From $1.5B to $30B: What Marvell's 13% Custom-Chip Target Requires

At Marvell's October 6, 2026 Investor Day, analyst Tore Svanberg asked a direct question: with relatively few custom-chip suppliers, why did the company target only 13% share? CEO Matt Murphy confirmed the arithmetic. The approximately $30 billion FY2031 custom-revenue midpoint divided by Marvell's roughly $235 billion calendar-2030 custom-market estimate produces about 13%. These figures describe a company plan; they do not measure actual market share.

The starting point reveals the growth requirement. Management put FY2026 custom revenue at about $1.5 billion, making the $30 billion target a twentyfold expansion. Its forecast of roughly 55% annual custom-market growth from CY2025 to CY2030 compounds to about 8.95 times the starting market. The company's FY2026 to FY2031 custom-revenue growth range is 75% to 85%. The fiscal and calendar windows do not align exactly, but the direction is clear: custom revenue must expand faster than the market Marvell models. Treating 13% alone as evidence of conservatism misses that requirement.

Murphy described selective participation in a few long-term flagship full-custom accelerators, or XPUs, alongside a broader set of designs for chips working around them. These XPU-attach products include network controllers tailored to a cloud customer's data paths, and chips extending memory or storage capabilities. The statement describes the portfolio overall; it does not disclose a separate flagship-program count for every customer.

Reuse is the attraction of the attach model. Marvell says these semi-custom products build on existing IP already shipped at scale, and that in many cases it works with customers on architecture and leads front-end design. An attach function may also persist across accelerator generations, extending its opportunity beyond XPUs designed by Marvell itself. Both reuse and persistence depend on continued customer adoption. They do not establish locked-in future revenue; buyer bargaining, competing products and production timing still affect conversion.

Current-year guidance provides a nearer test than the five-year target. Reported FY2027 first- and second-quarter consolidated revenue was $2.4178 billion and $2.7393 billion, totaling $5.1571 billion in H1. Using management's approximately $12 billion full-year guidance implies an H2 requirement of about $6.8429 billion. Prior-year H2 actual revenue was $4.2932 billion, making the required year-over-year increase approximately 59.4%.

The latest reported quarter grew approximately 36.5% year over year. A quarterly rate and a half-year rate cover different windows; comparing them makes the required H2 acceleration visible. This is a consolidated-company calculation, not a custom-silicon segment forecast: interconnect, switching and storage contribute to the total. Once Q3 revenue is reported, subtracting it from the $6.8429 billion requirement will show what Q4 must deliver.

Management says its FY2031 forecast includes more than 35 sockets at the top four US hyperscalers. That count places the plan across multiple product programs. It does not disclose the value, production date or irrevocable purchase commitment of each program, so it cannot simply be translated into an order backlog. Design adoption, production and recognized revenue remain distinct developments to track.

The useful tests work at two levels: near-term reported company revenue against current guidance, and subsequent disclosure of custom revenue, attach mix and named production ramps. A revised market definition or target also requires recalculating the ratio under the new basis. The 13% alone cannot explain competitive strength. The route from $1.5 billion to $30 billion and which products start contributing to it gives the strategy a concrete, testable substance.