NVIDIA and Cisco have introduced the Ultra-Dense Small Form Factor Pluggable (USFP), a high-density optical module capable of delivering 400Gb/s per lane. This new technology aims to support the evolving hardware needs of next-generation AI factories by utilizing pluggable transceivers.
As the demand for AI models and inference grows, these factories require efficient connectivity among computing systems while managing network complexities, power consumption, and costs. The USFP module allows for increased data transfer efficiency without necessitating a complete overhaul of existing data center infrastructure.
The USFP module is designed to provide four times the bandwidth and front-panel density compared to existing 200Gb/s-per-lane pluggable modules, while also achieving approximately 1.3 times greater power efficiency. This advancement means AI developers can scale computational resources more effectively within existing network configurations, resulting in reduced energy consumption related to networking.
“Building the world’s largest AI factories starts with rethinking the smallest of connections. With USFP, NVIDIA and Cisco are building an ecosystem to reinvent pluggable optics — connecting more computing with greater efficiency to power the next generation of AI factories,” stated Gilad Shainer, Senior Vice President at NVIDIA.
Bill Gartner, Senior Vice President and General Manager of Optical Systems and Optics at Cisco, added, “AI builders need an interconnect architecture that can scale performance without locking them into a more complex operating model. Cisco and NVIDIA recognized that scaling pluggable optics to 400Gb/s per lane would require more than another incremental extension of today’s module designs.”
For the development of the USFP, NVIDIA and Cisco are collaborating with TE Connectivity and Amphenol to produce the necessary cages, connectors, and modules. Additional partners are anticipated to join this initiative, creating a shared foundation that encourages the development of high-density pluggable infrastructure for 400Gb/s-per-lane fabrics.
This partnership builds on previous efforts between NVIDIA and Cisco to merge NVIDIA Spectrum-X Ethernet networking with Cisco’s Silicon One-based networks. These advancements aim to simplify AI infrastructure deployment across various sectors, including enterprises and cloud solutions. Moreover, NVIDIA and Cisco plan to establish an open industry specification for USFP through an upcoming community-led multi-source project.
