The largest players in the technology sector continue to reshape the rules of the game. This time, it is about a deal that goes far beyond classic partnerships: Google has received an option to purchase up to 58.97 million shares of chipmaker Marvell Technology. The potential deal is valued at $12.2 billion, which automatically makes this move one of the most significant in the field of specialized AI processors.
Deal Structure: From Guarantees to Direct Financial Incentives
The terms of the agreement, disclosed in an 8-K report filed with the SEC, demonstrate an unconventional approach. Only about 1.36 million shares will be unlocked in equal portions over the first year. The bulk of the warrant—240 tranches—is tied to actual purchases: each tranche becomes available after Google's orders generate $500 million in revenue for Marvell. To fully exercise the option, sales volume must reach $120 billion—a figure that speaks for itself.
The exercise price is set at $206.58 per share, and the warrant will remain valid until August 2033. If Google fully exercises the option, it will enter the top five largest shareholders of Marvell. The market has already reacted: Marvell's shares have risen, while competitor Broadcom's stock has fallen by more than 5%. However, as analysts, including William Kerwin of Morningstar, rightly note, this is not a displacement of Broadcom but rather a diversification of suppliers amid Google's explosive growth in AI computing needs.
Reducing Dependence on Nvidia and the New Market Logic
Google has been developing its own TPUs for years, and Marvell is becoming a key partner not only in creating accelerators but also in developing networking solutions, memory controllers, and interfaces for linking chips into giant computing clusters. This move is part of a strategy to lower the cost of AI computing and reduce dependence on Nvidia's dominance.
However, the most interesting aspect is the financial engineering. The Google-Marvell deal is not an isolated case. In October, AMD entered into a similar agreement with OpenAI, combining purchases with a stock option. Nvidia went even further, providing $105 billion in guarantees for OpenAI's data center in Ohio and investing $1.5 billion in SB Energy. Such schemes blur the lines between supplier, buyer, investor, and lender, creating complex chains of interdependence.
Scale of Capital and Risks to the System
Total AI spending this year will exceed $730 billion, and to finance such infrastructure, Nvidia, together with Apollo, BlackRock, and Goldman Sachs, is creating platforms to attract more than $500 billion in third-party capital. Private funds are already viewing data centers as a separate asset class. But there is a downside to this coin: the Bank for International Settlements warns of macro-financial risks, including opaque agreements and the risk of rehypothecation of assets.
IMD Professor Amit Joshi astutely notes that competitive advantage in AI now depends not on model quality but on who can finance infrastructure more cheaply. However, in my view, it is precisely deals like Google-Marvell that create a dangerous precedent: companies become hostages to each other, and in the event of a liquidity crisis, the entire structure could collapse like a house of cards. This is not just an evolution of the market—it is a fundamental shift that will require a new level of regulatory attention.