NVIDIA NVLink Fusion: A Unified Platform for Custom XPUs and AI Factories

This video introduces NVIDIA NVLink Fusion, a platform designed to connect custom XPUs and CPUs to the NVIDIA AI infrastructure. It explains how this technology addresses the challenges of deploying specialized AI chips at scale by providing a standardized rack-scale architecture.

■ Key Components of NVLink Fusion
- Integration of NVLink scale-up networking, Photonics, NVHBM, and NVLink-C2C
- Built on the proven MGX rack-scale architecture

■ Benefits for Silicon Innovators
- Increased performance and faster time-to-market
- Lower deployment risk and support for both XPUs and GPUs in the same AI factory

■ Partner Perspectives
- NVIDIA platform is vertically integrated but horizontally open
- Focus on custom compute while leveraging NVIDIA's proven infrastructure

This video is intended for technology professionals and decision-makers interested in AI infrastructure and custom chip deployment. Viewers will gain an understanding of how NVLink Fusion can streamline the creation of AI factories and reduce integration challenges.