Live Webinar, 10 June, 3pm - AI Training Clusters: Building the Optical Backbone
Webcast supported by
Date & Time: Wednesday, June 10, 2026, 3:00 PM BST
Duration: 60 minutes
Cost: Free
Contact for your FREE certificate of attendance: production@europascience.com
The race to build 100,000-GPU training clusters is exposing the optical interconnect as a critical variable in AI infrastructure performance. As model complexity and dataset scale push cluster sizes to unprecedented levels, the demands placed on the optical backbone—in terms of bandwidth density, latency, topology and failure resilience—have outgrown conventional data centre network design.
Every architectural decision, from fabric topology to transceiver selection to cabling strategy, now carries system-level consequences. Rail-optimised topologies, the InfiniBand versus Ethernet debate, and the evolving role of active optical cables are reshaping how network architects approach the physical layer, and the stakes of getting it wrong have never been higher.
With hyperscalers committing billions to AI-ready infrastructure, the industry needs a clear-eyed view of what a training cluster actually demands from its optical interconnect, and how to build a backbone that can deliver it at scale.
What You'll Learn
✓ How AFL's fibre solutions can address the unique demands of GPU cluster interconnect
✓ Physical layer architecture for 100,000+ GPU training clusters
✓ Cabling strategies: density, routing and serviceability at hyperscale
✓ Transceiver selection and optical component performance requirements
✓ Rail-optimised topologies and their implications for infrastructure design
✓ InfiniBand versus Ethernet: implications for fibre connectivity
✓ Active optical cables and their role in training cluster performance
✓ Building resilient optical backbones for mission-critical AI infrastructure
✓ Future-proofing interconnect design as AI cluster sizes continue to evolve
Who Should Attend?
This webcast is designed for:
- Network architects designing AI fabric infrastructure
- Infrastructure engineers specifying interconnect components
- Data centre planners evaluating cluster topologies
- Optical connectivity solution providers
- Systems engineers deploying AI training infrastructure
- Technology decision-makers in hyperscale environments
- Anyone seeking to understand how GPU cluster growth reshapes optical connectivity requirements
Speaker
Antonio Castano, Market Development Director, Global Sales, AFL
Antonio joined AFL in 2025, bringing more than 15 years of international experience across data centres, carrier networks and connectivity. He has a proven track record of success in technical sales, product management and business development, and now leads a talented global team delivering innovative fibre solutions. Known for strategic leadership, ownership and cross-functional execution, Antonio blends operational strategy and market development to turn complex challenges into customer value.
Why This Webinar Matters
The optical backbone is no longer a commodity layer in AI infrastructure. As hyperscalers commit billions to 100,000+ GPU training clusters, every decision about fibre, connectors, topology and cabling directly impacts model training speed, cost per GPU, and system reliability.
This webcast provides the technical depth and practical perspective needed to understand these trade-offs and inform infrastructure decisions.
With thanks to our sponsor
