Distributor A: The 37% Nobody Can NameThin moat

Marvell's fiscal 2026 10-K names one distributor at 37% of total net revenue. Two years earlier the same relationship was 24%; last year it was 34%¹. No direct customer comes close — the largest is 14%.

This is not what most people picture when they think about who buys Marvell's chips, and the gap between the picture and the filing is the point. Distributors in this industry are not merely logistics: they hold inventory, extend credit, manage the long tail of smaller customers, and in some arrangements sit between Marvell and very large end customers as well. A single one at more than a third of revenue is a genuine dependency, on a counterparty whose own economics and incentives Marvell does not control.

The specific hazard is that a distributor's purchases and its end customers' consumption are different things that can diverge for several quarters. Channel inventory builds when demand is expected and unwinds when it disappoints, and the supplier sees the correction late and all at once. In a cycle as steep as the current AI build-out, that lag is not a technicality.

What makes this specific relationship harder to read is its trajectory: 24% to 37% in two years means the concentration is deepening, not stabilising, and the latest quarterly filing shows it still rising: the same distributor was 45% of revenue in the first half of fiscal 2027 and 44% in the June quarter, against 35% and 34% a year earlier². Watch that percentage in each annual filing. Rising means more revenue arriving through a party Marvell cannot see past; a sudden fall would likely mean an inventory correction already under way.