Google in talks with Marvell to develop custom AI chips
Marvell Technology's stock jumped nearly 5% Monday after reports that Google is negotiating with the chip designer to build two new processors for running AI models more efficiently.
The potential deal would include a memory processing unit to work alongside Google's tensor processing unit and a new TPU specifically designed for AI inference, according to The Information.
Google and Marvell did not respond to requests for comment.
Why this matters for tech infrastructure
The move reflects a broader industry shift. Google, Meta, and other large technology companies are spending heavily to reduce reliance on external chip suppliers, particularly Nvidia, as AI demand drives up costs.
Tech companies collectively plan to spend at least $630 billion this year on AI infrastructure.
Google currently works with Broadcom to design its chips. A deal with Marvell would signal the company is diversifying its suppliers as demand for specialized processors grows.
Russ Mould, investment director at AJ Bell, said the strategy makes sense: "Rivals will want to grab a piece of the market and customers want to diversify their sources of supply to spread technological and supply chain risk."
The competitive landscape
Meta extended its Broadcom deal last week, paying the supplier $2.3 billion last year for AI chip design and services. Both Marvell and Broadcom help clients design custom chips as data center demand accelerates.
Nvidia invested $2 billion in Marvell last month to make it easier for customers to integrate custom AI chips with Nvidia's networking equipment and processors.
Marvell's stock has gained 64% this year after declining 23% in 2025. The company expects revenue to approach $15 billion in fiscal 2028.
Marvell trades at 33.35 times forward earnings estimates, compared with 27.84 for Broadcom.
For IT and development professionals, understanding custom chip strategies matters. Learn more about Generative AI and LLM infrastructure and AI for IT & Development to see how these hardware decisions affect software deployment and system design.
Your membership also unlocks:
