09 Mar 2027
14:15 - 15:45
Expo Theater I
The rapid growth of large-scale AI is driving a shift toward distributed, scale-across data-center architectures, where workloads span multiple sites to balance compute, power, and cooling constraints. This evolution introduces stringent requirements on latency, bandwidth, and synchronization, especially for distributed training. This panel will explore recent advances in data-center interconnect (DCI) architectures, transport protocols, and optical technologies, including emerging fiber solutions such as hollow-core and multi-core fibers. The discussion will bring a system-level perspective on enabling efficient, low-latency AI infrastructure beyond a single data center.