Google Secures $12.2B Option to Buy Marvell Shares in AI Chip Deal
The agreement extends a custom silicon partnership through 2033, targeting AI inference and networking products as an alternative to Nvidia.

Google has secured the option to purchase up to $12.2 billion worth of Marvell Technology shares as part of an expanded partnership focused on custom artificial intelligence chips, according to a securities filing disclosed this week.
The arrangement allows Google to buy as many as 58,970,907 Marvell shares at $206.58 each, with the stake tied to purchasing commitments extending through fiscal year 2033. Marvell's stock climbed 8% following the announcement, while shares of Broadcom—Google's longtime custom chip partner—fell approximately 4%.
Expanding Beyond TPUs
The expanded agreement centers on products that integrate with Google's tensor processing unit infrastructure, according to Marvell's filing. The partnership will encompass AI inference accelerators, storage controllers, and network interface controllers—components critical to managing data flow in large-scale AI systems.
This builds on an initial deal reported in April, when The Information first disclosed that Google and Marvell were collaborating on tensor processing units and memory processing units for AI workloads. That earlier announcement had already driven a significant rally in Marvell shares.
The Custom Silicon Strategy
Google's deepening relationship with Marvell represents part of a broader industry shift among hyperscale cloud providers. Amazon, Meta, and Microsoft have similarly invested in developing custom chips tailored to their specific AI infrastructure needs, seeking cost advantages and performance optimizations that general-purpose processors cannot deliver.
For the past decade, Google has primarily worked with Broadcom on custom silicon development. The companies expanded that partnership in April as well, though the Marvell agreement signals Google is diversifying its supplier base for AI-specific components.
The strategy reflects mounting pressure to reduce dependence on Nvidia, whose GPUs have become the industry standard for training and running large AI models but command premium prices and face persistent supply constraints.
Why it matters
This equity-linked chip deal illustrates how the economics of AI infrastructure are reshaping semiconductor partnerships. By tying share purchases to long-term procurement commitments, Google gains supply certainty while Marvell secures both revenue visibility and a powerful strategic investor. The structure also highlights the capital intensity required to build competitive AI systems—even tech giants are using creative financing to lock in critical components through the end of the decade. For Marvell, the partnership validates its position in AI networking and inference acceleration, markets where it has invested heavily to compete beyond traditional data center connectivity.
The details were first reported by CNBC based on regulatory filings.
This is an original analysis by the Omega editorial team. Source reporting: AI Watch.
Want systems like this working for your business?
Book a Call

