Marvell and NVIDIA Collaborate to Offer Next-Gen AI Infrastructure with NVLink Fusion Technology
Marvell Technology, Inc., a leading semiconductor company specializing in data infrastructure, has announced a new partnership with NVIDIA to offer NVIDIA NVLink Fusion technology.
This collaboration will enable customers using Marvell's custom cloud platform silicon to integrate NVLink Fusion, providing enhanced flexibility and choice in developing cutting-edge AI infrastructure solutions. This partnership aims to drive next-generation AI capabilities, improving the performance and scalability of AI data centers.
What NVLink Fusion Brings to the Table
NVIDIA’s NVLink Fusion technology is designed to integrate custom XPU silicon with NVIDIA NVLink connectivity, offering a rack-scale architecture that includes hardware, software, and other advanced technologies.
This integration offers customers the ability to scale AI infrastructure solutions rapidly, with improved bandwidth, performance, and efficiency. It provides hyperscalers, companies that operate at vast scales, a path to efficiently scale their AI infrastructure while leveraging existing investments in NVLink.
Marvell’s Expertise in Custom Cloud Platforms
By bringing together system and semiconductor design competency with leading-edge process manufacturing, Marvell creates high-performance platforms that address a wide range of infrastructure requirements.
These platforms encompass solutions like high-bandwidth memory (HBM), optical I/O, custom silicon, and die-to-die interconnects. Along with NVIDIA's NVLink Fusion, these solutions enable faster deployment of sophisticated AI systems.
Nick Kucharewski, Senior Vice President and General Manager of the Cloud Platform Business Unit at Marvell, said, "Marvell and NVIDIA are working together to advance AI factory integration.”
He added, "Through this collaboration, we offer customers the flexibility to rapidly deploy scalable AI infrastructure with the bandwidth, performance and reliability required to support advanced AI models."
Also read: NVIDIA and Foxconn to Build Taiwan's AI Factory Supercomputer
Accelerating Hyperscaler Solutions
The combination of Marvell custom silicon and NVLink Fusion offers hyperscalers an accelerated path to custom-scale solutions for AI model training and agentic AI inference.
These capabilities are vital for AI workloads driven by learned knowledge and reasoning. By integrating their proprietary XPUs into NVIDIA’s rack-scale systems and end-to-end networking platform, cloud providers can efficiently scale AI factories to meet the growing demands of AI development.
