Particle.news
Download on the App Store

Nvidia Invests $2 Billion in Marvell to Expand Its AI Data‑Center Stack

The tie-up points to Nvidia’s push to control more of the AI factory from chips to networking.

Overview

  • - Nvidia’s partnership with Marvell, disclosed in coverage Wednesday, includes a $2 billion stake and will add Marvell’s custom accelerator chips and scale-up networking into Nvidia’s NVLink Fusion rack‑scale system.
  • - NVLink Fusion links many processors inside a rack so they act like one large computer, and Marvell’s role will supply the specialized silicon and high-speed connections that make that possible.
  • - Nvidia remains financially strong after reporting record fiscal 2026 revenue of $215.9 billion, and CEO Jensen Huang framed the move as timed to an “inference” surge as more AI models run in production.
  • - AWS is reaccelerating with 24% year-over-year revenue growth to $35.6 billion and a large cloud backlog, and reports say Amazon plans roughly $200 billion in 2026 capital spending to build AI data centers and custom chips.
  • - AI stocks have pulled back this year, with one AI-focused ETF down about 9%, yet filings show hedge funds adding to Nvidia and Amazon as export controls limit Nvidia’s China sales but long-term infrastructure demand stays intact.