NVIDIA and Cisco Introduce New Pluggable Optics Form Factor to Scale AI Factories

What was announced

NVIDIA and Cisco introduced Ultra-Dense Small Form Factor Pluggable (USFP), described as a new high-density 400Gb/s-per-lane optical module intended for pluggable transceivers in data centers. The announcement states USFP is designed to enable a move from 200 Gb/s per lane to 400Gb/s-per-lane Ethernet AI fabrics and to provide up to 12.8Tb/s of overall capacity in an OSFP-sized footprint.

Sources: S1

Technical details highlighted by the issuers

According to the release, the first planned USFP-based transceivers will pack 16 full-duplex lanes in an octal small form factor (OSFP)-sized footprint, with the ability to expand to 32 and more lanes in the future. NVIDIA and Cisco say USFP delivers 4x the bandwidth density and 1.3x the power efficiency of 200Gb/s-per-lane pluggable modules and supports power dissipation of up to 100 watts without a dedicated embedded cooling circuit inside the transceiver. The design uses a new two-dimensional array connector with a two-step insertion feature and is presented as compatible with both switches and SuperNICs.

Sources: S1

Ecosystem and deployment claims

The companies said they are working with TE Connectivity and Amphenol to build cages, connectors and modules for USFP and plan to create an open industry specification via a community-led multi-source project. NVIDIA and Cisco further stated the initiative builds on efforts to ensure NVIDIA Spectrum-X Ethernet networking and Cisco Silicon One-based Ethernet networks interoperate.

Sources: S1

Why the companies say it matters (inference)

The release frames USFP as a response to AI factories' need for more bandwidth per connection and more connections per switch; the stated benefits include higher front-panel bandwidth density in the same network footprint and lower network power per bit versus current 200Gb/s-per-lane pluggable modules. If adopted broadly, the issuers argue, the form factor could allow operators to increase compute scale without adding network tiers.

Sources: S1

What remains uncertain

• All technical performance, efficiency and deployment benefits are reported by the issuers in the press release and are not independently verified here.
• Timing for vendor products, customer availability, and broader ecosystem adoption are not specified in the supplied source.
• Statements about future lane-count expansions, power and density outcomes describe planned or potential capabilities rather than completed field deployments.