NVIDIA and Cisco announced plans on Oct. 12, 2026, for a new high-density optical module form factor called Ultra-Dense Small Form Factor Pluggable (USFP).
The module is designed to support 400Gb/s-per-lane electrical signals, allowing AI factory operators to scale Ethernet fabrics to a total capacity of 12.8Tb/s. By increasing bandwidth per lane and lane counts, the technology allows for the connection of more 1.6Tb/s endpoints without adding network tiers, which typically increase latency and power costs, according to NVIDIA Newsroom.
USFP replaces the performance profile of 200Gb/s-per-lane optics. NVIDIA stated that USFP provides 4x the bandwidth density and approximately 1.3x the power efficiency of those earlier modules. The module uses an octal small form factor (OSFP)-sized footprint and initially supports 16 full-duplex lanes, with plans to expand to 32 or more lanes in the future, NVIDIA said.
The module features a two-dimensional array connector with a two-step insertion design. According to NVIDIA, this design is intended to improve signal integrity and reduce contact wear by engaging high-speed contacts only after the module is inserted. USFP supports power dissipation of up to 100 watts, achievable without a dedicated embedded cooling circuit within the transceiver, which reduces switch design complexity. The announcement does not include pricing or a release date.
NVIDIA and Cisco are collaborating with Amphenol and TE Connectivity to produce USFP connectors, cages, and modules. The companies intend to establish an open industry specification through a community-led multi-source project, NVIDIA said.
This development follows other collaborations between the two companies to integrate Cisco Silicon One-based Ethernet networks with NVIDIA Spectrum-X networking. Cisco said in August it was expanding its Secure AI Factory with NVIDIA and Supermicro to include rack-scale computing solutions.