside is more dependent on long-dated customer wins and network architecture shifts.