#prefill-and-decode

#prefill-and-decode

AWS and Cerebras are disaggregating AI inference into prefill and decode components, with AWS Trainium optimized for prefill processing and Cerebras wafer-scale chips excelling at decoding.