Google and Marvell Deepen AI Chip Alliance With $12.2 Billion Stock Option
The race to control the hardware powering artificial intelligence is entering a new phase—and Google has just made a potentially significant move.
Marvell Technology has granted Alphabet’s Google a warrant to purchase up to approximately $12.18 billion worth of Marvell shares as part of an expanded agreement to develop custom artificial-intelligence chips. The arrangement gives Google the right to acquire as many as 58.97 million Marvell shares at $206.58 per share, rather than representing an immediate $12.2 billion investment.
The agreement ties Google’s potential equity position to the commercial success of the companies’ custom-chip program. Reuters reported that the warrant could ultimately make Google one of Marvell’s largest shareholders if the applicable conditions are met through fiscal 2033.
This Is More Than a Stock Deal
At the heart of the agreement is an increasingly important battleground in the AI industry: custom silicon.
Instead of relying entirely on off-the-shelf graphics processors, major technology companies are increasingly designing specialized chips tailored to their own AI workloads.
Google has been pursuing this strategy for years through its Tensor Processing Units (TPUs), specialized processors designed to accelerate machine-learning workloads.
Under the new arrangement, Marvell is expected to develop a range of custom technologies for Google’s AI infrastructure, including AI inference accelerators, storage controllers, networking and memory-interface controllers, and near-memory computing technologies.
That matters because AI inference—the process of actually running trained AI models to generate responses—is becoming an enormous computing workload as services such as generative AI scale globally.
Why Google Wants More Custom Chips
Google has been building its own AI-chip ecosystem as it tries to control more of the technology stack behind its AI services and Google Cloud infrastructure.
The company has invested heavily in its TPU family while continuing to use GPUs from Nvidia and other specialized hardware.
Google’s strategy is increasingly about having multiple sources of advanced computing capacity, rather than depending on a single chip supplier.
Earlier this year, Reuters reported that Google was already in discussions with Marvell over additional chips designed to run AI models more efficiently. Those discussions reportedly included a memory-processing chip designed to work alongside Google’s TPUs and another TPU-oriented processor.
The latest agreement turns those earlier discussions into something much more consequential.
The Broadcom Question
The deal is also significant because Broadcom has been a major partner in Google’s custom AI-chip strategy.
In April, Broadcom signed a long-term agreement with Google to develop and supply future generations of custom AI chips and other components for Google’s next-generation AI racks through 2031.
That means Google’s expanded relationship with Marvell is not simply about adding another supplier—it potentially signals a broader effort to diversify the companies building the specialized silicon behind Google’s AI infrastructure.
Markets reacted immediately.
Marvell shares jumped more than 10% in premarket trading after the announcement, while Broadcom shares moved lower as investors assessed what the expanded Marvell relationship could mean for its position in Google’s custom-chip ecosystem.
The $12.2 Billion Figure Needs Context
The headline number is eye-catching, but it needs to be understood correctly.
Google is not handing Marvell $12.2 billion today.
Instead, Marvell has issued Google a warrant giving it the right to purchase the shares at a predetermined exercise price. The potential value of the shares is approximately $12.18 billion based on the stated $206.58 exercise price.
The warrant is structured around the future commercial relationship between the two companies. Reporting indicates that portions of the warrant vest over time and that additional portions are connected to revenue generated from jointly developed products.
In other words, the arrangement effectively links Google’s potential ownership stake to the growth of the chip business.
Why Marvell Is Becoming a Bigger AI Infrastructure Player
Marvell has increasingly positioned itself as a major supplier of custom silicon and data-center infrastructure as hyperscalers build their own AI accelerators.
The company’s custom-chip business has become an important growth area as cloud providers seek processors tailored to their specific workloads.
Reuters reported earlier this year that Marvell’s custom-chip business was expected to surpass $10 billion in revenue in fiscal 2029, underscoring the scale of the opportunity investors see in specialized AI silicon.
The company is also involved in other major hyperscaler programs, making the Google agreement part of a much larger shift toward customized AI infrastructure.
The Bigger Battle: Nvidia vs. Custom Silicon
The strategic backdrop is even bigger than Google and Marvell.
Nvidia remains the dominant force in AI accelerators, but hyperscalers increasingly want alternatives that can be optimized for their own architectures, workloads and economics.
Google’s TPUs are one of the most prominent examples of that strategy.
Custom chips can potentially give large cloud companies greater control over performance, power consumption, supply and cost. But designing and manufacturing advanced AI silicon is extremely complex, and companies still depend on a broad ecosystem of semiconductor designers, manufacturers, networking companies and memory suppliers.
That is why partnerships such as Google’s agreement with Marvell are becoming increasingly important.
What This Could Mean for the AI Chip Industry
The deal could have implications far beyond Marvell’s stock price.
For Google, it potentially provides another important partner for expanding its custom AI silicon capacity.
For Marvell, it could secure a deeper relationship with one of the world’s largest AI infrastructure buyers while creating a pathway for potentially significant future revenue.
For Broadcom, however, the development raises questions about how Google’s custom-chip supplier network could evolve.
And for Nvidia, the growing push toward proprietary AI accelerators represents another sign that the industry’s biggest technology companies are looking for ways to reduce their dependence on general-purpose GPUs.
That does not mean Nvidia is being pushed aside. Rather, the AI infrastructure market is becoming increasingly fragmented, with GPUs, TPUs and custom accelerators competing—and often working alongside one another.
The Real Story Is What Happens Next
The most important part of Google’s Marvell deal may not be the $12.2 billion headline figure.
It is the structure behind it.
By tying a potential multibillion-dollar equity position to future chip development and commercial activity, the agreement creates a financial relationship that could deepen as Google’s AI infrastructure expands.
And that raises the bigger question:
Is Google simply adding another supplier—or is it quietly building a much broader custom-chip ecosystem designed to challenge the industry’s dependence on Nvidia and strengthen its negotiating power with every major semiconductor partner?
For now, the answer is not fully clear.
But one thing is becoming increasingly difficult to ignore: the next phase of the AI race may be decided not only by who builds the smartest models, but by who controls the chips underneath them.
Leave a Reply