NVIDIA and Cisco Unveil High-Density Optical Transceiver
The new USFP technology supports 400 Gb/s-per-lane connectivity to increase bandwidth density for future AI data center networks.
Updated on Oct. 21, 2026 in Data Centers
Live Poll
Does increasing AI hardware density in your area improve network efficiency for local services?
NVIDIA and Cisco have introduced the Ultra-Dense Small Form Factor Pluggable (USFP) optical transceiver, a new hardware component designed to increase bandwidth density in AI data centers. The technology, which is currently in the announcement phase, aims to support higher-capacity Ethernet networks for large-scale AI infrastructure.
Why it matters
As AI clusters demand increasingly high-speed connectivity, the physical space and power constraints within switch front panels have become primary bottlenecks. This collaboration seeks to set an open industry specification that scales bandwidth capacity without requiring a proportional increase in power consumption.
The USFP modules integrate 16 full-duplex lanes into an OSFP-sized footprint, delivering a total capacity of 12.8 Tb/s. The hardware is designed to dissipate up to 100W while providing four times the bandwidth density of current 200 Gb/s-per-lane optics.
The players
NVIDIA
A designer of graphics processing units and high-performance networking hardware essential for training and deploying AI models.
Cisco
A developer of enterprise networking equipment and infrastructure systems, including high-capacity switches for data centers.
The details
The transceiver utilizes a two-dimensional array connector paired with a two-step insertion mechanism. This approach ensures high-speed electrical contacts only engage once the module is fully seated, protecting the interface during installation. The design manages thermal output by handling up to 100W without the need for a dedicated embedded cooling circuit.
Timeline
October 21, 2026: NVIDIA and Cisco unveiled the USFP optical technology.
The Tech Race
This collaboration challenges the current density limits of the OSFP form factor by integrating significantly higher lane counts into established footprints. It follows a broader industry trend of scaling Ethernet connectivity to keep pace with the massive throughput requirements of modern AI factories.
Data center operators can expect this hardware to provide four times the bandwidth density once the technology reaches commercial production. Users should track the upcoming open industry specification to determine when compatible hardware will be ready for integration into large-scale AI networks.
The takeaway
The move to consolidate 16 high-speed lanes into a single module signals a shift toward higher-density optical interconnects in AI compute clusters. Observers should track the release of the open industry specification and the addition of new partners to the development initiative.
Further reading
For more on the infrastructure supporting current AI clusters, visit our Data Centers section.
Source note: This article includes information reported by eeNews Europe.
Live Poll
Does increasing AI hardware density in your area improve network efficiency for local services?
