Nvidia will invest US$2 billion in Silicon Valley chip manufacturer Marvell to strengthen network technology and connect its AI chips to larger data centers. Two semiconductor companies said on Tuesday that they will cooperate to develop silicon photonics technology and use optical technology to upgrade data center network systems and increase data transmission rates.
The deal between Nvidia and Marvell will also help the technology giant integrate self-developed customized AI chips into Nvidia’s data center systems.
Marvell designs dedicated AI acceleration chips for American technology giants such as Amazon, which can replace Nvidia's general-purpose graphics processor (GPU).
The two companies said the collaboration will enable seamless integration of such customized AI chips with Nvidia GPUs, networks and storage systems.
NVIDIA is leveraging its leading position in the field of AI processors to expand into a comprehensive platform for AI data centers, consolidating its position in the core infrastructure where technology giants are spending billions of dollars competing for AI dominance.
Earlier this month, Nvidia launched its first dedicated chip designed specifically for AI inference (the process in which an AI model responds to user queries and generates output such as text, code, etc.). This is the first time that its core AI chip products have expanded beyond GPUs.
On Tuesday, Marvell's stock price surged 12.8% in New York; the cumulative increase in the past 12 months exceeded 50%, with a market value of US$86.6 billion.
Matt Murphy, CEO of Maxwell, said that cooperation will help customers improve scalability and efficiency: "Our deepening cooperation with NVIDIA reflects the increasing importance of high-speed connections, optical interconnects and acceleration infrastructure in the large-scale development of AI."
Both Nvidia and Marvell have recently consolidated their positions in the AI data center market through mergers and acquisitions.
Last month, Marvel completed its $3.3 billion acquisition of Celestial AI. The photonic technology developed by the latter can connect hundreds of thousands of AI acceleration chips.
Large-scale clustering of AI chips has become the key to building and deploying cutting-edge AI systems (such as Google Gemini, Anthropic Claude Code, and OpenAI ChatGPT).
Since acquiring network equipment maker Mellanox for $6.9 billion in 2019, Nvidia has gradually expanded its business beyond GPUs.
In December last year, Nvidia reached an approximately US$20 billion deal with AI chip start-up Groq to absorb its core talents and license its technology, laying the foundation for the launch of new inference chips in March this year.