NVIDIA Unveils NVLink Fusion for Industry to Build Semi-Custom AI Infrastructure With NVIDIA Partner Ecosystem

Core Insights - NVIDIA has introduced NVLink Fusion, a new silicon technology that enables industries to create semi-custom AI infrastructure, leveraging a vast ecosystem of partners [1][12] - The technology allows for the integration of NVIDIA GPUs with CPUs from companies like Fujitsu and Qualcomm, facilitating the development of high-performance AI factories [2][6] Industry Impact - A significant transformation is occurring in data centers, necessitating a fundamental rearchitecture to incorporate AI into every computing platform [3] - NVLink Fusion provides cloud providers with a scalable solution to expand AI capabilities, supporting up to millions of GPUs and delivering throughput of up to 800 Gb/s [3][4] Partner Collaborations - Key partners adopting NVLink Fusion include MediaTek, Marvell, Alchip Technologies, Astera Labs, Synopsys, and Cadence, all of which are working to create custom AI silicon [2][4][13] - MediaTek and Marvell emphasize their collaboration with NVIDIA to redefine AI factory integration and deliver scalable technologies for cloud-scale AI [5] Technological Advancements - The fifth-generation NVIDIA NVLink platform offers compute-dense racks with a total bandwidth of 1.8 TB/s per GPU, significantly outperforming PCIe Gen5 by 14 times [7] - NVIDIA Mission Control software enhances AI factory operations by automating the management of AI data centers and workloads, streamlining deployment and validation processes [9] Availability and Future Prospects - NVLink Fusion silicon design services and solutions are currently available from several partners, indicating a strong market readiness for this technology [14]