Keysight Showcases AI Infrastructure Validation Advances Across Scale-Up, Scale-Out, Scale-Across Networks to Inference at 2026 OCP

New solutions, large-scale AI workload emulation, multi-vendor interoperability, and end-to-end inference validation demonstrations on show to help accelerate next-generation AI infrastructure

SANTA ROSA, Calif.--(BUSINESS WIRE)--Keysight Technologies, Inc. (NYSE: KEYS) will showcase a broad portfolio of AI infrastructure validation solutions at the 2026 OCP Global Summit in San Jose, October 12 – 15. Working with leading industry partners at the event, Keysight will demonstrate how it is helping technology developers, infrastructure providers, hyperscalers, and operators validate increasingly complex AI systems under realistic conditions before deployment.

At booth #C83, Keysight will show its latest innovations for validating rapidly evolving AI infrastructure, as the industry moves toward 224G SerDes and 1.6T Ethernet interconnects, open scale-up connectivity, larger scale-out fabrics, scale-across architectures spanning data centers, and increasingly demanding inference workloads. Keysight industry collaborations at OCP 2026 will include joint demonstrations with Alibaba Cloud, Cisco, HPE, and NVIDIA.

Ram Periakaruppan, Vice President and General Manager, Network Test & Security Solutions, Keysight, said: “AI infrastructure is evolving from individual components into highly interconnected systems spanning Scale-Up, Scale-Out, Scale-Across, and production inference. At 2026 OCP, we will show how Keysight is helping the industry recreate realistic AI environments before deployment so engineering teams can understand system behavior at scale, identify bottlenecks earlier, and accelerate the transition across design to deployment continuum.”

Keysight highlights at 2026 OCP will include:

• Two new 1.6T Ethernet test platforms: Keysight will show the SERT 1600GE data center switch and edge router test solution, and the AresONE Plus 1600GE platform, the industry’s first to deliver eight 1.6T Ethernet ports in a single 2U appliance and provide stateful, realistic AI workload generation with mix-of-collectives emulation.
• Open Ethernet Scale-Up interoperability: A demonstration of multi-vendor interoperability of emerging Ethernet Scale-Up technologies, with the first live public demonstration of Link Layer Retry (LLR) running over 200G SerDes-based 1.6T speed.
• Scale-Across validation: Demonstrating resiliency and performance validation of RoCEv2 AI workloads across Scale-Across networking, enabling engineering teams to evaluate how latency, congestion, buffering, packet loss, and resiliency affect workload-level performance as AI infrastructure extends across data centers.
• Large scale 64x800GE Mix of Collectives emulation: Keysight will emulate 64x800GE NIC and GPU endpoints driving a mix of AI collective workloads across a 2x2 CLOS fabric. The joint showcase enables developers to evaluate fabric utilization, congestion, load distribution, and workload performance under more representative AI traffic before deploying large-scale GPU infrastructure.
• Scale-Out validation with MRC/SRv6: Keysight will demonstrate multi-path MRC traffic using SRv6 uSID-based source routing, showcasing resilient path and plane failover, congestion mitigation through packet trimming, and sustained application performance.
• Inference scale and performance: Keysight AI Inference Builder will generate realistic inference workloads against a live inference stack while correlating workload performance with compute, GPU, memory, and network telemetry.

Keysight will collaborate with industry leaders to jointly demonstrate how realistic workload emulation and system-level validation can help the industry identify issues sooner, evaluate interoperability, and gain confidence in AI infrastructure before deployment.

Will Eatherton, SVP Engineering, Data Center and Internet Infrastructure, Cisco, said: “As AI clusters push beyond the power and physical limits of a single data center, scale-across is becoming a critical new dimension of AI networking. This is more than traditional data center interconnect: distributed AI workloads require massive bandwidth, resiliency, and the ability to absorb traffic bursts without compromising application performance. Cisco Silicon One P200 is purpose-built for this transition, combining 51.2 Tbps of bandwidth with deep buffering and Cisco 800G ZR+ coherent pluggable optics help deliver predictable performance as AI workloads extend across data centers. In addition to using real GPU server clusters, our collaboration with Keysight helps validate these architectures under realistic RoCEv2 workloads, congestion, and network conditions. In parallel, a 64x800GE mixed-collective workload demonstration with Cisco Nexus One brings the same workload-level validation to scale-out fabrics, helping customers understand how their AI networks will perform before deploying GPU infrastructure at scale.”

Throughout the conference, Keysight experts will participate in 2026 OCP speaker sessions, covering AI infrastructure validation, scale-up networking, digital twins, switching, and accelerator validation. Featured speaking sessions include:

• AI Infrastructure Validation at Scale - Ram Periakaruppan, Keysight. Monday, October 12, 12:00pm PT | SJCC Concourse Level, Room 211.
• Beyond Physical Digital Twins: Why Logical Digital Twins Are Essential for AI Factory Design, Validation, and Operations - Abhijit Lahiri, Keysight, and Avi Alkobi, NVIDIA. Tuesday, October 13, 2:05pm PT | Concourse Level, Room 220B.
• Tier 1 Ranking Test: A Vendor-Neutral Framework for Benchmarking Switching ASICs via SAI in SONiC - Manodipto Ghose, Keysight, and Shenglin Zhu, Alibaba. Wednesday, October 14, 8:40am PT | Concourse Level, Room 230A.
• Validation Methodologies for ESUN Ecosystem - Manoj Wadekar, Meta, and Venkat Pullela, Keysight. Wednesday, October 14, 8:50am PT | Concourse Level, Room 220B.
• Standardizing Ethernet Scale-Up Validation: Test Strategy and Current Progress on LLR/CBFC - Haiyang Zheng, Alibaba Cloud, and Yuanwen (Daisy) Sun, Keysight. Thursday, October 15, 9:45am PT | Concourse Level, Room 210AE.
• Cluster-less Validation of AI Accelerator Silicon Under Real Workloads - Matt Bergeron, Meta; Ashutosh Kumar and Cosmin-Nicolae Prunaru, Keysight Technologies. Thursday, October 15, 12:55pm PT | Concourse Level, Room 220B.

For more information about Keysight technology showcases, joint partner demonstrations, and other activities at 2026 OCP Global Summit, visit Keysight at OCP.

Resources

• AI Infrastructure Month at Keysight

About Keysight Technologies

Keysight (NYSE: KEYS) serves technology innovators as a mission-critical design enablement partner for the world’s most complex engineering challenges. By connecting market-leading design, emulation, and test solutions across the full life cycle, Keysight helps engineering teams accelerate innovation, reduce risk, and bring new technologies to market faster. Customers across AI infrastructure, communications, industrial automation, aerospace and defense, automotive, semiconductor, and general electronics rely on Keysight to bridge virtual design and physical reality, enabling confident decisions earlier. Learn more at www.keysight.com.

Keysight Media Contacts

Mark Price
Americas
mark.price@keysight.com

Fusako Dohi
Asia
fusako_dohi@keysight.com

Jenny Gallacher
Europe
jenny.gallacher@keysight.com