11 Mar 2027
11:45 - 12:45
Expo Theater III
There have been unprecedented AI compute demands for optical interconnect products especially high-speed optical transceivers for recently years. With computing power doubling every 3-4 months, AI scaling advances much faster than before and may require our industry to evolve at a much faster pace. While AI brings excitement and opportunities to our red hot industry, there are many challenges to come as well. Can optical interconnects keep up the speed with AI scaling? Can existing technologies move faster enough to meet emerging AI requirements? Are there any disruptive technologies required? Do we need to change our way to develop new technologies and products? This panel will bring together industry influential experts from various segments including hyperscale end users, system vendors, optical transceiver and chip companies to extensively address those questions.
The intention of this panel is to have end-users to speak loud, what new optical interface like CPO, NPO, XPO, CPX, LPO, LRO etc is likely going to take off after n+1 years? What are the key requirements and deployment issues to tackle with? How should vendors improve to address those issues? with various new innovative products ready, what else vendors need to address to keep end-users happy enough to issue PO?
Key questions:
- AI is exploding rapidly, does this also drive Optics to continuously explore and break through its technical boundaries and bottlenecks, such as power consumption, cost, reliability, etc.?
- Which optical technology solution will really accelerate the development of large-scale AI clusters, CPOs, NPOs, XPOs, LPOs, pluggable modules or beyond both?
- 800G has been deployed massively with approaching 100Mpcs mark annually for 2027, and pluggable modules have been focused on transition to 1.6T taking off. What is the progress of technologies such as 400G/Lane electrical interface rate? how will these technologies play a leading role in scaling AI data centers further?
- Both CPO, NPO, XPO and LPO/LRO have their own advantages, but also face various problems. For example, the reliability of CPO needs to consider both testability and serviceability, while the performance of LPO depends too much on the adaptation with the master chip. How to resolve the performance differences of different master chips and different channels?
- With so many MSAs involved for industry, when CPO/NPO, XPO, CPX vs. LPO/LRO are going to meaningfully impact Data Center Optical Interconnect?
Presented By:
-
Frank Chang
Chief Technology Officer, Source Photonics, United States
-
Scott Wilkinson
Cignal AI, United States
-
Rui Lu
Alibaba Group , China
-
Die Wan
Baidu, China
-
Nan Ge
Meta, United States
-
Sasha Gazman
Microsoft, United States
-
Eric Bernier
Huawei, Canada