Marvell Technology granted Google an option to acquire a $12.2 billion equity stake as part of an expanded custom AI silicon partnership.
Marvell Technology shares jumped more than 12% in premarket trading on Wednesday after the chipmaker announced an expanded partnership with Google to develop custom semiconductors for artificial intelligence applications. As part of the agreement, Google received a warrant that could allow it to acquire up to $12.2 billion worth of Marvell stock.
Under the terms outlined in a regulatory filing, Marvell issued Google a warrant on August 18 to purchase up to 58.97 million shares at an exercise price of $206.58 per share. If Google exercises the entire warrant, the stake would be valued at approximately $12.18 billion according to Reuters calculations. Based on LSEG data, such a position would elevate Google to Marvell’s fifth-largest shareholder. However, the majority of the warrant is contingent upon future procurement. Approximately 1.36 million shares will vest through equal quarterly installments during the first year of the agreement. The remaining shares will vest through fiscal 2033, tied to discretionary purchases by Google and its affiliates. Specifically, one tranche will vest for every $500 million in revenue generated from Marvell’s custom products under the partnership.
The collaboration, formally expanded on July 29, covers a broad spectrum of custom semiconductor programs integrated into Google’s Tensor Processing Unit (TPU) ecosystem. The initiative will encompass AI inference accelerators, storage controllers, network interface controllers, memory interface controllers, and near-memory computing technology. “The expanded partnership spans a comprehensive range of custom silicon programs that attach to the TPU ecosystem, including AI inference accelerators, storage controllers, network interface controllers, memory interface controllers, and near-memory compute,” Marvell stated in its filing. These components are designed to support Google’s rapidly scaling AI infrastructure, with a particular emphasis on inference systems that run already-trained AI models.
The deal arrives as major technology firms deploy unprecedented capital into AI infrastructure, increasingly seeking alternatives to costly general-purpose processors. While Google has long relied on its proprietary TPUs for AI workloads, the broader industry is accelerating development of application-specific silicon. Inference has emerged as a critical growth area as demand for deployed AI services expands. Custom silicon enables companies to optimize hardware for specific workloads, manage costs, and reduce dependence on widely used competitors like Nvidia’s AI accelerators.
The expanded Marvell-Google alliance intensifies competition within the custom AI chip sector, directly challenging Broadcom, which has served as Google’s primary supplier and development partner for custom TPU chips. Following the announcement, Broadcom shares dipped more than 2% in premarket trading, while Marvell gained more than 11%. Broadcom maintains a long-term agreement with Google to develop and supply future generations of custom AI chips and related components for next-generation AI racks through 2031.
The shifting dynamics reflect broader industry trends, with major technology companies projected to spend more than $700 billion on AI infrastructure this year, a sharp increase from the roughly $400 billion invested last year. Google’s shares remained largely unchanged ahead of the market open, underscoring investor focus on the evolving landscape of specialized AI hardware and Google’s expanding procurement requirements.