Tuesday, 01 April,
16:00
–
17:00
Theater II
Session Description:
A new generation of optical (and electrical) interconnects has been motivated by the recent explosive interest in AI, particularly LLM compute applications, which has dominated the data-center investment since 2023. The expected growth (reportedly 5-10x every 2 years) in the AI infrastructure requires, over the coming years, the related "backend" networks to scale much faster than previous data-center infrastructure through a combination of faster chips (GPU, TPU, NPU, etc) and interconnects (Ultra Ethernet, Infiband, NVlink, UAlink, PICe, UCIe, etc). The OFC 2025 IEEE panel will aim to review the most critical enabling technologies for this next-generation AI/ML interconnects and, more specifically, focus on debating, beyond the hype, the emerging photonics ecosystem and the role of current and future IEEE technologies (e.g., 200 PAM4 pluggable) as well as other proposed innovations (e.g., "slower and wider" optical chiplets).
Organized by:
Panelists
Ram Huggahalli, Microsoft, United States
Ashkan Seyedi, NVIDIA, United States