Break the Bottleneck: High-capacity Optics for Next-Gen GPU Clusters
(Image: Shutterstock/Shutterstock AI)
Webcast supported by
In association with
AI model training is outpacing network capacity at an unprecedented rate. As GPU clusters scale toward tens of thousands of accelerators, the optical layer has become the primary constraint on system performance and operational efficiency.
Traditional network architectures designed for general-purpose data centre traffic cannot meet the bandwidth density, power efficiency, and fabric-level scalability demands of AI workloads.
With hyperscalers investing billions to build AI-ready infrastructure, the industry is racing toward 1.6T OSFP-XD transceivers, co-packaged optics, and fundamentally new optical switching paradigms—but the path forward raises critical questions about interoperability, power consumption, and architectural trade-offs.
Whether you're directly involved in AI infrastructure deployment or need to understand how these architectural shifts will impact the broader optical networking industry, this webcast provides the technical depth and strategic perspective to inform your decision-making.
Calling All:
- Data centre network architects planning optical infrastructure upgrades to support AI training clusters and seeking guidance on 1.6T transceiver deployment strategies
- Product managers and engineers at optical component manufacturers, hyperscalers, and systems vendors developing or evaluating next-generation interconnect technologies
- CTO office and infrastructure executives at cloud providers, telecommunications operators, and enterprises assessing technology roadmaps and capital investment priorities for AI-ready networks
- Optical systems designers working on co-packaged optics, silicon photonics integration, and laser source architectures for high-bandwidth applications
- Network operations teams responsible for deploying and maintaining optical infrastructure in large-scale data centre environments where power efficiency and density are critical constraints
We'll cover:

- Mastering the 1.6T technology landscape: Understand the hardware specifications, interoperability challenges, and deployment timelines for next-generation high-speed pluggable transceivers designed specifically for GPU cluster connectivity

- Decoding optical architecture trade-offs: Learn how centralised versus external laser source architectures impact power efficiency, thermal management, and overall system cost in AI-scale fabrics

Optimising for fabric-level efficiency: Discover how optical switching technologies can improve power consumption, network resiliency, and scalability across large AI training environments

- Leveraging hyperscale deployment experience: Gain practical insights drawn from Lumentum's experience delivering photonic technologies at volume to the world's largest data centre operators

- Preparing for architectural shifts: Understand how AI workload requirements are fundamentally changing optical network design principles and what this means for procurement, deployment, and operational strategies

- Navigating interoperability requirements: Understand the technical performance characteristics: stable light sources, efficient modulation, power budgets; required to support AI-scale networks and ensure vendor ecosystem compatibility
Speakers:

Christian Urricariet
Senior Director of Product and Strategic Marketing, Lumentum
Christian Urricariet leads product strategy and market development for Lumentum's data centre and AI connectivity solutions. With deep expertise in optical transceiver technologies, silicon photonics, and hyperscale network architectures, Christian has been at the forefront of defining next-generation optical interconnect roadmaps.
His work spans high-speed pluggable transceivers, co-packaged optics integration, and optical switching technologies designed to meet the performance and efficiency demands of AI-scale data centres. At Lumentum, Christian works directly with hyperscale operators and systems vendors to translate emerging AI workload requirements into deployable photonic solutions.
Having trouble registering? Please get in touch at webcasts@europascience.com.
Interesting in sponsoring this webcast? Get in touch with kris.collins@europascience.com.