Google is significantly expanding its partnership with semiconductor company Marvell Technology as the tech giant moves to strengthen its custom artificial intelligence chip infrastructure and diversify its AI hardware supply chain.
The agreement gives Google the option to purchase up to 58.97 million Marvell shares at $206.58 per share, representing a potential stake worth approximately $12.2 billion if fully exercised. The arrangement is tied to Google’s purchases of Marvell’s custom semiconductor products.
Deal Could Generate Up to $120 Billion for Marvell
The partnership could become significantly larger over the coming years.
Under the agreement, Marvell could generate as much as $120 billion in revenue from custom products through fiscal 2033, provided Google reaches the purchasing and performance milestones attached to the deal.
The warrant is structured around Google’s purchases. Shares progressively vest as Google reaches specified revenue thresholds, meaning the potential $12.2 billion equity position is not an immediate cash investment by Google.
The arrangement therefore creates a close financial link between Google’s demand for custom AI silicon and Marvell’s potential long-term growth.
Marvell to Support Google’s TPU Ecosystem
The agreement extends beyond the development of a single AI processor.
Marvell will work on custom semiconductor products designed to support Google’s Tensor Processing Unit (TPU) ecosystem, Google’s internally developed hardware for training and running artificial intelligence workloads.
The products covered include AI inference accelerators, storage controllers, network interface controllers, memory interface controllers and near-memory computing technology.
This is important because modern AI systems depend on much more than the central processor. Networking, memory and storage technologies all play a critical role in moving huge amounts of data through AI data centres.
Google Looks Beyond Nvidia
The agreement comes as the world’s largest technology companies race to secure enough computing power to support the rapid growth of generative AI.
While Nvidia remains a dominant force in AI accelerators, companies including Google, Amazon and Microsoft have increasingly invested in custom silicon to improve performance, manage costs and gain greater control over their AI infrastructure.
Google has developed its own TPUs for years, but relies on semiconductor partners to help build the wider hardware ecosystem around those processors.
The expanded Marvell relationship gives Google another important supplier as it continues scaling its AI infrastructure.
Broadcom Comes Under the Spotlight
The announcement has also raised questions about Google’s relationship with Broadcom, which has been a major partner in Google’s custom TPU programme.
Following the announcement, Marvell shares rose sharply, while Broadcom shares fell as investors assessed whether Marvell could capture a larger portion of Google’s custom-chip business. Reuters reported that Marvell’s shares gained nearly 8%, while Broadcom declined by more than 5%.
However, the agreement does not necessarily mean Google is replacing Broadcom.
Analysts and industry reports indicate that Google continues to maintain relationships with multiple chip suppliers, suggesting the company is instead building a broader and more diversified AI semiconductor supply chain.
A Major Signal for the AI Chip Industry
The Google-Marvell agreement highlights how the AI boom is reshaping the semiconductor industry.
The competition is no longer simply about who can manufacture the fastest AI processor. Companies are increasingly competing across the entire AI infrastructure stack — from accelerators and memory to networking, storage and data-centre connectivity.
For Marvell, the partnership could provide a major source of long-term growth. For Google, it represents another step in its strategy of developing and scaling its own AI hardware while maintaining multiple supply-chain relationships.
With the potential for $120 billion in custom-chip purchases through 2033, the agreement could become one of the most significant supplier relationships in Google’s expanding AI infrastructure strategy.
The bigger story: As AI models become more powerful and computationally demanding, the race to build the infrastructure behind them is becoming just as important as the race to build the AI models themselves.
