VIAVI Solutions Inc. announced on June 23, 2026, the launch of its new Ultra Ethernet Transport (UET) validation solution for AI fabrics. This offering, integrated into the VIAVI TestCenter platform, brings traffic generation and analysis capabilities to hyperscalers, cloud and neocloud providers, and network equipment manufacturers participating in the Ultra Ethernet ecosystem.

The Ultra Ethernet Stack was purpose-built for large-scale AI and High-Performance Computing workloads. The UEC 1.0 specification introduces a new UET protocol designed to deliver advanced congestion control and massive scalability, accelerating multi-vendor adoption of Ethernet as the primary transport for AI fabrics.

VIAVI's GPU-free solution emulates the transport layer of UET, replicating realistic, stateful AI traffic patterns at scale without requiring dedicated GPU infrastructure. The platform supports reliable ordered and unordered delivery, packet trimming, congestion control, and dynamic multipathing. It provides full-fidelity emulation of AI workloads including collective communications and large language model flows, while enabling comprehensive validation of load-balancing mechanisms such as ECMP, packet spraying, and flowlet switching.

Aniket Khosla, Vice President of Product Management for Optical Transport and High-Speed Ethernet at VIAVI, stated that as AI clusters scale to millions of endpoints, relying on physical GPUs alone for network validation is no longer practical. He emphasized that rigorous, realistic testing is essential to the success of Ultra Ethernet, and that the new solution enables customers to deploy scalable, high-performance AI fabrics faster and more cost-effectively.

Mahesh Subramaniam, Senior Director of Product Management for AI Data Centers at HPE, noted that as AI data center fabrics grow in size and complexity, traditional validation approaches struggle to keep pace. He highlighted that VIAVI TestCenter's realistic traffic emulation and granular visibility enable faster deployment and reliable scaling of Ultra Ethernet, citing HPE's collaboration with Juniper using the QFX5240 platform and Junos Evolved.