Session

DescriptionAI systems are rapidly outgrowing the capabilities of purely electronic interconnects and compute substrates. As models scale toward multi-chiplet and ultra-large distributed systems, the energy, latency, and bandwidth limits of electrical signaling have become critical bottlenecks. Integrated photonics, with its ultra-high bandwidth, low latency, and energy-efficient communication, is essential for sustaining AI's growth. However, realizing the full potential of large-scale Electronic-Photonic Integrated Chiplets (EPICs) requires a unified and mature Electronic-Photonic Design Automation (EPDA) ecosystem. Currently, EPDA is fragmented and lacks the mature co-design, co-simulation, auto-layout, verification, and design-for-manufacturability (DFM) infrastructure that enabled VLSI's growth. Establishing EPDA as a first-class design paradigm is both urgent and foundational.

This special session brings together leading researchers and industry innovators to explore the emerging EPDA landscape from design to silicon. The session will feature three key talks:

1. Lightmatter will present advanced electro-optic co-simulation frameworks for wafer-scale photonic fabrics, demonstrating how tightly co-designed electronics and photonics can overcome bandwidth and energy bottlenecks in next-generation multi-chiplet AI systems.

2. Prof. Jiaqi Gu (ASU) will discuss breakthroughs in automated placement/routing and AI-driven design-for-manufacturability for photonic inverse design, addressing scalability and yield challenges in EPIC layout synthesis.

3. AIM Photonics will offer a foundry-level perspective on heterogeneous integration, advanced packaging, and PDK evolution, focusing on process-design co-optimization and ecosystem development for manufacturable EPIC technologies.

These talks highlight photonics' rising importance in heterogeneous, chiplet-based AI systems and the transformative role of EPDA in enabling scalable, fabrication-ready EPICs. This session is timely for DAC and aims to catalyze deeper co-evolution across academia, industry, and foundries.