Google has secured an option to acquire a stake in Marvell Technology, valued at up to $12.2 billion, enhancing its AI processor development.
Google has secured the right to acquire nearly 59 million shares of Marvell Technology, a deal valued at up to $12.2 billion, as part of a broader partnership aimed at enhancing AI processor development. This information was revealed in a recent filing with the U.S. Securities and Exchange Commission (SEC).
The arrangement highlights the evolving financial dynamics in the AI sector, where chipmakers and tech giants are increasingly forming complex partnerships that go beyond simple transactions. These collaborations often include stock options, guarantees, and joint financial initiatives to bolster computing infrastructure.
Details of the Agreement
This partnership will focus on multiple aspects of Google’s Tensor Processing Unit (TPU) infrastructure, which serves as specialized AI accelerators. Included in the deal are processors for data storage, networking solutions, and memory interfaces.
Of the total shares, approximately 1.36 million will be available in equal portions during the first year following the agreement's signing. The remaining shares are contingent on Google’s revenue generation for Marvell, with each of the 240 tranches becoming available after Google achieves an additional $500 million in revenue, culminating in a total sales target of $120 billion to fully unlock the option.
Google has the right to purchase these shares at $206.58 each until August 18, 2033. If fully exercised, this stake could position Google as Marvell's fifth-largest shareholder, as noted by industry sources.
Following the announcement, Marvell's stock price saw an uptick, while shares of competitor Broadcom dropped over 5%.
“This is a significant win for Marvell,” commented Morningstar analyst William Kerwin, although he emphasized that this does not imply Broadcom's exclusion from Google's supply chain.
Strategic Moves Against Nvidia
The demand for specialized processors is surging as companies aim to decrease reliance on Nvidia’s accelerators while managing AI computing costs. Google has been working on its own TPUs designed for training AI models, with Marvell set to play a crucial role in both creating these accelerators and integrating various chips into systems.
This partnership arrives amidst a restructuring of Google’s AI division, reflecting a shift towards increased influence from Google Cloud leaders. The agreement not only positions Marvell as a significant customer but also ties its financial success to Google's spending, fostering a vested interest in Marvell's growth.
Similar financial arrangements are becoming standard in the AI industry. For instance, AMD recently formed a deal with OpenAI that combined substantial processor purchases with a stake acquisition opportunity.
AI Infrastructure Financing Trends
The intricate relationships among AI participants are exemplified by Nvidia’s recent agreement with OpenAI, which includes a $105 billion guarantee for a new data center project in Ohio, alongside a direct investment of $1.5 billion in the project.
This scenario creates a unique predicament for Nvidia, as it finances a facility that will heavily rely on its own accelerators, though CEO Jensen Huang refuted claims that this constituted circular financing.
As spending on AI infrastructure soars, exceeding $730 billion this year, major firms like Nvidia are collaborating with investment giants to attract substantial external capital for computing infrastructure development.
These trends indicate a shift in how AI is perceived, moving from software-centric models to capital-intensive operations reliant on data centers and chip manufacturing.
Risks and Future Outlook
While these arrangements facilitate rapid infrastructure growth, they also raise concerns about interdependence among companies within the AI market. The Bank for International Settlements has flagged potential macro-financial risks linked to the AI boom, emphasizing the dangers of opaque financial agreements that could lead to circular financing.
In light of these developments, industry experts, including Liao Heng of Huawei, have warned that the physical scalability of large AI chips is nearing its limits, underscoring the need for innovative solutions in the rapidly evolving tech landscape.