Test Platforms Target 1.6T AI Data Center Networks

Keysight Technologies will demonstrate testing and validation technologies for high-speed AI data center infrastructure at the Optical Fiber Communication Conference (OFC) 2026. The demonstrations focus on enabling the development and validation of 1.6-terabit Ethernet networks used in large-scale AI computing environments.

Using the KAI Data Center Builder platform, engineers can emulate AI workloads to evaluate how networking hardware, optical interconnects and compute systems behave under realistic operating conditions.

Photonic Integrated Circuit Testing
One demonstration highlights verification methods for photonic integrated circuits used in optical networking components. Keysight systems enable testing of optical devices such as micro-ring modulators (MRM), Mach–Zehnder modulators (MZM) and photodetectors (PD).

The platform supports testing frequencies beyond 170 GHz and includes automated workflows for radio-frequency and photonic characterization. Engineers can also evaluate bias conditions and thermal effects that influence optical device performance.

Electro-Optical Validation from Device to System
Keysight will also demonstrate a unified electro-optical (EOE) validation workflow that links device-level measurements to system-level network performance. The workflow correlates electrical waveform analysis, optical signal quality metrics and end-to-end link performance.

By combining these measurements in a single validation process, engineers can identify integration issues earlier in the development cycle and reduce debugging time during system deployment.

High-Speed Interconnect Testing
To support emerging AI data center architectures, Keysight will present validation tools for 224G interconnect technologies, including Linear Pluggable Optics (LPO) and retimer-less linear interfaces.

These solutions emulate reference receivers and validate signal integrity across high-speed optical and electrical interfaces used in AI computing clusters.

Additional demonstrations will show optical networking research targeting future 3.2-terabit networks. The platform can generate 224-GBaud PAM-4 signals and perform measurements with bandwidths up to 120 GHz, supporting research into next-generation optical transmission systems.

1.6T Optical Networking for AI Workloads
Keysight will demonstrate validation methods for 1.6-terabit optical transmitters used in hyperscale data centers. These testing platforms aim to reduce measurement complexity while increasing throughput in large-scale optical device validation environments.

The company will also present interoperability testing for Ultra Ethernet networking technologies, including LLR and CBFC mechanisms designed for deterministic AI cluster communication. These tests support both 112G and 224G serializer-deserializer (SerDes) interfaces used in high-density networking equipment.

AI Workload Emulation and Infrastructure Optimization
The KAI Data Center Builder platform enables engineers to emulate AI workloads across simulated networking infrastructure. This approach allows system architects to evaluate how compute workloads interact with networking hardware and optical interconnects.

By combining workload emulation with hardware validation, engineers can benchmark network performance, identify bottlenecks and optimize AI data center architectures before deployment.

Edited by Industrial Journalist, Romila DSilva – AI Powered

www.keysight.com