Google expands AI chip partnership with Marvell, adds option to buy $12.2bn in shares

Google expands AI chip partnership with Marvell, adds option to buy $12.2bn in shares
Google’s Marvell AI move

Google is widening its artificial intelligence chip supply strategy through a deeper tie-up with Marvell as demand for custom semiconductors accelerates across large data centre operators. The agreement includes collaboration on tensor processing unit-related hardware and gives Google the right to buy up to $12.2 billion of Marvell stock.

Highlights

  • Marvell will work with Google to develop custom semiconductor products tied to the TPU ecosystem, including AI accelerators and controllers, under a new partnership.
  • Google received a warrant to buy up to 58.9 million Marvell shares at $206.58 each, with vesting linked to $500 million joint revenue tranches between Q3 2027 and 2033.
  • Marvell shares rose 8% on the announcement, extending a year-long rally, as it positions against Nvidia and expands its AI infrastructure reach through photonics and recent acquisitions.

Custom chip agreement and financing structure

As reported by Financial Times, Marvell said on Wednesday that it will work with Google to develop custom semiconductor products linked to the TPU ecosystem, including AI inference accelerators, storage controllers, network interface controllers, memory interface controllers and near-memory compute.

According to a regulatory filing, Marvell issued a warrant that allows Google to purchase up to 58.9 million shares at $206.58 each. Some 1.3 million shares vest in equal quarterly instalments during the first year after the warrant is executed, while the rest vest through discretionary purchases from Marvell's third quarter of 2027 through the end of 2033, with one tranche tied to every $500 million in revenue from jointly developed products.

The broader partnership adds to Google's existing TPU supply arrangements. In April, Google signed an agreement with Broadcom to develop TPUs and other components for its AI infrastructure through 2031, while Google is also beginning to sell its TPUs, which were previously used largely in its own data centres, to outside customers.

Market implications for the AI semiconductor sector

Marvell, which helps hyperscalers including Amazon design specialised AI accelerator chips, is positioning itself as an alternative to Nvidia's general-purpose graphics processing units. Its shares rose 8% on Wednesday after the announcement, extending gains that have more than tripled the stock's value over the past 12 months, while Broadcom shares fell 5%.

The deal comes as competition intensifies around the hardware underpinning generative AI systems. Earlier this month, Financial Times reported that Google has built a financing operation to supply more than $150 billion of AI chips to Anthropic.

Marvell is also broadening its role in AI infrastructure through photonics. Nvidia said in March that it was investing $2 billion in Marvell and that the two companies would work together on silicon photonics to speed data flows in data centres; earlier this year, Marvell acquired Celestial AI, whose technology helps connect AI chips inside large computing clusters used for systems such as Google's Gemini, Anthropic's Claude Code and OpenAI's ChatGPT.

U.S. regulators’ review of compute derivatives contracts outlined how a formal market for instruments tied to computing power could help companies and investors hedge AI infrastructure costs and supply risks. Our publication noted that the consultation focuses on issues such as market oversight, manipulation concerns, customer protections, and the idea of perpetual compute futures as an early step toward potential rules for this emerging asset class.

This material may contain third-party opinions, none of the data and information on this webpage constitutes investment advice according to our Disclaimer. While we adhere to strict Editorial Integrity, this post may contain references to products from our partners.