Piper Sandler initiates Marvell stock with overweight rating on datacenter growth

Piper Sandler initiated Marvell Technology with an Overweight rating and a $270 price target, projecting EPS to compound at roughly 45% annually to $19 by calendar 2030. The bullish thesis centers on Marvell's datacenter connectivity franchises, hyperscaler relationships and a reported $120 billion Google agreement supporting TPU attach programs beginning in fiscal 2029. Marvell's Q2 revenue of $2.739 billion and $0.94 EPS modestly exceeded expectations, while Q3 guidance also surpassed estimates on data-center strength; however, some analysts remain cautious given the stock's elevated valuation and 79x P/E.

Analysis

MRVL’s upside case is increasingly a multi-year AI-interconnect duration trade rather than a near-term semiconductor-cycle trade. The key valuation question is whether custom connectivity revenue earns software-like visibility or remains project-based, customer-concentrated hardware revenue; the latter would cap the terminal multiple well below the level implied by long-dated earnings projections. A material Google program would also tighten MRVL’s dependence on one hyperscaler’s TPU roadmap, making quarterly design-win commentary and program ramp timing more consequential than aggregate data-center demand.

The October 6 analyst day is a near-term catalyst, but expectations appear elevated after a cluster of upward target revisions. The stock needs quantified evidence on backlog conversion, gross-margin trajectory, and customer-funded development economics to avoid a "show-me" reaction; vague long-term TAM expansion is unlikely to support further rerating. Over the next 1-3 months, MRVL can outperform if management bridges custom-silicon/connectivity revenue to fiscal 2028-29 with credible margins; absent that bridge, consensus revisions may be too far ahead of realizable earnings.

Second-order beneficiaries of accelerated optical attach adoption include COHR and LITE, while AVGO is the most direct competitive risk in custom ASIC/connectivity given its entrenched hyperscaler relationships. The contrarian view is that the market may be underpricing execution risk in co-packaged optics and customer qualification cycles: these architectures can expand content per cluster but shift revenue recognition later than investors expect. A delay of even two quarters in a concentrated program would matter disproportionately to MRVL’s forward multiple.

Key Decisions for Investors

• Maintain MRVL as a catalyst watch rather than chase before October 6; initiate only if management provides revenue and gross-margin milestones through fiscal 2028. Use a 8-10% stop below the post-event support level, with a 15-20% upside target on credible estimate revisions.
• Pair trade for the analyst-day window: long MRVL / short AVGO in equal dollar amounts only if MRVL’s relative valuation remains below AVGO on forward AI-infrastructure growth. The thesis is incremental custom-connectivity share; exit if MRVL cannot identify program timing or if AVGO discloses comparable design-win momentum.
• For lower single-name risk, accumulate COHR or LITE on confirmation that optical attach deployments are moving from qualification to volume production. Do not treat MRVL commentary alone as confirmation; require orders, datacenter revenue guidance, or margin evidence from the optical suppliers.
• Set an earnings-revision alert: reduce or avoid MRVL if next-year consensus EPS stops rising after the analyst day, or if data-center guidance implies growth is being driven by inventory/one-time program milestones rather than recurring production ramps.