AI Computing Center
The objective is to construct an all-optical, high-speed network spanning both inter- and intra-data center domains. This architecture encompasses cross-center interconnects, internal cluster networking, and the deep convergence of both, enabling the sharing of computing power and network resources. Based on compute-network collaboration, the network facilitates the flexible interconnection of local and remote resources and dynamically builds distributed clusters across multiple locations. It addresses critical requirements for ultra-broadband transmission, lossless low-latency communication, and agile computing resource scheduling for various workloads, including inference and training.
UX IC's transceiver chips, working in synergy with optical chips, serve as the "Data High-Speed Rail" within AI computing clusters. They provide high-speed, low-latency electro-optical conversion and data transmission between compute nodes, underpinning the collaborative scheduling of large-scale parallel computing. Furthermore, these solutions are deployed for data transmission between AI computing clusters, ensuring seamless scalability.
* The list includes products under development.
