Google Bets $12.2 Billion on Marvell as AI Chip Race Intensifies

Key Highlights of the News

• Marvell has given Google a warrant to buy up to 58.97 million Marvell shares worth about $12.2 billion.
• The agreement covers custom chips for Google's AI infrastructure, including processors, networking and storage components.
• Marvell could generate up to $120 billion in cumulative revenue through fiscal 2033 if Google meets the required purchase milestones.
• Marvell shares jumped while Broadcom shares fell, highlighting concerns over Google's growing push to diversify its AI chip supply chain.

The AI boom is creating demand for new processors, data centres and specialised infrastructure. Explore upGrad's Artificial Intelligence courses to understand the technologies driving this shift.

Google Gives Marvell a $12.2 Billion Equity Option

Google has received a warrant from Marvell Technology to purchase up to 58.97 million Marvell shares.

The shares carry an exercise price of $206.58 each, putting the potential value of the warrant at about $12.2 billion. The arrangement forms part of an expanded commercial agreement between the companies.

The agreement focuses on developing custom silicon for Google's growing artificial intelligence infrastructure. Marvell will work with Google on custom products spanning AI processors, networking components and storage-related technologies. The warrant is structured around the commercial success of the partnership rather than being an upfront investment.

Google's ability to receive the full number of shares depends on purchase and revenue milestones tied to the custom products developed with Marvell. Reuters reported that the agreement could generate up to $120 billion in revenue for Marvell through fiscal 2033. That figure is not guaranteed revenue. It represents the level of cumulative product purchases needed for the warrant's performance-based shares to fully vest.

The structure effectively links Google's potential ownership stake to how much business it generates for Marvell. If the custom-chip relationship scales substantially, Google could become one of Marvell's largest shareholders. Reuters said full exercise could make Google Marvell's fifth-largest investor.

Why Google Is Expanding Its Custom Chip Supply Chain

The agreement comes as Google increases its reliance on internally designed chips for AI workloads.

Google develops Tensor Processing Units, or TPUs, for training and inference workloads. The company has historically worked with Broadcom to design those processors. The new Marvell relationship points toward a broader custom-silicon strategy. Earlier reports indicated that Google and Marvell were discussing two AI chips. One was described as a memory-processing unit, while another was designed for AI inference.

Custom chips can help hyperscalers optimise hardware for specific workloads. They can also give companies greater control over performance, power consumption and supply chains. That matters as AI workloads require increasingly large computing clusters.

Google is also trying to make its TPUs easier for developers to use. Its TorchTPU initiative is designed to improve compatibility with PyTorch, a widely used machine-learning framework. That could reduce some of the software barriers around adopting Google's chips.

The broader strategy reflects a major shift in AI infrastructure. Hyperscalers are increasingly designing or commissioning specialised processors instead of relying entirely on general-purpose GPUs.

Broadcom Takes the Market's Immediate Hit

The announcement immediately affected investor sentiment across the custom-AI-chip market. Marvell shares rose sharply after the agreement became public. Reuters reported an increase of nearly 8%, while other market reports put the gain closer to 10%.

Broadcom moved in the opposite direction. Its shares fell more than 5% as investors considered whether Google's expanded relationship with Marvell could reduce Broadcom's future share of Google's custom-chip business.

However, the market reaction does not mean Google is abandoning Broadcom. Reuters reported that analysts viewed the arrangement primarily as Google's attempt to diversify its AI supply chain. The agreement could therefore create another supplier rather than replace Broadcom entirely.

That distinction is important because Google's AI infrastructure requirements are expanding rapidly. The company needs processors, networking, memory and connectivity technologies across increasingly large data-centre deployments.

Marvell is positioning itself across several of those components. The company has also been expanding its custom-silicon capabilities through acquisitions and partnerships. In December, Marvell announced its $3.25 billion acquisition of Celestial AI to strengthen its photonics technology. Marvell also received a $2 billion investment from Nvidia earlier this year.

That investment was aimed at making it easier for customers to combine Marvell's custom AI chips with Nvidia's networking and central-processing technologies. The Google agreement therefore arrives as Marvell attempts to establish itself as a major supplier in the custom AI silicon market.

For Google, meanwhile, the deal provides another route to scale its AI hardware ecosystem. The bigger significance is the changing structure of AI computing. Instead of one dominant processor architecture serving every workload, hyperscalers are increasingly building specialised systems around their own AI requirements.

Conclusion

Google's agreement with Marvell is more than a large chip-development contract. The $12.2 billion warrant ties Google's potential ownership in Marvell to the success of their custom-chip business. The structure also gives Marvell a significant incentive to scale the partnership.

For Google, the deal adds another important supplier to its AI hardware strategy. The move also shows how aggressively hyperscalers are building custom silicon as AI computing demand continues to grow.

The immediate market reaction suggests investors see Marvell as a stronger competitor in this market. It also shows why Broadcom's position as a major custom-AI-chip supplier is receiving closer scrutiny.

As AI infrastructure expands, organisations also need stronger expertise around AI security and technology risks. The IIT Delhi Cyber Security and AI Certificate Programme can help professionals build knowledge across these closely connected areas.