Cisco and NVIDIA have proposed USFP, an open pluggable optical format designed to increase AI data-centre bandwidth while preserving field replaceability, multi-vendor supply and familiar deployment models.
Cisco and NVIDIA have set out plans for USFP, a new pluggable optical form factor intended to support the next wave of AI infrastructure as data-centre operators push for higher bandwidth without abandoning familiar deployment practices. The companies said the Ultra-Dense Small Form-Factor Pluggable is designed to carry 400Gb/s per lane, with a route to as much as 12.8Tb/s of aggregate capacity, while keeping the serviceability and multi-vendor characteristics that have long made pluggable optics attractive.
The launch reflects the pressure created by AI workloads that now extend beyond model training into inference and more complex agentic systems, which rely on constant movement of data between accelerators, storage and applications. Cisco argued that the answer is not simply more speed, but a full-stack approach in which silicon, systems, optics, software and operations are designed together. In that framing, USFP is meant to let customers expand capacity without redesigning networks from scratch or discarding existing operating models.
According to Cisco and NVIDIA, the module will use 16 lanes at 400Gb/s to deliver 6.4Tb/s initially, with a path to 32 lanes and 12.8Tb/s. The companies said that this would deliver four times the bandwidth density of current 200Gb/s-per-lane optics and about 1.3 times the power efficiency, a combination aimed at reducing the power and cooling burden that has become a central constraint in AI data centres. The design is also intended to preserve the practical advantages of pluggable hardware, including field replaceability and established supply chains.
Cisco has positioned the project as part of a broader push for open AI infrastructure, and said it is working with NVIDIA on an open industry specification and a multi-source hardware ecosystem. TE Connectivity and Amphenol are among the companies named in the effort, underscoring the intention to build supplier diversity around the new module. That matters because many operators still weigh not only raw performance, but also interoperability, lifecycle support and the ability to scale networks without locking themselves into a single proprietary path.
Disclaimer: This content is intended for informational purposes only. Readers are advised to exercise their own judgement, conduct due diligence, or consult a qualified expert before acting on any information provided.
