Marvell Grants Google Warrants for Up to $12.2 Billion in Stock in Expanded Custom AI Chip Deal
The AMW Read
An equity-warrant structure tying Marvell's economics to Google's custom silicon roadmap is a novel deal mechanic that deepens a known hyperscaler-chipmaker relationship with segment-level competitive implications for the custom-ASIC supply chain.
Marvell Grants Google Warrants for Up to $12.2 Billion in Stock in Expanded Custom AI Chip Deal
Marvell has granted Google warrants to purchase up to $12.2 billion of its stock as part of an expanded custom AI chip supply agreement between the two companies. The equity component ties Marvell's compensation more directly to the scale and duration of the chip supply relationship rather than a conventional purchase-order arrangement, and signals a deepening of Google's custom silicon roadmap for its AI data centers.
Custom silicon has become central to how hyperscalers manage AI infrastructure costs, and Google is among the more aggressive builders of application-specific chips for its own data centers, working alongside merchant suppliers such as Marvell and Broadcom. Structuring warrants into a chip supply deal gives Google potential equity upside in a key vendor while giving Marvell a longer runway of committed demand — a mechanic that echoes a broader pattern of hyperscalers using their balance sheets to lock in scarce compute-adjacent supply. Per the AI Market Watch index, Marvell-related coverage in our pipeline rose to five items in the last 90 days versus one in the prior 90, name-matched over pipeline-ingested sources only, consistent with an accelerating cadence of Marvell-Google dealmaking.
For investors, the warrant structure effectively makes Google a contingent shareholder in one of its own key chip suppliers, aligning incentives around volume while concentrating more of Marvell's equity story around a single hyperscaler relationship. Builders and investors tracking the custom-ASIC supply chain should watch whether equity-linked supply agreements spread to other hyperscaler-chipmaker pairs, since that would mark further consolidation of AI chip demand around a small set of preferred vendors.