Shares of Marvell Technology (NASDAQ:MRVL) jumped 11% on Tuesday following a joint announcement with NVIDIA that combines a $2 billion investment from NVIDIA with a strategic technology partnership.
The multiyear collaboration ties Marvell into NVIDIA’s AI factory and AI-RAN ecosystem by way of NVIDIA NVLink Fusion, a rack-scale platform designed to let customers develop semi-custom AI infrastructure while remaining compatible with NVIDIA’s NVLink ecosystem.
Under the terms outlined by the companies, Marvell will contribute custom XPUs together with scale-up networking that is compatible with NVLink Fusion. NVIDIA will supply complementary technologies including the Vera CPU, ConnectX network interface cards, BlueField data processing units, NVLink interconnects and Spectrum-X switches, alongside rack-scale AI compute resources.
The agreement also calls for joint work on silicon photonics technology and sets an explicit goal to repurpose telecommunications networks as AI infrastructure through NVIDIA Aerial AI-RAN aimed at 5G and 6G deployments.
Jensen Huang, founder and CEO of NVIDIA, framed the arrangement around rising demand for inference and token generation, saying: "The inference inflection has arrived. Token generation demand is surging, and the world is racing to build AI factories. Together with Marvell, we are enabling customers to leverage NVIDIA’s AI infrastructure ecosystem and scale to build specialized AI compute."
Marvell’s chairman and CEO, Matt Murphy, emphasized that the expanded collaboration signals the increasing role of high-speed connectivity, optical interconnects and accelerated infrastructure in scaling AI deployments.
The NVLink Fusion platform is described as enabling customers that develop custom XPUs to create heterogeneous AI infrastructure that remains fully compatible with NVIDIA systems. That compatibility is intended to facilitate integration across NVIDIA GPU, LPU, networking and storage platforms.
What this means
- Marvell becomes a supplier of custom processing units and NVLink Fusion-compatible networking elements within NVIDIA’s rack-scale approach to AI infrastructure.
- NVIDIA provides a suite of enabling technologies, from CPUs and NICs to DPUs and switches, to support rack-scale AI compute compatible with NVLink Fusion.
- Both companies will collaborate on silicon photonics and on efforts to convert telecom networks into AI-ready infrastructure using NVIDIA Aerial AI-RAN for future 5G and 6G networks.