Marvell Shares Jump as Google Eyes Custom AI Chips, Reuters Says

Marvell Technology’s shares rose about 7% in premarket trading after reports that Alphabet’s Google is in talks with the chip designer to develop two custom AI chips, a move that would deepen the push among hyperscalers to control more of their own infrastructure. One proposed chip would be a memory processing unit built to complement Google’s tensor processing unit, while the other would be a new TPU designed to run AI models more efficiently.

The details matter because the AI hardware race is no longer just about buying more Nvidia GPUs. Google and other large cloud operators are trying to design silicon that lowers costs, improves performance and reduces reliance on outside suppliers. The newest chips would be aimed at inference, the stage when an AI model responds to prompts, a part of the market that is becoming central to cloud demand as usage scales beyond training alone.

For Marvell, a deal would be valuable not only because of the immediate revenue potential, but because it would position the company as a strategic design partner inside one of the most important supply chains in AI. Marvell’s fiscal 2026 revenue reached $8.195 billion, up 42% from the prior year, and fourth-quarter revenue came in at $2.219 billion, underscoring how closely its business is already tied to the buildout of AI infrastructure.

Google already works with Broadcom on chip design, and Broadcom disclosed expanded multi-year AI chip and networking relationships with Alphabet in early April. If Google adds Marvell as another partner, it would suggest a broader effort to diversify the custom silicon stack rather than depend on a single outside supplier. That kind of diversification can matter as much as raw speed, especially when chip demand is surging and design cycles increasingly shape who captures the economics of AI.

The report also fits into a wider industry pattern. Big technology companies, including Google and Meta, are moving quickly to reduce dependence on external chipmakers while trying to keep pace with Nvidia’s higher-priced offerings. Marvell has also been signaling broader ambitions across that ecosystem, with investor materials highlighting a partnership with Nvidia around NVLink Fusion. For investors, that combination makes Marvell less of a one-off trade on rumor and more of a barometer for how far custom AI silicon is spreading through the market.