A new trend is emerging in the artificial intelligence industry: financial relationships between giants are extending far beyond simple buy-and-sell deals. Google Corporation has received the exclusive right to acquire up to 58.97 million shares of chip manufacturer Marvell Technology. This involves a warrant with a potential value estimated at $12.2 billion, making this deal one of the most significant in the field of specialized AI processors.
An option as a tool for strategic partnership
The structure of the agreement, disclosed in a filing with the SEC, demonstrates an unconventional approach to incentives. Only about 1.36 million shares will be unlocked evenly over the first year. The remaining volume is divided into 240 tranches, each of which becomes available only after Google's purchases generate an additional $500 million in revenue for Marvell. Full exercise of the warrant is possible only upon reaching a cumulative sales volume of $120 billion—a figure that underscores the scale of ambition on both sides.
The exercise price is set at $206.58 per share, and the option expires on August 18, 2033. If the warrant is fully exercised, Google could enter the top five largest shareholders of Marvell. The market reacted positively to the news: Marvell's shares rose, while the stock of key competitor Broadcom lost more than 5% of its value.
Reducing dependence on Nvidia
Behind this agreement lies Google's strategic goal of reducing its dependence on the dominant accelerator supplier, Nvidia. Its own TPUs (Tensor Processing Units) have been the foundation of the corporation's AI infrastructure for many years, but now Marvell is being brought in not only for the production of the chips themselves but also for creating critical components: networking solutions, memory controllers, and interfaces for combining thousands of accelerators into unified computing clusters.
It is important to note that analysts do not view this deal as displacing Broadcom from Google's supply chain. Rather, the hyperscaler is diversifying its supplier base amid explosive growth in demand for computing power. Such schemes, combining purchase guarantees with stock options, are becoming the new norm. A recent example is the agreement between AMD and OpenAI, where the ChatGPT developer also received the right to a stake in the chip manufacturer.
Financial engineering on the edge of risk
The intertwining of interests becomes even more evident in the case of Nvidia and OpenAI. Recently, Nvidia agreed to provide a $105 billion guarantee for OpenAI's data center in Ohio, investing an additional $1.5 billion in SB Energy, the company implementing this project. This structure creates an unusual chain: Nvidia is effectively financing the construction of infrastructure that will then become the largest buyer of its own accelerators.
The combined spending of the largest technology companies on AI this year will exceed $730 billion. Against this backdrop, Nvidia, together with investment giants Apollo, BlackRock, and KKR, is creating platforms to attract more than $500 billion in third-party capital. The chip manufacturer itself is ready to provide guarantees totaling up to $125 billion.
The Bank for International Settlements (BIS) has already warned of macro-financial risks from this kind of "circular financing," pointing to the opacity of terms and the risk of rehypothecation of the same assets. AI has finally transformed from a software industry into a capital-intensive one, where competitive advantage is determined not by model quality but by the ability to cheaply finance infrastructure construction.
My comment: We are witnessing a fundamental shift in the AI economy: companies are no longer just buying chips; they are investing in each other, creating complex cross-obligations. This accelerates infrastructure construction but simultaneously builds in systemic risks—if one participant in this web fails to meet its obligations, a chain reaction could affect the entire industry. Investors should closely monitor the disclosure of terms in such deals.