Cisco and NVIDIA have announced plans to develop Ultra-Dense Small Form Factor Pluggable (USFP), a high-density optical module designed to support the next generation of AI factories.
The new module will deliver 400Gb/s per lane, enabling AI infrastructure operators to connect more computing systems while managing network complexity, power consumption and costs.
As AI models become larger and demand for inference grows, data centres need greater network capacity without significantly increasing their infrastructure footprint. USFP aims to address this by helping AI factory builders move from 200Gb/s-per-lane to 400Gb/s-per-lane Ethernet AI fabrics while maintaining a deployment model already used across multiple generations of data centre networks.
Comparable in size to an OSFP module, USFP is designed to deliver four times the bandwidth and front-panel bandwidth density, along with approximately 1.3 times the power efficiency of 200Gb/s-per-lane pluggable modules. This would allow operators to scale computing capacity within the same network footprint while reducing the network’s share of overall AI factory power consumption.
“Building the world’s largest AI factories starts with rethinking the smallest of connections,” said Gilad Shainer, Senior Vice President at NVIDIA. He added that the collaboration aims to develop an ecosystem that connects more computing systems with greater efficiency.
Bill Gartner, Senior Vice President and General Manager of Optical Systems and Optics at Cisco, said the industry needs interconnect architecture that can increase performance without adding operational complexity. He noted that USFP is designed to improve bandwidth density while retaining the serviceability, deployment flexibility and operational simplicity of pluggable optics.
Building an Ecosystem for AI Factories
Cisco and NVIDIA are working with TE Connectivity and Amphenol to develop USFP cages, connectors and modules. Additional partners are expected to join the initiative.
The shared form factor will provide suppliers with a common foundation for developing high-density pluggable infrastructure for 400Gb/s-per-lane networks, supporting a broader hardware ecosystem for AI factory operators.
The collaboration also builds on the companies’ existing work to ensure interoperability between NVIDIA Spectrum-X Ethernet networking and Cisco Silicon One-based Ethernet networks. The effort aims to simplify AI infrastructure deployment for enterprises, neocloud providers and sovereign cloud operators.
Cisco and NVIDIA also plan to establish an open industry specification for USFP through an upcoming community-led multi-source project.
