Cisco promotes scalability and openness on OCP 2026
PR Newswire
41m ago
Ai Focus
Cisco releases new chip, system, and optical interconnection innovations on OCP 2026, advancing from rack-level computing to open AI infrastructure with distributed AI capacity; the company also plans to develop a new generation of 400G per-lane pluggable optical modules with NVIDIA to create open industry standards.
Helpful
No.Help

Cisco Systems ( Cisco, NASDAQ ticker: CSCO) launched new chip, system, and optical interconnection innovations on Sunday at Open Compute Project ( OCP) Global Summit 2026, advancing from rack-level computing to open AI infrastructure with distributed AI capacity. Cisco stated that these releases demonstrate how the company is transforming networking into a cohesive, programmable architecture while maintaining the openness and choices that customers expect in the era of superintelligence.

Cisco states that these new capabilities further reflect the company's commitment to continuously delivering an open technology stack from chips to Kubernetes. This is crucial for connecting the infrastructure of hyperscale cloud service providers, neoclouds, and enterprises. The company also noted that strong market adoption has verified its leadership position and highlighted customers' deep trust in the company as the demand for AI computing power accelerates rapidly.

Kevin Wollenweber, Senior Vice President and General Manager of Cisco's Data Center and Internet Infrastructure, stated: "AI is redefining what infrastructure expansion means. It's no longer just about individual data centers or clusters; it's about how to connect computing on a large scale without sacrificing performance or resilience. Cisco's strengths in chips, networking, optics, systems, and security put us in a unique position to provide customers with the open infrastructure they need as AI continues to evolve."

“Scale-Across” connects global AI capacity

Scale-across has expanded the capability for cross-facility coordination of AI workloads. Building on the technology of traditional data center interconnection (DCI), it has introduced significantly different workload requirements. When accelerators across sites participate in the same coordination task, congestion, latency, and recovery times directly affect computational utilization and task progress. Cisco states that the company is leveraging its leadership in chips, routing, and optical networking to address these needs with a packet-to-optical architecture.

Cisco's scale-across architecture combines Sil token issuance with a deep buffer system and coherent optical interconnects, linking AI from different locations into a unified cluster. The company claims that this provides customers with a single, unified computing fabric, supported by the performance, deep programmability, and built-in security of the Sil token-issued One architecture. Cisco states that its industry-first 51.2T system, driven by Sil tokens and One P200, has been deployed in production environments by leading hyperscale cloud service providers.

Anand Ramakrishna, Vice President of Enterprise and Head of Engineering at Microsoft, stated: "Microsoft and Cisco have long maintained a partnership based on engineering collaboration and innovation. We are pleased to announce that today, Microsoft is putting into production systems driven by Cisco Sil tokens. As our infrastructure needs grow, P200's robust architecture provides the speed and enhanced resilience required to support the continuous expansion of our cloud and high-performance services. This milestone represents the next phase of our long-term collaboration, ensuring that we continue to deliver the performance and reliability needed in the modern digital age."

Dennis Cai, the person in charge of network infrastructure at Alibaba Cloud Intelligence Group and Vice President, stated: "Alibaba is proud to announce that our next-generation backbone network eCORE has reached an important milestone. We have deployed on a large scale our self-developed Quantum Cat routers, which are powered by Cisco Sil tokens issued by One P200. These routers have brought 51.2 Tbps of routing capacity to our infrastructure. They feature a fully shared deep buffering architecture, strong programmability, and excellent energy efficiency. This solution provides the exact performance and scale required for today's high-intensity cloud and AI workloads. We have combined Cisco's chips with Alibaba's customized systems and SONiC operating system. This combination proves the true value of open networks. It provides the flexibility and operational capabilities needed for some of the most demanding networks around the world. We look forward to deepening our cooperation with Cisco, Open Compute Project ( OCP ), and the SONiC community. Together, we will continue to advance open, high-performance networks for the AI era."

For the era of 1.6T, expanding AI from a rack to fabric.

Within data centers, the AI scale also imposes higher demands on chips, systems, and optical interconnects in terms of bandwidth, density, and energy efficiency. To meet the aggressive bandwidth requirements of the next generation of AI, Cisco's Sil token issuance n One product portfolio is providing next-generation network performance for large-scale AI. As a member of OCP ESUN, Cisco states that it is committed to promoting the development of open, interoperable, high-performance AI networks in large-scale scenarios and is actively collaborating with ESUN to establish 1.0 and future scale-up workload specifications.

Cisco's earlier this year announced 102.4T Sil token issuance n One G300 is now available for customers to sample, and the first batch of systems driven by G300 has also begun to be shipped to customers. Cisco is also expanding its Sil token issuance n One product portfolio, launching the first 200G SerDes exchange chip designed for top-of-rack ( ToR ) deployment with 51.2T capacity. The new Sil token issuance n One G302 provides 51.2T capacity and integrates Cisco's 1.6T optical components and direct chip liquid cooling to improve efficiency. G302 is intended to drive ToR exchange systems, which are located at the rack level and are responsible for aggregating server traffic and connecting it efficiently to larger data center networks. Cisco states that this will optimize the way servers are connected and integrated, bringing unprecedented front-end scale, density, and throughput to hyperscale cloud service providers. The company plans to begin shipping to customers by the end of 2026.

Cisco and NVIDIA launch new pluggable optical modules

Cisco also announced a plan in collaboration with NVIDIA to develop Ultra - Dense Small Form Factor Pluggable ( USFP ), which is a new type of high-density optical module that supports 400Gb/s per channel. Compared to other methods of increasing the bandwidth of optical modules, USFP relies on a mature operating mode to achieve 400Gb/s per channel. Cisco expects that USFP will bring about 4 times the bandwidth density and 1.3 times the energy efficiency. Compared to optical modules with 200Gb/s per channel, it will improve the efficiency and scalability of AI factories without sacrificing maintainability, deployment flexibility, or the ease of operation of today's pluggable optical modules. To promote widespread industry participation and adoption, Cisco and NVIDIA plan to develop open industry standards for USFP through an upcoming community-led, multi-supplier project.

Extended Security AI Network

SONiC ( Software for Open Networking in the Cloud ) is an open-source operating system that forms the foundation for many large-scale cloud service providers, as well as today's neoclouds, service providers, and enterprises. It covers the ecosystem of AI builders. Cisco customers can utilize Cisco-verified SONiC and Nexus One, which represents the next generation of data center architecture and supports an open approach for multi-site to rack-level environments. This support will be extended to customers through Live Protect starting from mid-2027. By leveraging Cisco Runtime Technology, Live Protect can reduce the exposure window that customers typically face between vulnerability disclosure and patching of Cisco products and versions. Cisco states that Live Protect will give Cisco SONiC customers time to test and deploy permanent software fixes without affecting operations.

Control and recovery on a scale of AI

Out-of-band management systems are a type of system relied upon by network operators; even if the production network fails, they can provide remote access, enabling repairs during power outages or network attack incidents. However, at the scale of AI, out-of-band management systems can become a mixture of legacy systems, tools, and proprietary interfaces, making it more difficult to manage and repair thousands of GPU nodes. Cisco states that the new Cisco 8001 is a management system driven by Sil tokens, available in 10G or 1G versions, which can address this issue by improving performance and introducing SONiC into this critical network. Operators can continue to use the familiar API, automation, and operations and maintenance methods from fabric to the rack, without the need to build and maintain a separate management environment. Customers can now deploy a simpler, programmable management network that can scale with the AI infrastructure and remain available when needed most.

Cisco Sil Token Issuance n One: The most scalable and programmable unified network architecture in the industry

Cisco states that the Sil token issuance n One represents the most scalable and programmable unified network architecture in the industry, covering application scenarios such as AI, ultra-large-scale clouds, data centers, enterprises, and service providers. This architecture was introduced in 2019 and has already played a key role in multiple large networks around the world. With today's innovations, the Sil token issuance n One has proven its ability to scale from 100G to 100T with a single, unified chip architecture.

Supported Resources:

  • Executive Blog: Cisco and NVIDIA Announce USFP : Expanding the Path to AI Connectivity
  • Executive Blog: SONiC's Next Chapter with Cisco : Open Choice, Production Confidence

About Cisco

Cisco (NASDAQ : CSCO) is a global technology leader that is revolutionizing the way organizations connect and are protected in the AI era. For over 40 years, Cisco has been securely connecting the world. With its industry-leading AI driven solutions and services, Cisco helps customers, partners, and communities unleash innovation, enhance productivity, and increase digital resilience. Centered around its mission, Cisco is committed to creating a more connected and inclusive future. For more information, visit The Newsroom, and follow @Cisco on X.

Tip
$0
Like
0
Save
0
Views 17
WalletJYS reminds readers to view blockchain rationally, stay aware of risks, and beware of virtual token issuance and speculation. All content on this site represents market information or related viewpoints only and does not constitute any form of investment advice. If you find sensitive content, please click“Report”,and we will handle it promptly。
Submit
Comment 0
Hot
Latest
No comments yet. Be the first!
Related
This AI startup claims to be able to generate approval-ready blueprints within minutes
The founder of Blueprints AI, Al Drams, stated that the company can use AI to generate construction blueprints that meet specifications and are ready for approval within minutes, based on two-dimensional sketches or PDF. The company will showcase the second-generation software at TechCrunch Disrupt and claimed that it will not replace certified architects.
TechCrunch
·2026-10-12 23:40:49
7
AMD Launches the New Generation Ryzen AI Halo Developer Platform, Upgrading to Max + PRO with 495MB to 192GB of Memory
AMD Today, local time, a new generation of Ryzen AI Halo developer platform based on the Ryzen AI Max + PRO 495 processor was released. The new product has improvements in performance across CPU, GPU, and NPU; it supports up to 192GB of LPDDR5X-8533 memory, of which a maximum of 160GB can be allocated as shared video memory. The Linux version is priced at $7,299.99, and a version pre-installed with Microsoft Project Zenith will be launched later on.
The Block
·2026-10-12 23:40:45
6
ASUS released the XA P2N - E2 in OCP 2026, equipped with NVIDIA Vera CPU, targeting Agentic AI workloads.
ASUS launches XA P2N - E2 servers on OCP Global Summit 2026, equipped with NVIDIA Vera CPU, targeting agentic AI and reinforcement learning workloads. It also showcases a broader AI infrastructure product portfolio that covers rack-level AI systems, networking, storage, and cooling.
PR Newswire
·2026-10-12 23:40:43
7
Wells Fargo downgrades rating of Adobe from 'Overweight' to 'Underweight', targets price cut to $225, concerned that competition from AI erodes core business
Wells Fargo downgrades rating of Adobe from 'Overweight' to 'Underweight', targets price reduced from $270 to $225, stating that the company has shown growth pressures in the third quarter, and that net new ARR in the fourth quarter needs to nearly double compared to the third quarter to meet annual targets; at the same time, intensified AI competition, lack of price hikes, and management transition also increase the uncertainty of the 2027 fiscal year guidance.
The Block
·2026-10-12 23:31:19
11
AMD releases a AI-optimized Helion rack version that is purely CPU and can be scaled up to 32,768 cores.
AMD announced today local time the launch of a Helion rack solution for proxy-based artificial intelligence computing loads. This rack is designed based on OCP design specifications and the open platform concept of AMD Helios rack solutions. Partners and customers can flexibly combine a variety of CPU and GPU computing resources according to their own needs. The pure CPU version, based on the sixth-generation EPYC processor, can be scaled up to 32,768 cores. AMD plans to contribute the AMD Helion specifications to OCP, and states that it is working with ecosystem partners to develop the Helion rack, which is scheduled to be officially launched in November 2026.
The Block
·2026-10-12 23:31:18
11
View More