Google and Marvell have reached a chip customization cooperation agreement, which not only covers a broad range of AI chip customization business but also achieves an unprecedented deep commercial interest alignment between the two parties through a warrant valued at up to $12.2 billion.
The intensifying competition in AI infrastructure is driving tech giants and chip manufacturers to explore closer cooperation models. Recently, Google and Marvell reached a chip customization cooperation agreement, which not only covers a broad range of AI chip customization business but also achieves an unprecedented deep commercial interest alignment between the two parties through a warrant valued at up to $12.2 billion.
The "Double Lock" of Capital and Business
According to disclosures, Marvell issued Google warrants allowing it to purchase up to approximately 58.97 million common shares, with an exercise price set at $206.58 per share (approximately RMB 1390.3), representing a potential total value of about $12.2 billion. If fully exercised, Google could become Marvell's fifth-largest shareholder.
However, this is not a simple financial investment or cash order. The vesting of these warrants is designed as a highly customized "performance-based bet" mechanism. Apart from a small number of shares vesting quarterly at the beginning of the agreement, the vesting of the vast majority of shares is strictly tied to future actual purchase volumes: starting from Marvell's fiscal year 2027 Q3 through fiscal year 2033, each time Marvell recognizes $500 million in revenue from Google's custom chip business, Google can unlock a batch of shares (240 batches in total).
In simple terms, the larger Google's procurement scale, the higher its potential equity incentive.
This means that only when Marvell's cumulative recognized revenue from Google's custom chip business potentially reaches $120 billion will this $10-billion-plus equity promise be fully realized. This mechanism cleverly ties Google's purchasing intent to Marvell's R&D and delivery capabilities, ensuring long-term strategic synergy between the two parties over the next seven years.
The "Flanking Maneuver" Focusing on the TPU Ecosystem
On the technical cooperation level, Marvell is targeting the core peripheral aspects of Google's TPU (Tensor Processing Unit) ecosystem. The product lines covered by the agreement include AI inference accelerators, storage controllers, memory interface controllers, network interface controllers, and near-memory computing chips.
This layout is highly strategic. In the operation of AI large models, in addition to the core computing chips, data storage, movement, and network interconnection are also bottlenecks determining the overall efficiency of data centers. Marvell's deep expertise in these "supporting chip" areas can effectively help Google improve the data throughput rate and reduce the energy consumption of TPU clusters.
It is worth noting that Marvell is not directly involved in designing the core computing chips of TPUs but has chosen to complement the ecosystem around the core computing power. This "flanking maneuver" showcases both Marvell's technical prowess and its flexible approach in supply chain cooperation.
Supply Chain Diversification and Industry Landscape Reshaping
After the news was announced, Marvell's stock price surged significantly, rising up to 13.54%. The closing price was $237.27, up 9.85%, giving the company a total market capitalization of $207.8 billion. Meanwhile, the stock price of its long-time competitor, Broadcom, showed a noticeable decline. Currently, the company's stock closed at $362.48, down 4.61%, with a total market capitalization of approximately $1.72 trillion.
For a long time, Broadcom has been a core partner for Google in the custom chip field. Google's high-profile introduction of Marvell and the grant of equity incentives this time is widely interpreted by the market as a key move by major cloud vendors to proactively promote supply chain diversification amid surging AI computing power demand.
Analysts point out that Broadcom previously held the majority share of Google's TPU orders. However, with Marvell's entry, this landscape is likely to evolve into a more balanced situation of "two giants standing side by side." Broadcom's market share will shift from "dominating the field" to "a relative lead."
Despite the risk of share dilution, this does not mean Broadcom's business with Google will face a cliff-like decline. In the core custom AI chip design field, Broadcom is still regarded by the industry as "best-of-breed."
The current macro background is that global AI computing power demand is experiencing explosive growth, and the overall market is in a state of "supply shortage." Google's introduction of Marvell is more about meeting its own increasingly vast computing power demand rather than simply to "sideline" Broadcom. As industry experts stated, this reflects the overall expansion of the TPU sales pie. Broadcom's fundamentals remain extremely strong, with its AI semiconductor revenue maintaining rapid growth, and management has reiterated its ambitious goal of AI revenue exceeding $100 billion in fiscal year 2027. Therefore, even if the share ratio decreases, the absolute revenue scale of Broadcom's business with Google is still expected to grow.
For Google, distributing the supply rights of core chips to multiple suppliers can not only effectively reduce reliance risks on a single vendor but also gain stronger bargaining power in the fierce scramble for production capacity. For Marvell, securing the capital backing and long-term order expectations of a tech giant is undoubtedly a significant milestone in establishing its leading position in the AI custom chip race.
Overall, this "union" between Google and Marvell transcends the traditional buyer-seller relationship, marking a new phase where upstream and downstream partners in the AI industry chain are moving from simple supply-demand cooperation towards deep integration of capital and business. In an era where computing power equals power, whoever can build a more elastic and efficient custom chip supply chain will grasp the true initiative in the next wave of the artificial intelligence tide.
