Cisco and Nvidia Unveil USFP Optical Module With 12.8 Tbps Bandwidth for AI Datacenters
16 Sources
[1]
Cisco, Nvidia team up to build fatter, faster pluggable
AI datacenters are a rats nest of fiber optic cables stitching together thousands of AI accelerators, NICs, switches, and everything else required to train and run the latest and greatest modes. Cisco hopes to untangle this mess with a new pluggable optic developed in collaboration with Nvidia
[2]
Nvidia and Cisco unveil next-gen network interface built for the AI data center boom
Serving tech enthusiasts for over 25 years. TechSpot means tech analysis and advice you can trust. In Layman's Terms: New data center projects are emerging everywhere, with no end in sight. Traditional data center interconnects are becoming increasingly inefficient, which is why Nvidia and Cisco
[3]
Tiny optical module boosts data center bandwidth 4x without extra gear
NVIDIA and Cisco have announced plans for a new optical module that could deliver up to 12.8 terabits per second (Tbps) of capacity, offering four times the bandwidth density of existing 200-gigabit-per-second-per-lane technology for data center networks. Called Ultra-Dense Small Form Factor
[4]
Cisco and NVIDIA Announce USFP: Expanding the Path to AI Connectivity
The Ultra-Dense Small Form-Factor Pluggable (USFP) is industry's first pluggable form factor designed to help customers scale AI networks, providing a path to up to 12.8T of aggregate capacity while preserving proven deployment and serviceability models As AI evolves from answering prompts to
[5]
NVIDIA and Cisco Introduce New Pluggable Optics Form Factor to Scale AI Factories
New Transceivers Enable AI Builders to Scale Compute Within the Same Network Footprint While Reducing Networking's Share of the Power Budget * Ultra-Dense Small Form Factor Pluggable (USFP) is designed to be the first module to support 400Gb/s-per-lane electrical signals, enabling capacity up to
[6]
AI infrastructure networking moves from copper to optics
Copper's days are numbered inside AI racks as Cisco bets on optics AI infrastructure networking is approaching a physical turning point as the copper links that tie GPUs together inside a rack strain under ever-faster signaling speeds. The move to light-based interconnects could redraw the
[7]
Cisco Systems scales programmable AI network capacity with new silicon and repluggable optics innovations
Cisco Systems scales programmable AI network capacity with new silicon and repluggable optics innovations Cisco Systems Inc. took to the stage at the annual Open Compute Project Global Summit today, unveiling a new suite of open silicon and optical networking systems as it delivered one of the
[8]
OCP 2026: Cisco Expands AI Networking Stack With New Silicon One And Nvidia Optics Push
At OCP Global Summit 2026, Cisco unveiled new Silicon One chips and systems, expanded SONiC support and partnered with Nvidia on a new open optics standard, all aimed at helping customers scale AI infrastructure while maintaining openness, interoperability and choice. This year at the 2026 Open
[9]
NVIDIA And Cisco Cram 4x OSFP's Bandwidth Density Into An OSFP-Sized Module, Pushing AI Fabrics To 12.8 Tb/s With USFP
NVIDIA & Cisco have partnered to create USFP, an ultra-dense and compact form factor that drives up to 400 Gb/s speeds. NVIDIA's USFP Is A Compact & Pluggable Switch That Harnesses Optics To Drive Up To 400 Gb/s Per Lane & 12.8 Tb/s Capacity The need for advanced fabric, networking, and
[10]
NVIDIA and Cisco Introduce High-Density USFP Modules to Scale AI Infrastructure
NVIDIA and Cisco are collaborating with TE Connectivity and Amphenol to build USFP cages, connectors and modules, with additional partners expected to join. NVIDIA and Cisco announced plans for Ultra-Dense Small Form Factor Pluggable (USFP), a new high-density 400Gb/s-per-lane optical module
[11]
Nvidia and Cisco announce new optical module for AI data centers By Investing.com
SAN FRANCISCO - Nvidia (NASDAQ:NVDA) and Cisco (NASDAQ:CSCO) announced today plans for Ultra-Dense Small Form Factor Pluggable (USFP), a high-density optical module designed for AI data center networks, according to a press release statement. The USFP module supports 400Gb/s-per-lane electrical
[12]
Cisco unveils AI networking silicon and optics with NVIDIA By Investing.com
SAN JOSE, Calif. - Cisco Systems Inc. (NASDAQ:CSCO) announced new networking silicon and a collaboration with NVIDIA to develop a high-density optical module specification at the Open Compute Project Global Summit on Monday. The company said its Silicon One P200-powered 51.2T systems are now
[13]
Nvidia and Cisco develop compact optical module for AI data centers
Nvidia and Cisco have announced the development of a new ultracompact pluggable optical module designed to strengthen the capabilities of data centers specializing in artificial intelligence. The technology is expected to increase infrastructure capacity without expanding the space occupied by
[14]
NVIDIA Corporation and Cisco Introduce New Pluggable Optics Form Factor to Scale AI Factories
NVIDIA Corporation and Cisco announced plans for Ultra-Dense Small Form Factor Pluggable (USFP), a new high-density 400Gb/s-per-lane optical module designed to equip the hardware ecosystem for the next generation of AI factories using pluggable transceivers. NVIDIA and Cisco are working with TE
[15]
Cisco, Nvidia Tackle AI's Growing Data Center Bottleneck
Cisco is partnering with Nvidia on a new generation of optical networking technology designed to handle the exploding amount of data moving through AI data centers. Cisco President and Chief Product Officer Jeetu Patel explains why increasingly powerful AI agents could consume far more network
[16]
Nvidia, Cisco Release New Optical Module With Higher Capacity
Nvidia and Cisco are introducing a new optical module designed to support the networking needs of growing artificial-intelligence infrastructure. The two companies on Monday announced the Ultra-Dense Small Form Factor Pluggable, a product they described as the first optical module to support
Share
Copy Link
Cisco and Nvidia announced USFP (Ultra-Dense Small Form Factor Pluggable), the first 400G-per-lane optical module designed to deliver up to 12.8 Tbps bandwidth for AI datacenters. The next-gen network interface provides 4x the bandwidth density and 1.3x better power efficiency than current 200Gbps-per-lane modules, enabling AI infrastructure to scale without adding network tiers or overhauling existing deployment models.
Cisco and Nvidia Introduce USFP to Transform AI Connectivity
Cisco and Nvidia have jointly announced plans for the Ultra-Dense Small Form Factor Pluggable (USFP), a next-gen network interface designed to address the escalating bandwidth demands of AI datacenters
1
2
. Unveiled at the Open Compute Project (OCP) global summit, USFP represents the first pluggable optics form factor capable of supporting 400G-per-lane electrical signals, delivering up to 12.8 Tbps of aggregate capacity4
5
. This development addresses a critical challenge facing AI infrastructure: connecting thousands of AI accelerators, NICs, and switches while managing power consumption, network complexity, and operational costs1
.Delivering 4x Bandwidth Density and Enhanced Power Efficiency
USFP promises to deliver four times the high-bandwidth density of existing 200Gbps-per-lane pluggable optics while providing approximately 1.3x greater power efficiency
2
3
. The first planned USFP-based optical module will pack 16 full-duplex lanes into a footprint roughly the size of existing OSFP modules, delivering 6.4 Tbps of aggregate capacity4
. By doubling both the lane count and bandwidth per lane, USFP enables AI connectivity solutions to connect more 1.6 Tbps endpoints without adding network tiers, which typically increase latency, power consumption, and costs5
. Future versions could expand to 32 full-duplex lanes or more, potentially achieving the full 12.8 Tbps bandwidth capacity2
3
.Simplifying AI Datacenter Cabling and Infrastructure
Source: TechSpot
The USFP optical module addresses the tangled web of fiber optic cables that characterize modern AI datacenters
1
. A single USFP pluggable transceiver could consolidate eight GPUs worth of cabling into a single fiber optic bundle, significantly simplifying infrastructure1
. On the switch side, a future 204.8 Tbps switch would require only 16 USFP ports to support 128 1.6 Tbps endpoints1
. This consolidation helps AI infrastructure operators scale compute within the same network footprint while decreasing networking's share of the overall AI factory power budget5
. The technology supports AI workloads ranging from training and inference to emerging agentic AI systems that demand continuous communication across accelerators, applications, storage, and services4
.Engineering Innovations Enable Higher Performance
Achieving 400G-per-lane performance requires significant engineering advances. USFP employs a two-dimensional array connector with a two-step insertion mechanism that engages high-speed contacts only after the module is inserted, reducing contact wear and improving signal integrity
3
5
. The design supports power dissipation of up to 100 watts without requiring a dedicated cooling circuit inside the transceiver, simplifying switch construction while accommodating the heat generated by higher-bandwidth optical connections3
5
. Cisco acknowledges this represents a significant jump from today's high-end pluggable optics, which typically pull around 20 watts1
. The architecture also supports optical designs that reduce latency and power consumption by simplifying signal processing5
.Building an Open Industry Standard Ecosystem
Source: Wccftech
Cisco and Nvidia are collaborating with TE Connectivity and Amphenol to develop connectors, cages, and modules for the USFP interface, with additional partners expected to join
2
5
. The companies plan to establish USFP as an open industry standard through a community-led multi-source project, ensuring anyone can use the technology in their designs1
4
. This approach preserves the independently field replaceable modules, mature supply chains, and interoperability across suppliers that data center operators value4
. "Building the world's largest AI factories starts with rethinking the smallest of connections," said Gilad Shainer, senior vice president at Nvidia3
5
.Related Stories
Nvidia's Strategic Interest Beyond Co-Packaged Optics
Nvidia's participation in USFP development is notable given the company has largely moved toward co-packaged optics in its most recent Spectrum-X and Quantum-X switches, which deliver roughly 30 percent power savings by eliminating pluggables on the switch side
1
. However, Nvidia's NICs still rely on pluggable optics to offer integrators greater flexibility, particularly for smaller deployments using top-of-rack switches where direct attach copper cabling remains viable1
. The timing aligns with Nvidia's Vera Rubin Ultra platform, teased in early 2025, which promises 115.2 Tbps scale-out bandwidth per rack for 600 kW racks packing 144 GPUs, translating to approximately 3.2-6.4 Tbps of bandwidth per GPU package1
. Cisco's initial USFP spec calls for 16 400 Gbps links per pluggable, delivering exactly 6.4 Tbps of bandwidth1
. The first USFP transceivers will support both switches and SuperNICs, specialized accelerator cards that move data directly between GPUs using the RoCE protocol2
.Microsoft Adopts Cisco Silicon for Multi-Datacenter AI
Source: SiliconANGLE
Alongside the USFP announcement, Cisco revealed that Microsoft is using its Silicon One P200 routing chips to distribute AI workloads across multiple datacenters in what Nvidia calls "scale across networking"
1
. Cisco touts the P200 as the first 51.2 Tbps systems deployed in production by leading hyperscalers1
. Microsoft was among the first major hyperscalers to embrace scale-across networking with its Fairwater clusters in Wisconsin and Georgia, using this approach to overcome local power constraints1
.What This Means for AI Infrastructure Evolution
USFP represents a strategic bet that pluggable optics will remain relevant alongside co-packaged optics and other emerging optical architectures in AI infrastructure
4
. Bill Gartner, senior vice president at Cisco, emphasized that "AI builders need an interconnect architecture that can scale performance without locking them into a more complex operating model"5
. The technology allows organizations to embrace next-generation optical technology without overhauling their data center infrastructure or abandoning established operational models4
5
. However, the companies have not disclosed independent test results, production volumes, pricing, or commercial launch dates for USFP modules3
. Watch for how quickly the ecosystem of TE Connectivity, Amphenol, and other manufacturers can deliver compatible components, and whether USFP's power efficiency claims hold up in production deployments as AI factories continue scaling toward increasingly power-constrained environments.References
Summarized by
Navi
[1]
[3]
Related Stories
Nvidia Unveils Revolutionary Silicon Photonics Switches for AI Data Centers
19 Mar 2025•Technology
Nvidia Unveils Groundbreaking Silicon Photonics Technology for Next-Gen AI Data Centers
25 Aug 2025•Technology
Cisco unveils Silicon One G300 chip to challenge Broadcom and Nvidia in AI networking race
10 Feb 2026•Technology
Recent Highlights
1
Anthropic AI Development Doubles Speed as Self-Improving Systems Near Breakthrough
Technology
2
Google's AI Policy Forces Schools to Accept AI Overviews or Lose Search Access Entirely
Technology
3
Jensen Huang Scandal: Nvidia CEO's Affair with Suspected Chinese Spy Raises National Security Concerns
Policy and Regulation
Recent Highlights
Today's Top Stories
Don’t drown in AI news. We cut through the noise - filtering, ranking and summarizing the most important AI news, breakthroughs and research daily. Follow topics that matter to you and stay ahead.
