NVIDIA Invests $2 Billion in Marvell to Accelerate Custom AI Infrastructure via NVLink Fusion

A Landmark Strategic Partnership

NVIDIA and Marvell Technology announced a landmark strategic partnership today, anchored by a $2 billion investment from the AI giant. This collaboration centers on the launch of NVIDIA NVLink Fusion, a platform designed to integrate Marvell’s custom silicon into NVIDIA’s dominant AI ecosystem. The deal marks a shift toward heterogeneous computing architectures to meet the soaring global demand for specialized AI factories.

Strengthening the AI Fabric

The $2 billion capital injection gives NVIDIA a strategic stake in Marvell, estimated at roughly 3 to 4 percent of the company’s market capitalization. This move ensures deep technical alignment as the industry moves toward specialized hardware for massive generative AI workloads. The investment was revealed alongside plans to co-develop next-generation infrastructure for hyperscale data centers.

Unleashing NVLink Fusion

Through the NVLink Fusion platform, customers can now build heterogeneous AI infrastructure that combines Marvell’s custom accelerators with NVIDIA’s full hardware stack. This includes support for the Vera CPU, ConnectX networking, and BlueField DPUs, all connected via high-bandwidth proprietary fabric. Marvell will specifically provide custom XPUs and scale-up networking solutions compatible with this ecosystem.

Advancing Silicon Photonics and Telecom

Beyond the data center, the duo will collaborate on silicon photonics to replace traditional copper wiring with light-based data transfers. They also plan to transform telecommunications networks by integrating NVIDIA Aerial AI-RAN technology into Marvell’s 5G and 6G silicon solutions. This initiative aims to turn edge networks into distributed AI infrastructure for global carriers.

Leadership Perspectives

NVIDIA founder Jensen Huang noted that the world is racing to build AI factories and this partnership enables customers to scale specialized compute. Marvell CEO Matt Murphy added that the alliance reflects the growing importance of high-speed connectivity and optical DSPs in scaling global AI. Both leaders emphasized that the collaboration provides greater choice for customers building next-generation infrastructure.

Market and Industry Impact

Investors responded positively to the news, sending Marvell’s stock up as much as 13 percent during mid-day trading according to reports from Marvell. The partnership is seen as a major step toward supporting cloud providers who require bespoke silicon optimized for specific AI tasks. By locking these custom designs into the NVLink ecosystem, NVIDIA secures its role as the primary architect of future compute clusters.