Google has struck a deal that gives it the right to buy up to $12.2 billion worth of shares in Marvell Technology, as the two companies deepen a partnership to build the custom chips powering Google’s AI systems. The arrangement, disclosed in a securities filing, is tied directly to how much AI hardware Google actually buys from Marvell over the next several years — not a straightforward cash investment.
Under the terms, Google receives a warrant to purchase nearly 59 million Marvell shares at $206.58 each. Only a small portion, about 1.4 million shares, vests automatically in the first year. The rest unlocks in blocks, with one tranche becoming available for every $500 million in chips Google buys from Marvell, running through Marvell’s 2033 fiscal year. If Google hits every target, Marvell stands to collect roughly $120 billion in revenue from the relationship, and Google would become the chipmaker’s fifth-largest shareholder.
The deal covers a broad set of hardware that supports Google’s Tensor Processing Units (TPUs) — the custom chips Google has used internally since 2016 to run its AI models — including components that handle storage and move data across networks. Marvell had already been working with Google in a smaller capacity; this expands that relationship significantly and gives Marvell a bigger foothold alongside Broadcom, which has been Google’s main custom chip partner until now. Marvell’s stock jumped nearly 10% on the news, while Broadcom shares fell more than 5%.
The bigger picture is a familiar one this year: major AI companies are increasingly tying their chip purchases to equity stakes in their suppliers, rather than relying purely on standard purchase orders. A similar arrangement happened in October 2025, when AMD agreed to supply OpenAI with chips worth tens of billions of dollars a year in exchange for giving OpenAI the option to buy roughly a 10% stake in AMD.
For everyday tech buyers, deals like this matter indirectly — cheaper, more efficient custom chips are part of what could eventually make AI features in phones, Chromebooks, and other Google-powered devices faster and less reliant on the cloud. But the immediate impact is financial, and Google’s actual stake depends entirely on how much hardware it ends up buying.