Marvell and Google deepen AI chip partnership with expanded custom silicon deal
Marvell Technology and Google have expanded their collaboration on custom semiconductor products tied to artificial intelligence systems. The arrangement gives Google the right to purchase up to roughly $12.2 billion in Marvell shares. It forms part of a broader push by major technology firms to secure diverse sources of specialized chips for AI workloads.
The commercial agreement was signed in late July. It covers development of multiple custom silicon components that attach to Google’s Tensor Processing Unit ecosystem. Marvell disclosed the warrant details in a regulatory filing. The partnership reflects ongoing efforts by hyperscalers to broaden their supplier base for AI infrastructure.

Key Takeaways:
- Google and Marvell have expanded their collaboration on custom semiconductor products for AI systems.
- The deal includes a $12.2 billion share purchase option, potentially making Google a major Marvell shareholder.
- The partnership covers development of AI inference accelerators and other essential components for Google’s TPU ecosystem.
- This move aligns with the trend of tech firms diversifying their AI chip supply chains.
- Marvell’s existing relationships with other cloud providers are strengthened, expanding its reach in AI infrastructure.
Scope of the custom silicon programs
The expanded partnership includes AI inference accelerators. It also covers storage controllers, network interface controllers, memory interface controllers and near-memory compute solutions. These components support the broader systems that run Google’s TPUs.

Google has long relied on partners such as Broadcom for TPU-related work. The new arrangement with Marvell adds another design and supply channel. Marvell already works with other large cloud providers on custom silicon. The Google programs extend that portfolio into additional AI-related functions.
Industry observers note that custom silicon demand continues to grow as companies seek alternatives or complements to general-purpose accelerators. The deal positions Marvell to supply a wider range of chips that operate alongside Google’s in-house processors.
Equity warrant tied to product revenue
In connection with the collaboration, Marvell issued Google a warrant to buy up to 58,970,907 shares of common stock. The exercise price is set at $206.58 per share. Full exercise would represent approximately $12.2 billion and could make Google one of Marvell’s larger shareholders.
A portion of the shares, about 1.36 million, vests in equal quarterly installments over the first year. The remaining shares vest in tranches linked to actual purchases. One tranche becomes available for every $500 million in qualifying custom products revenue recognized by Marvell from Google and its affiliates.
The performance-based vesting runs from Marvell’s third quarter of fiscal 2027 through the end of fiscal 2033. Full vesting of the performance shares would correspond to cumulative qualifying revenue of roughly $120 billion over that period. The warrant remains exercisable until mid-August 2033, subject to the vesting conditions.
This structure links the equity opportunity directly to commercial activity. Google gains potential ownership upside as it places orders. Marvell gains visibility into longer-term demand for the co-developed products.
Diversification trends in the AI chip supply chain
Large technology companies continue to pursue multiple suppliers for specialized silicon. Google’s TPU program has historically centered on a primary partner. Adding Marvell expands options across accelerators, networking, storage and memory interfaces.
Similar arrangements have appeared elsewhere in the sector. Chip designers and hyperscalers have used equity warrants or investment rights alongside supply commitments. These deals aim to align incentives while supporting capacity expansion for AI infrastructure.

Marvell’s existing relationships with other major cloud platforms mean it now participates in custom programs across several leading AI builders. The latest agreement adds Google to that list. Analysts have described the move as expanding the overall opportunity set rather than simply shifting share among existing suppliers.
Demand for AI accelerators and supporting chips remains elevated. Companies are investing heavily in data center capacity. Securing reliable design and manufacturing partners has become a priority for firms building large-scale AI systems.
Fortune Prime Global continues to monitor developments across technology and semiconductor markets as part of its coverage of global financial trends. Readers interested in core market concepts can explore resources such as Forex Trading Basics, which outlines foundational principles relevant to understanding broader economic and sector dynamics.
Market reaction and related sector activity
Marvell shares rose sharply following the disclosure. Reports indicated gains ranging from high single digits to more than 10 percent in trading sessions after the announcement. The reaction reflected investor assessment of the potential revenue scale and the endorsement from a major customer.
Broader semiconductor names showed mixed performance on the same day. Some peer stocks experienced pressure as attention focused on the new partnership. Memory chipmakers and other suppliers remain active in capital return programs and capacity planning amid fluctuating demand signals.
In parallel developments, other technology firms continue to expand financing tools and infrastructure investments. These moves occur against a backdrop of elevated capital spending by cloud providers and ongoing supply chain adjustments.
Historical context for custom silicon partnerships
Custom chip programs have grown in importance as AI workloads increase in scale and specialization. Hyperscalers design or co-design processors optimized for their specific models and data center architectures. Partners handle design services, packaging and coordination with foundries.
Marvell has built a presence in this area through earlier agreements. The Google expansion adds programs focused on inference and supporting infrastructure. Inference workloads, which involve running trained models to generate outputs, represent a significant and growing portion of AI compute demand.

The warrant mechanism provides a transparent link between commercial success and equity participation. It differs from pure supply contracts by giving the customer a direct stake that scales with volume. Similar structures have been used in other recent semiconductor and AI-related agreements.
Longer-term framework through 2033
The revenue measurement period extends into early 2033. This multi-year horizon matches the long development and deployment cycles typical of advanced semiconductor projects. Design, validation, production ramp and system integration often span several years.
Marvell will develop the custom products under the commercial agreement. Google retains ownership of the intellectual property for the chips, consistent with typical arrangements in this segment. Marvell contributes design expertise and takes the designs through manufacturing partners.
The arrangement does not displace existing relationships. It adds capacity and design bandwidth across multiple chip categories that support TPU-based systems. Full realization of the potential revenue depends on Google’s purchasing decisions over the coming years.
Neutral summary of the development
Marvell and Google have formalized an expanded partnership covering custom silicon for AI systems. The deal includes a warrant that could allow Google to acquire a substantial stake in Marvell, with vesting tied to product purchases over an extended period. The collaboration adds another supplier channel for components linked to Google’s TPU ecosystem.
The announcement highlights continuing efforts by major technology companies to diversify sources of specialized chips. It also illustrates the use of equity-linked structures to support long-term commercial relationships in the semiconductor sector. Further details on product development and order volumes will emerge as the programs progress through their planned timelines.
People Also Ask:
What is the significance of Google and Marvell’s partnership?
The partnership allows Google to diversify its AI chip supply chain and enhances its TPU ecosystem with custom components developed by Marvell.
How does the equity warrant work in this deal?
Google has the right to purchase up to $12.2 billion in Marvell shares, with vesting linked to revenue from custom products developed through the partnership.
Why are tech companies diversifying their AI chip suppliers?
Diversification helps technology firms secure reliable sources of specialized silicon, crucial for supporting large-scale AI workloads and infrastructure.
What components are included in the expanded partnership?
The partnership covers AI inference accelerators, storage controllers, network interface controllers, memory interface controllers, and near-memory compute solutions.











