Particle.news
Download on the App Store
3 ARTICLES
·
last mo.

Nvidia Invests $2 Billion in Marvell to Link Custom AI Chips to Its Platform

The pact signals Nvidia’s move to keep its ecosystem central as customers pursue custom processors by pairing it with faster, lower‑power data links built with silicon photonics.

Image

Overview

  • Nvidia put $2 billion into Marvell and agreed to integrate Marvell’s custom AI chips with Nvidia’s networking gear and CPUs, including support for NVLink Fusion, Nvidia’s high‑speed chip‑to‑chip link.
  • The companies will co‑develop optical interconnects and silicon photonics that use light instead of copper to move data faster while cutting power use in large AI systems.
  • Marvell will deliver custom processors and network solutions compatible with NVLink Fusion, and Nvidia will provide CPUs, network interface cards and interconnect technology.
  • Industry estimates point to at least $630 billion in AI infrastructure spending this year by companies such as Alphabet and Meta, which could fuel demand for servers and networking parts Marvell supplies.
  • Marvell projected revenue nearing $15 billion in fiscal 2028, about 40% higher than today, citing expected gains from tighter alignment with Nvidia and broader AI data center buildouts.