Photonics for AI: Challenges and Applications

Artificial intelligence (AI), machine learning, and edge computing are driving unprecedented demand for high-performance, energy-efficient computing hardware. However, electronic accelerators face critical limitations in speed, scalability, and energy consumption as transistor scaling slows down. Neuromorphic photonics — the use of photonic circuits and systems to emulate neural architectures — has emerged as a compelling approach to overcome these barriers by harnessing the inherent parallelism, high bandwidth, and ultrafast response of light.

To fully realize the potential of neuromorphic photonic architectures, breakthroughs are required across the underlying silicon photonic platforms. Recent advances in silicon photonic devices (e.g., MEMS), heterogeneous integration, and novel optical materials such as barium titanate (BTO) and two-dimensional materials (e.g., MoS₂, graphene) are paving the way toward this goal. These innovations are enabling compact, CMOS-compatible photonic computing systems that can deliver high-speed, low-latency, and energy-efficient information processing — driving next-generation capabilities in AI acceleration, signal processing, sensing, and LIDAR.

This workshop aims to bring together experts from academia, research institutes, and industry across the world to explore current challenges, technological enablers, and future directions in neuromorphic photonics. It will also serve as a platform to identify opportunities for deeper cooperation in the next generation of computing and communication systems.

Key topics of this workshop include:

Technology enablers

• Emerging photonic device technologies for neuromorphic architectures
• MEMS and phase-change devices for reconfigurable photonic systems
• Heterogeneous integration of InP with Silicon Photonics platforms
• Novel materials: Barium Titanate (BTO), SiOC, MoS₂, and graphene-based modulators and detectors
• Advances in on-chip photonic interconnects

Architectures and applications

• Architectures and applications
• Photonic implementations of spiking neural networks and analog optical computing
• Optical neural network accelerators for AI inference and training
• Photonic LIDAR and signal processing systems leveraging neuromorphic architectures
• Cross-layer optimization and co-design between algorithms, hardware, and photonic circuits

Workshop Organizers

• Miltiadis Moralis-Pegios, Aristotle University of Thessaloniki (AUTH), Greece
• Kyoungsik Yu, Korea Advanced Institute of Science and Technology (KAIST), Republic of Korea
• Ruud Oldenbeuving, IMEC, Netherlands

Speakers

Hollow-Core Fiber Technology: Recent Advances, Challenges, and Future Directions

Hollow core fiber (HCF) technology has progressed from a laboratory curiosity to a strong candidate for next-generation optical networks and advanced photonic systems. Recent breakthroughs in HCF designs have achieved record-low losses (<0.1 dB/km), well below conventional solid-core silica fibers, while maintaining ultra-low latency and greatly reduced nonlinear impairments. Combined with growing industrial investment and early field deployments in cloud and data-center networks, these advances suggest that HCF is approaching a tipping point towards wider commercial adoption.

At the same time, HCF are enabling transformative capabilities across multiple application domains. In quantum technologies, they provide low-noise, low-latency channels for distributing quantum and classical signals. In high-power and ultrafast photonics, HCFs supports kilowatt-class single-mode transmission at 1 µm and beyond. Gas-filled HCFs underpin the rapidly growing areas of “gas photonics”, enabling long-interaction-length nonlinear optics and highly sensitive gas sensing and spectroscopy.

Despite this rapid progress, several key technological challenges remain: scalable, low-cost fabrication; robust cabling and splicing; integration with conventional fiber infrastructure and photonic components; and long-term reliability and standardization. This workshop will bring together leading experts from industry and academia to: Review the latest breakthroughs in HCF design, fabrication, and system integration. Identify common technology bottlenecks and cross-cutting solutions across applications. Formulate a coherent vision for the next 5-10 years of HCF research, standardization, and commercial deployment.

Participants will gain a concise overview of the state of the art, exposure to real deployment experiences, and insights where their own research and product roadmaps can intersect with emerging HCF capabilities. The workshop format is explicitly designed to foster open, interactive discussion between fiber manufacturers, system vendors, and end-users.

Workshop Organizers

• Yongmin Jung, Optoelectronics Research Centre, University of Southampton, UK
• Stephanos Yerolatsitis, University of Bath, UK

Intelligent Optical Networks: From Optical Sensing to AI

As digital services continue to expand and user expectations for high-speed and highly reliable connectivity increase, optical networks are required to evolve beyond conventional data transmission. Future networks must be capable of perceiving their own operational conditions, interpreting network states, and responding autonomously in real time.

This workshop aims to explore how AI can enable intelligent decision-making and automated control in optical networks. By utilizing sensing data collected from the network, AI techniques can detect anomalous traffic behavior, identify performance degradation, and anticipate potential failures. Technologies such as digital twins, predictive analytics, and closed-loop control are discussed as key enablers for proactive network operation, improved service quality, and reduced operational complexity.

This workshop also aims to investigate the role of optical sensing and signal monitoring in enhancing network awareness. Fiber sensing, optical performance monitoring, and in-home service quality measurements provide valuable insights into physical impairments, including signal loss, fiber bending, and component aging, as well as end-user experience. AI-based models are essential for interpreting these heterogeneous sensing data, enabling early fault detection, risk prediction, and more accurate assessment of service quality.

Experts from telecom operators, equipment manufacturers, research institutes, and universities will join this workshop. Together, they will explore what kinds of sensing data, AI models, and operational methods are needed for next-generation optical networks. The goal is to better understand how AI and sensing can make networks more reliable, more efficient, and easier to manage—and to identify key research questions for moving the technology forward.

Workshop Organizers

• HanHyub Lee, ETRI, Republic of Korea
• Shen Shikui, China Unicom, China
• Noboru Yoshikane, KDDI Research, Japan
• Luca Valcarenghi, Scuola Superiore Sant'Anna, Italy

Speakers

Baud-Rate Scaling in Optical Transmission: Technical Limits and Practical Challenges

Recent years have seen a rapid evolution in optical transmission capacity, striving toward 1.6 Tbps and beyond for both IMDD and digital coherent technologies. As physical bandwidth becomes a precious resource, industry and standardization efforts have shifted focus toward higher baud-rate operation as the primary lever for capacity growth.

However, as symbol rates continue to climb, we face a critical intersection of fundamental and practical bottlenecks. These include increased sensitivity to component bandwidth, implementation penalties, and signal integrity degradation—all of which become significantly more pronounced at ultra-high baud rates. These challenges complicate not only performance optimization but also the realization of scalable, robust, and cost-effective optical interfaces.

This workshop provides a forum to explore the boundaries of baud-rate scaling. We will address key questions:

• How far can we push symbol rates with emerging device and DSP technologies?
• What are the critical trade-offs between performance, power efficiency, and implementation complexity?
• How will these constraints shape future system architectures and standardization?

By bridging perspectives from device physics, DSP design, and system architecture, this workshop aims to define the roadmap for next-generation high-capacity optical transmission.

Workshop Organizers

• Asuka Matsushita, NTT, Japan
• Sun Hyok Chang, ETRI, Republic of Korea

Speakers