> ## Content Index > Fetch the complete content index at: https://zipmex.com/llms.txt > Use this file to discover other available public pages before exploring further. # NVIDIA and Cisco plan USFP optics: 4× bandwidth claimed, about 1.3× efficiency - URL: https://zipmex.com/blog/nvidia-and-cisco-plan-usfp-optics-4x-bandwidth-claimed-about-1-3x-efficiency/ - Published: 2026-10-13T01:42:23.000Z - Updated: 2026-10-14T13:09:04.000Z - Description: NVIDIA and Cisco plan USFP pluggable optics at 400Gb/s per lane. See the claimed 4x density and about 1.3x efficiency, and what is still unverified. - Author: ZIPMEX - Tags: AI, #zipmex-newsroom NVIDIA and Cisco on Oct. 12 [announced plans](https://nvidianews.nvidia.com/news/nvidia-and-cisco-introduce-new-pluggable-optics-form-factor-to-scale-ai-factories?ref=zipmex.com) for Ultra-Dense Small Form Factor Pluggable (USFP), a new optical module designed for 400 gigabits per second (Gb/s) per lane. USFP is a high-density module design for pluggable transceivers in the next generation of AI factories. Its two headline numbers do not move together. Against 200Gb/s-per-lane pluggable modules, the companies claim four times the bandwidth but only about 1.3 times the power efficiency. **Key facts from the announcement** - **Lane rate:** 400Gb/s per lane, against a 200Gb/s-per-lane baseline. - **First modules:** 16 full-duplex lanes in an octal small form factor (OSFP)-sized footprint, for switches and SuperNICs. - **Later modules:** Room to expand to 32 and more lanes. - **Overall capacity:** 12.8 terabits per second (Tb/s). - **Power:** Up to 100 watts of dissipation, with no separate embedded cooling circuit in the transceiver. - **Partners:** TE Connectivity and Amphenol, with more partners expected to join. ## Is four times the bandwidth four times better? Not by NVIDIA and Cisco's own figures. Their comparison sets an OSFP-sized USFP module against 200Gb/s-per-lane pluggable modules. On that baseline, USFP is claimed to deliver four times the bandwidth. Front-panel bandwidth density also rises fourfold. Power efficiency improves by approximately 1.3 times. The bandwidth and density figures share one multiple, and power efficiency has a much smaller one. ![Screenshot of the nvidianews.nvidia.com page with the cited passage highlighted](https://storage.ghost.io/c/d4/f3/d4f34881-0300-4bbe-b8df-78a26d28be4d/content/images/2026/10/zipmex-shot-m-bf0ed5f9a7e68203.jpg) Screenshot of [nvidianews.nvidia.com](https://nvidianews.nvidia.com/news/nvidia-and-cisco-introduce-new-pluggable-optics-form-factor-to-scale-ai-factories?ref=zipmex.com), captured Oct 13, 2026\. Highlighted: the passage this section cites. ## What NVIDIA and Cisco announced The first planned USFP transceivers would pack 16 full-duplex lanes into an OSFP-sized footprint, the joint announcement says. They would serve both switches and SuperNICs. The design leaves room to grow to 32 lanes and more in later versions. With USFP, customers will be able to reach 12.8Tb/s of overall capacity, it says. ![Screenshot of the nvidianews.nvidia.com page with the cited passage highlighted](https://storage.ghost.io/c/d4/f3/d4f34881-0300-4bbe-b8df-78a26d28be4d/content/images/2026/10/zipmex-shot-m-b7025a5fe5e7db2f.jpg) Screenshot of [nvidianews.nvidia.com](https://nvidianews.nvidia.com/news/nvidia-and-cisco-introduce-new-pluggable-optics-form-factor-to-scale-ai-factories?ref=zipmex.com), captured Oct 13, 2026\. Highlighted: the passage this section cites. USFP and baseline lane rates/ DATAUSFP and baseline lane ratesNVIDIA and Cisco announcement, Oct. 12, 2026USFP planned400 Gb/sCited baseline200 Gb/s0 Gb/s125 Gb/s250 Gb/s375 Gb/s500 Gb/sCompany announcement; USFP is planned and the baseline is cited.Source: NVIDIA, Oct 12, 2026 ![Planned USFP: 400 Gb/s per lane; cited pluggable-module baseline: 200 Gb/s per lane, according to the NVIDIA and Cisco announcement.](https://storage.ghost.io/c/d4/f3/d4f34881-0300-4bbe-b8df-78a26d28be4d/content/images/2026/10/zipmex-d578039861be7572.png) Planned USFP lane rate compared with the announcement’s cited pluggable-module baseline. data: NVIDIA newsroom, Oct. 12, 2026 · ZIPMEX chart; data: nvidianews.nvidia.com ## The numbers side by side Beyond the lane rate, the figures come in units that do not mix. Overall capacity is stated in Tb/s. Density and power efficiency appear only as multiples of the 200Gb/s-per-lane baseline. Power dissipation is given in watts. Lane counts are stated for USFP but not for the baseline modules. ## USFP against 200Gb/s-per-lane modules, measure by measure How do planned USFP modules compare with 200Gb/s-per-lane pluggable modules on each measure NVIDIA and Cisco state? | Module | Lane rate (Gb/s per lane) | Full-duplex lanes per module (lanes) | Overall capacity claimed (Tb/s) | Bandwidth density vs baseline (×) | Power efficiency vs baseline (approx.) (×) | Power dissipation supported, up to (W) | Source | | --------------------------------------------- | ------------------------- | ------------------------------------ | ------------------------------- | --------------------------------- | ------------------------------------------ | -------------------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | First planned USFP modules | 400 | 16 | 12.8 | 4 | 1.3 | 100 | Source: [NVIDIA, Oct 12, 2026](https://nvidianews.nvidia.com/news/nvidia-and-cisco-introduce-new-pluggable-optics-form-factor-to-scale-ai-factories?ref=zipmex.com) | | Later USFP modules (32 and more lanes) | 400 | 32 | Not stated | Not stated | Not stated | Not stated | Source: [NVIDIA, Oct 12, 2026](https://nvidianews.nvidia.com/news/nvidia-and-cisco-introduce-new-pluggable-optics-form-factor-to-scale-ai-factories?ref=zipmex.com) | | 200Gb/s-per-lane pluggable modules (baseline) | 200 | Not stated | Not stated | Baseline | Baseline | Not stated | Source: [NVIDIA, Oct 12, 2026](https://nvidianews.nvidia.com/news/nvidia-and-cisco-introduce-new-pluggable-optics-form-factor-to-scale-ai-factories?ref=zipmex.com) | USFP values are planned or company-stated, and NVIDIA and Cisco give only the lane rate for the baseline modules. ## How the connector works USFP also brings a new connector, according to the announcement. It is a two-dimensional array connector. Insertion happens in two steps. The module is seated first. The high-speed contacts engage only after that. This reduces contact wear, improves signal integrity and makes insertion force more predictable, it says. ## Power and cooling The companies say USFP can handle power dissipation of up to 100 watts. The transceiver needs no separate, dedicated embedded cooling circuit to reach that level. In their account, this reduces switch design complexity and improves reliability. ## Who is building it NVIDIA and Cisco are working with TE Connectivity and Amphenol to build USFP cages, connectors and modules. More partners are expected to join. The two companies also plan an open industry specification for USFP. It would come out of a community-led multi-source project that they describe as upcoming. ## Analysis: a capacity gain more than an energy saving The two multiples describe a capacity gain more than an energy saving. In the companies' comparison, bandwidth rises fourfold. Power efficiency rises by about 1.3 times. If both figures hold, each module would move far more data per second. It would also draw more total power than a baseline module, not less. Front-panel bandwidth density carries the same fourfold claim as bandwidth. Support for up to 100 watts of power dissipation is a separate claim. It describes how much power dissipation USFP can support, not how much a module will draw in use. For a network planner, the claims point to more capacity per switch rather than a smaller power bill. ## What the announcement leaves open The 4× and approximately 1.3× figures are company claims without independent tests. No date is given for the first modules or for the specification project. The announcement also does not explain how the 12.8Tb/s total is counted. ## Where to read more [NVIDIA's newsroom post](https://nvidianews.nvidia.com/news/nvidia-and-cisco-introduce-new-pluggable-optics-form-factor-to-scale-ai-factories?ref=zipmex.com) on the USFP plan gives the full connector description, the named partners and the plan for an open specification. Sources: [NVIDIA, Oct 12, 2026](https://nvidianews.nvidia.com/news/nvidia-and-cisco-introduce-new-pluggable-optics-form-factor-to-scale-ai-factories?ref=zipmex.com) Promotion · ZIPMEX ### Trade **Nvidia** perpetuals on ZIPMEX 135 marketsCrypto · stocks · gold · oil · FXPTS on every tradePTS → ZIPMEX tokens [Trade NVDA-PERP](https://trade.zipmex.com/perps/PERP%5FNVDA%5FUSDC?utm%5Fsource=zipmex-news&utm%5Fcampaign=perps&utm%5Fmedium=article&utm%5Fcontent=story-186671) long or short · 24/7 [See all markets](https://trade.zipmex.com/perps/markets?utm%5Fsource=zipmex-news&utm%5Fcampaign=perps&utm%5Fmedium=article&utm%5Fcontent=story-186671) 135 markets on trade.zipmex.com Perpetual futures are leveraged derivatives. Prices can move fast and you can lose all of your margin. Not available in every jurisdiction. Not investment advice. ZIPMEX promotes trading on trade.zipmex.com. Promotions are labelled and kept separate from news coverage, which is selected and checked without regard to them. Perpetual futures are leveraged derivatives. Prices can move fast and you can lose all of your margin. Not available in every jurisdiction. Not investment advice.