NVIDIA Invests $2B in Marvell — NVLink Fusion Opens Door to Hybrid AI Infrastructure

NVIDIA and Marvell Technology announced a strategic partnership on March 31 that could reshape the AI infrastructure landscape. At its core is NVLink Fusion, a new platform enabling hyperscalers and custom ASIC designers to integrate their own processors with NVIDIA's high-speed NVLink interconnect.

NVIDIA is investing $2 billion in Marvell as part of the deal. Marvell contributes custom XPUs and NVLink Fusion-compatible scale-up networking solutions, while NVIDIA supplies Vera CPU, ConnectX NICs, BlueField DPUs, and Spectrum-X switches.

The NVLink Fusion chiplet delivers up to 1.8 TB/s of bidirectional bandwidth, enabling semi-custom AI infrastructure where heterogeneous silicon can coexist within a standardized rack design based on the OCP MGX architecture.

The partnership also covers silicon photonics and AI-RAN for 5G/6G networks, with the two companies working to transform telecom infrastructure into AI factories.

For CIOs, this means more flexibility in AI infrastructure choices. Organizations no longer need to choose between NVIDIA-only or fully custom ASIC deployments but can combine both in the same rack. This could drive down costs and enable faster adaptation to specific workloads.

Marvell stock surged 11 percent on the news.