Google has struck one of the largest chip partnerships of the AI boom, signing Marvell Technology to help develop its in-demand custom silicon and granting the search giant the right to buy a stake worth up to $12.2 billion, Reuters reported on Wednesday.
The deal is the latest signal that Big Tech is willing to take direct financial positions in the suppliers powering its AI build-out, reshaping the semiconductor supply chain around a handful of hyperscale customers. For readers tracking these shifting alliances, our breaking AI news coverage follows every major chip and model announcement as it lands.
The announcement lands amid a broader rally in AI infrastructure stocks, with investors rewarding any company that secures a direct role in hyperscaler compute plans. Under the agreement, Marvell has issued Google a warrant to purchase up to 58.97 million shares at $206.58 apiece — a package worth $12.18 billion if fully exercised, according to Reuters. If Google exercises the warrant in full, it would become the fifth-largest investor in Marvell.
What Marvell Will Build for Google
The partnership covers a broad range of technologies used alongside Google's tensor processing units (TPUs), including processors that run AI models, manage data storage, and move information across networks. Marvell has long specialized in this kind of infrastructure silicon, and the Google deal dramatically expands its custom-chip business.
Analysts say the deal could bring Marvell roughly $120 billion in revenue through fiscal 2033, provided Google hits the volume targets that the stake option depends on.
Investors reacted swiftly. Marvell shares jumped nearly 8% on the news, at one point popping as much as 10% in intraday trading, according to CNBC. Broadcom, which had been Google's main custom chip partner, fell more than 5%, while shares of Google parent Alphabet were little changed.
Why Custom Chips Are Winning
Demand for in-house chips such as Google's TPUs has surged as cloud providers hunt for cheaper alternatives to Nvidia's expensive graphics processors and for hardware better suited to inference — the process of running trained AI models, which now accounts for the bulk of AI computing costs.
A recent overhaul of Google's AI division also shifted power toward executives with closer ties to Google Cloud, putting a brighter spotlight on the custom chips and AI infrastructure that analysts say are increasingly central to that business.
Morningstar analyst William Kerwin called the agreement "a big win for Marvell," but cautioned against reading it as a blow to Broadcom. "I saw this news as a growing pie at Google for new sources, rather than a competitive displacement of Broadcom," he said in a note reported by Reuters. Morningstar went as far as calling Google the "white whale" that Marvell had long pursued.
Broadcom's decline was not only about the Google news. Yahoo Finance reported that the stock's 5% drop also reflected lingering concerns around its VMware software business and financing questions — though losing an exclusive grip on Google's custom-chip program was clearly the trigger for the day's rotation.
A Pattern of Big Tech Buying Into Its Suppliers
The Marvell-Google deal extends a fast-growing pattern in which AI's biggest spenders are taking equity stakes in the companies that supply them.
Earlier this month, Nvidia agreed to provide a backstop of up to $105 billion for a data-center project that OpenAI is leasing in Ohio. In October, AMD struck a similar arrangement, agreeing to supply OpenAI with AI chips worth tens of billions of dollars in annual revenue while giving the ChatGPT maker an option to buy a stake of up to roughly 10% in the chipmaker.
These intertwined relationships have raised concerns among regulators and analysts about concentration in the AI supply chain, where a few large customers increasingly underwrite the capital plans of their own vendors. The Reuters report noted that Wednesday's deal could add to those concerns.
What It Means for the AI Chip Market
For Marvell, the Google warrant represents a dramatic vote of confidence from one of the world's largest cloud-computing providers, coming after a period in which the company's custom-silicon ambitions faced skepticism following the loss of key customer programs.
For Google, the deal secures a second major custom-chip partner as TPU demand accelerates, reducing reliance on any single supplier at a moment when AI inference workloads are straining global compute capacity.
And for the broader market, the message is that custom silicon — not just Nvidia GPUs — will carry a growing share of the world's AI computation. Broadcom's share price reaction illustrates the stakes: even a company that retains the majority of Google's custom-chip business can lose billions in market value when a rival wins a seat at the table.
Stay Ahead in AI
Chip deals like this one are redrawing the map of the AI economy in real time. Follow the latest AI developments and keep up with every major announcement at AI Buzz Wire.
Read more AI news →
