⚠ The Foundry Sets the PriceModerate threat

Marvell owns no fabs. Every dollar of its $8,194.6 million of fiscal 2026 revenue came from wafers someone else manufactured¹, almost all of it from a single foundry at the leading edge. That model has been a genuine advantage — no capital, no cycle risk on plant, access to the best process available — and it comes with a dependency Marvell cannot design around.

Leading-edge capacity is allocated, not simply purchased, and the allocation favours volume. Marvell is bidding for the same 3nm and 2nm wafers as Apple, Nvidia, AMD and Qualcomm, several of which buy an order of magnitude more. In a constrained year the smaller buyer waits, and waiting at the newest node is the one cost this business genuinely cannot absorb: the whole optical thesis rests on being first, and being first requires wafers on schedule.

Price is the second half. Leading-edge wafer costs have risen sharply each generation, and a fabless designer's margin is what is left after the foundry has been paid.

Watch gross margin through a capacity-tight period. If it holds while wafer costs rise, Marvell is passing them on and its products are genuinely differentiated. If it compresses, the foundry — not Marvell — is capturing the value of the node.