Better Artificial Intelligence Stock: Arm vs. Marvell Technology

Marvell reported FY2026 revenue of $8.2B, up 42.1%, and forecasts sales of $20B by FY2028 and $70B–$90B by FY2031, citing demand for AI solutions. The article favors Marvell over Arm, noting Marvell’s lower forward P/E (65.2x vs. 119.5x) and P/S ratio (25.7x vs. 55.3x), alongside customer concentration and Taiwan manufacturing risks. Arm FY2026 revenue grew 22.8% to $4.9B, but its net margin declined to 18.4% from 19.8% in FY2025.

Analysis

The relative case for MRVL is a bet on sustained hyperscaler spending and rising silicon content per AI rack—not simply on AI demand. That creates upside if networking, interconnect and custom-silicon programs scale, but also makes revenue unusually sensitive to a handful of customers’ deployment schedules. AVGO is a credible competitor for custom silicon and networking dollars; an allocation shift or in-house design decision at a major customer could impair MRVL before aggregate AI spending slows.

The stated sales ambitions are the key catalyst and the key risk. Treat them as management targets, not an independently validated demand curve: verify the product mix, customer commitments, capacity and accounting basis behind the trajectory. Near term, guidance revisions and hyperscaler capex commentary matter more than headline valuation multiples. Over 6–18 months, execution and customer concentration will determine whether growth supports today’s expectations. A pause in orders, customer insourcing, or export restrictions would challenge the thesis; upside requires continued bookings and conversion into reported revenue and cash flow.

ARM offers a different exposure: licensing can benefit from server CPU adoption without owning the full manufacturing burden, but custom architectures and open alternatives threaten its take rate over time. Its premium valuation leaves less room for execution slippage. The contrarian point is that MRVL’s lower relative multiples do not make it cheap if ambitious targets are already embedded; the article’s strong preference understates forecast risk. Also compare dilution and cash generation on consistent definitions—SBC ratios using different denominators are not directly comparable.

Key Decisions for Investors

• Relative-value watch: consider a modest long MRVL / short ARM position only after validating MRVL’s target assumptions against customer commitments and quarterly guidance; size for high customer and forecast risk. Reassess if MRVL bookings or data-center growth decelerate, or ARM raises outlook on server CPU adoption.
• For the next 1–3 months, track hyperscaler capex plans, MRVL design-win conversion, and any evidence of customer insourcing or order deferrals. These are more actionable catalysts than the cited forward P/E comparison.
• Do not treat the FY2028/FY2031 sales targets as base case. Before adding outright MRVL exposure, reconcile reported versus adjusted earnings, revenue mix, customer concentration, and cash-flow/SBC treatment; absent confirmation, no aggressive position is warranted.
• Thesis failure signals: a material MRVL guidance cut, falling data-center bookings, meaningful customer substitution toward AVGO or internal designs, or export restrictions that constrain shipments. Conversely, repeated guidance increases with realized revenue would support scaling the relative position.