Photonic Fabric™ for Optical Interconnect Networks in AI
The unprecedented increase in bandwidth requirements for data movement on- and off-chip in response to exponentially increasing Artificial Intelligence (AI) model size and complexity cannot be adequately addressed by traditional electronics-based interconnects, nor can the optical access to the chip be limited to its perimeter. To fuel the next generation of AI-ready hardware, direct access to the point of compute is paramount, along with flexible resource allocation enabled by compute and memory resource pooling, which can be efficiently achieved only by using compact and temperature stable photonic technology.
Authors
- Trung Diep
- Prabhat Mishra (ORCID: https://orcid.org/0000-0003-3653-6221)
- Phil Winterbottom
- Sujit H. Ramachandra (ORCID: https://orcid.org/0000-0002-3306-5893)
- David Lazovsky
- Saurabh Vats
- Subal Sahni
- Nadav Bergstein
- Jie Ding (ORCID: https://orcid.org/0000-0002-6036-7301)
- Angelina Totović (ORCID: https://orcid.org/0000-0003-0267-7368)
- Matteo Staffaroni (ORCID: https://orcid.org/0000-0002-9538-0960)
- Andrew Gimlett
- Shifeng Yu
- Ankur Aggarwal (ORCID: https://orcid.org/0009-0003-6605-8758)
- Romanas Narevich
- Waleed Younis
- Ravi Mahatme
Institutions
- Santa Clara University (US)
Publication Details
- Published
- 2025-06-23
- DOI
- https://doi.org/10.1109/cleo/europe-eqec65582.2025.11109692
- Primary Topic
- Photonic and Optical Devices
- Type
- article
- Field-Weighted Citation Impact
- 0.00