Nvidia and Cisco jointly unveil USFP pluggable optics standard
- Boom: USFP supports 400 Gb/s per lane, scaling to 12.8 Tb/s total capacity
- Neutral: Cisco and Nvidia jointly announced USFP at OCP 2026 on October 12
- Boom: Cisco unveiled new ICs for AI fabrics alongside the USFP specification
- Boom: MGX 1.2 and DSX OS received updates tied to the USFP launch
- Boom: USFP is positioned as the first single-lane 400 Gb/s pluggable optical module
The story in full
On October 12, 2026, Nvidia and Cisco announced a new pluggable optics form factor called USFP, introduced at OCP 2026. The specification supports 400 Gb/s per lane and can scale to 12.8 Tb/s total capacity, targeting AI data centers and AI factory interconnects. Cisco also unveiled accompanying ICs for AI fabrics alongside the USFP announcement.
The USFP module is described as a first single-lane 400 Gb/s pluggable optical module, positioning it as a next-generation standard for AI fabric connectivity. The announcement was accompanied by updates to Cisco's MGX 1.2 and DSX OS platforms, suggesting the specification is being integrated into broader Cisco product lines.
Analysis
371 wordsOn October 12, 2026, Nvidia and Cisco jointly announced a new pluggable optics specification called USFP at the Open Compute Project conference. The standard introduces the first single-lane 400 Gb/s pluggable optical module, with total capacity scaling to 12.8 Tb/s. Cisco paired the announcement with new integrated circuits designed for AI fabric connectivity and issued updates to its MGX 1.2 and DSX OS platforms, suggesting the specification is being woven into existing Cisco product lines rather than offered as a standalone component.
The announcement matters because interconnect bandwidth has become a practical ceiling on how large and fast AI training clusters can grow. Moving from 200 Gb/s per lane to 400 Gb/s per lane effectively doubles single-lane throughput, and proponents describe it as quadrupling bandwidth density compared to current modules. That distinction, bandwidth per unit of physical space, is significant in dense data center environments where rack space and cabling complexity are real constraints. What remains in dispute is who actually benefits from that headroom: whether the gains flow back to general-purpose computing or remain locked inside the largest AI infrastructure deployments.
The pro-AI camp has received the announcement as a meaningful infrastructure milestone. Accounts including GadgetBond.com and CryptoFox News highlighted the four-times bandwidth density improvement, while K. Azwan framed USFP as delivering the open, resilient Ethernet foundations needed to scale distributed AI factories. The anti-AI camp is sharply skeptical of the broader direction. Pookleblinky tied hardware cost escalation to a future where gaming-grade computing requires tens of thousands of dollars in components, while Nihera argued that PC gaming has effectively collapsed under the weight of AI hardware prioritization. Hublai called for the physical destruction of Nvidia-based data centers, and IMPEACH and REMOVE questioned whether data center expansion is aimed at public benefit or large-scale manipulation. The middle ground, represented by Dr. Cassandra Mudgway, holds that even slowing the buildout would not be sufficient without structural commitments ensuring these facilities serve people and the environment.
The argument is unlikely to resolve around a single event, but adoption signals from other hardware vendors and any formal industry body endorsement of USFP as a shared standard would indicate whether the specification becomes a broad infrastructure baseline or remains a joint Nvidia-Cisco proprietary effort.
What Pro-AI voices are sayingNvidia and Cisco's USFP module is framed as a major networking leap, delivering 400Gb/s per lane and four times the bandwidth density of current modules to scale AI factory infrastructure.
Quote 1 of 5What Anti-AI voices are sayingCritics see AI data center expansion as serving harmful or exploitative ends rather than public benefit, with some calling for direct action against Nvidia-based facilities. A minority frames the hardware cost as making consumer technology unaffordable.
Quote 1 of 4What Middle Ground voices are sayingSimply pausing AI data center development is insufficient without ensuring these facilities are built to benefit people and the planet.
Top quoteAdd your take
0 reader votesSign in with Google to pick a side and post. Your vote moves the story's Doom / Boom score.
More Pro-AI reactions (4)
“おはようございます。 NVIDIAとCiscoが、新しい光通信の技術について発表しました。 AIの工場をより大規模に動かすための、 新しい仕組みを開発するそうです。 次世代のAIがどのように進化していくのか、 今からとても楽しみですね。 技術の進歩が、私たちの未来を どのように変えていくのでしょうか。 nvidianews.nvidia.com #news”
カマドさん(仮), Bluesky · 19:02 UTC“NVIDIAとCiscoが、AI工場をパワーアップさせる新しい通信部品を作るって発表したよ! その名も「USFP」!めちゃくちゃ高速なデータ転送ができるモジュールなんだって✨ これがあれば、次世代のAIインフラがもっとすごいことになるはず。 テクノロジーの進化が止まらない感じ、マジでアツいよね! 早朝からすごいニュースが入ってきたね!これからの動き、絶対チェックしておかないと損かも! nvidianews.nvidia.com #news”
昔が好きなヒストリア, Bluesky · 19:08 UTC“The new USFP form factor aims to deliver four times the bandwidth density of 200Gbps-per-lane optical modules.”
GadgetBond.com, Bluesky · 17:49 UTC“Delivers the resilient, open Ethernet foundations required to connect and scale massive, distributed AI factories”
K. Azwan, Bluesky · 15:25 UTC
More Anti-AI reactions (3)
“I love that PC Gaming is just dead because of a technology that is most prominent use case is... *checks notes* CSMA generation, racist content, destroying industry and dissolving human worker's rights, ruining cognitive function etc etc.”
Nihera, Bluesky · 13:04 UTC“maybe data centers and AI are less about a utopian future, and more about mind control on a massive scale.”
IMPEACH & REMOVE, Bluesky · 08:00 UTC“I suggest we repurpose the word and it's abbreviation "uke" (as in "nuke") to mean conventional weapons employed to trash #warlike installations. This includes ALL #nVidia based #datacenter whatsoever”
Hublai (charismatic megafauna), Bluesky · 15:16 UTC
No more Middle Ground reactions
Sources
15 articles from 15 outlets- 富途牛牛Reshaping the Landscape of AI Factory Interconnects: NVIDIA and Cisco Unveil First Single-Lane 400Gb/s Pluggable Optical Module
- HPCwireNVIDIA and Cisco Introduce New Pluggable Optics Form Factor to Scale AI Factories
- Fierce NetworkNVIDIA and Cisco intro pluggable optics form factor for AI factories
- StorageReview.comNVIDIA and Cisco USFP Optics Target 400Gb/s per Lane, With MGX 1.2 and DSX OS Updates at OCP 2026
- PluangNVIDIA and Cisco launch USFP optics for 400Gb/s lanes to boost AI data center efficiency and scale.
- Quiver QuantitativeNVIDIA and Cisco Announce USFP, a 400Gb/s-per-Lane Optical Module for Next-Generation AI Fabrics
- Investing.com CanadaNvidia and Cisco announce new optical module for AI data centers
- Cisco NewsroomNVIDIA and Cisco Introduce New Pluggable Optics Form Factor to Scale AI Factories
- NVIDIA NewsroomNVIDIA and Cisco Introduce New Pluggable Optics Form Factor to Scale AI Factories
- Cisco BlogsCisco and NVIDIA Announce USFP: Expanding the Path to AI Connectivity
- StreetInsiderNvidia and Cisco unveil new USFP optical module for AI data centers
- GamesBeatNvidia and Cisco unveil pluggable optics for AI factories
- GlobeNewswireNVIDIA and Cisco Introduce New Pluggable Optics Form Factor to Scale AI Factories
- Techzine GlobalNvidia and Cisco develop USFP form factor: up to 12.8 Tb/s capacity
- All About CircuitsCisco Unveils ICs for AI Fabrics and Nvidia-Codeveloped USFP Optics Spec

