NVIDIA and Cisco have announced a new pluggable optical module design aimed at connecting more computing equipment inside AI data centres, with up to 12.8 terabits per second of capacity. The proposed USFP standard targets more bandwidth in a familiar-sized module, potentially helping AI operators scale networks without adding another layer of switches.
NVIDIA Watch analysis
What happened
The companies say their Ultra-Dense Small Form Factor Pluggable design supports 400-gigabit-per-second electrical signals per lane. The first planned modules use 16 full-duplex lanes in an OSFP-sized footprint, with a path to 32 lanes or more. NVIDIA and Cisco say the design could deliver four times the bandwidth density and about 1.3 times the power efficiency of 200Gb/s-per-lane pluggable modules.
The companies are working with TE Connectivity and Amphenol on modules, cages and connectors, and plan to establish an open industry specification through a community-led, multi-source project. This is a design and ecosystem announcement, not a claim that finished USFP products are already on sale.
Why it matters
AI data centres need to connect growing numbers of computing systems while managing power, cost and network complexity. NVIDIA and Cisco say USFP could increase capacity within the same network footprint and avoid adding network tiers, which can bring extra links, power use and latency.
The design also aims to retain the serviceability of pluggable optics. Its connector is built for 400Gb/s-per-lane performance, and the companies say it can support power dissipation of up to 100 watts without a separate cooling circuit inside the transceiver. Those details matter to operators weighing higher density against the practical business of maintaining equipment.
Our read
The headline number is eye-catching; the more consequential part may be the planned multi-source specification. A shared design supported by several suppliers could give AI infrastructure builders more choice than a tightly controlled, single-vendor route. But the announcement describes an intended ecosystem, not one that exists at scale yet. The rack-level test will be whether the promised density and efficiency arrive without making deployment or servicing more awkward.
What to watch
- When the first USFP modules and systems become available.
- Whether the planned open specification attracts additional suppliers.
- How measured power, reliability and performance compare in real deployments.
Discussion spark: For the next generation of AI data centres, should buyers prioritise a shared, multi-supplier optics standard, or the performance of a more tightly controlled system?
Sources and evidence
- NVIDIA and Cisco Introduce New Pluggable Optics Form Factor to Scale AI Factories – NVIDIA Newsroom (12 October 2026, 15:03 UTC)
Independent WittyWires tracker for public updates about NVIDIA. Not affiliated with or endorsed by NVIDIA; this is not an official account.