OFC

How Optical Circuit Switching (OCS) Enables Large-scale GPU Clusters

For AI/ML training clusters, the Lumentum R300 OCS enables optical connectivity across the rack, allowing you to replace or bypass traditional spine switches. This creates dynamic, high‑bandwidth paths for GPU‑to‑GPU communicat...
Apr-23-2026

AI Fabrics Proven: Marvell Infrastructure + Keysight Testing

This AI data center simulation validates 800G/1.6T Ethernet fabrics using realistic AI traffic, congestion patterns, and UEC features—supporting pre‑deployment testing for vendors and hyperscalers. Keysight’s AresONE‑800GE emul...
Apr-16-2026

PCIe-based Scale-up Switching

For GPU scale‑up and high‑performance multi‑GPU systems, the Marvell® Structera‑S PCIe® 6.0 switch delivers PCIe‑based scale‑up with the highest lane count PCIe switch in the industry. With an industry‑leading 256‑lane radix, S...
Apr-06-2026

Extending In‑rack Copper for AI: A Hybrid 200G ACC/AEC Solution

Extending in‑rack copper reach is critical for scaling AI data center infrastructure—without increasing cost or power. This solution combines the simplicity and efficiency of active copper cable (ACC) with the extended reach of...
Apr-07-2026

Powering AI Fabrics: 1.6T Optical Interconnects + Open CPX MSA

For AI/ML cluster interconnects at 800G and 1.6T, extreme bandwidth, low latency, and lower power per bit are achieved using short‑reach optical solutions with 1.6T optical transceivers and 800G Coherent‑Lite O‑band modules. To...
Apr-06-2026

Inside the 1.6T DR8 PIC: Rack‑to‑Rack Silicon Photonics for AI Clusters

For rack‑to‑rack data center connectivity in AI clusters, the OpenLight 1.6T DR8 PIC (8 lanes × 200G = 1.6 Tbps) is a highly integrated silicon photonics engine designed for data‑center‑reach optical interconnects. This photoni...
Apr-06-2026

Marvell Ara T™: 1.6T PAM4 DSP Optimized for AI Optical Interconnects

Next‑generation AI networking fabrics and high‑performance data center interconnects demand extreme bandwidth with lower power and smaller footprints. The Marvell® Ara T™ 1.6T PAM4 DSP with Transmit Retimed Optics (TRO) is desi...
Apr-06-2026

The Hybrid Active Electrical and Copper Cable (AEC + ACC)

The solution integrates seamlessly with the Marvell optical connectivity portfolio and the RELIANT™ interconnect telemetry platform, providing scalable, efficient, and intelligent network fabric control for next‑generation GPU ...
Apr-06-2026

Scaling AI Data Centers with 1.6T & 800G Coherent Pluggable Optics

The rapid growth of generative AI is reshaping data center architecture. As infrastructure evolves from isolated facilities to multi‑site campuses and regional AI clusters, the need for high‑capacity, energy‑efficient data cent...
Apr-06-2026

Photonic Fabric™: Optical Scale‑Up for Next‑Generation AI Clusters

Next‑generation AI clusters demand unprecedented reach, bandwidth, ultra‑low latency, and energy efficiency. Photonic Fabric™ is an optical interconnect purpose‑built for AI scale‑up and memory transactions—delivering lower XPU...
Apr-06-2026