Nvidia's investment of US$2 billion in Marvell Technology aims to create a strategic alliance that will combine Marvell's custom chip designs with Nvidia's networking and processing ecosystem. Announced on March 31, this partnership is a response to the increasing demand for specialized AI hardware, as major tech companies are projected to invest over US$630 billion in AI infrastructure this year.
Following the news, Marvell's shares surged by nearly 7%, while Nvidia's stock rose by 2.7%. Under this collaboration, Marvell will provide custom "XPUs," which cover various data processing units and networking solutions using Nvidia’s NVLink Fusion platform. In return, Nvidia will supply essential technologies such as Vera CPUs, ConnectX NICs, and Spectrum-X switches.
The partnership seeks to create a heterogeneous AI infrastructure that will allow customers to design specialized processors compatible with Nvidia’s existing hardware. It will also focus on advancements in silicon photonics and optical interconnect solutions, addressing critical bottlenecks in large-scale AI systems. Additionally, the collaboration intends to improve telecommunications infrastructure by utilizing Nvidia’s Aerial AI-RAN technology for transitioning to AI-capable 5G and 6G networks.